Variability of Ignition Furnace Correction Factors (2017)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

22.1 Background Accurate determination of asphalt content and aggregate gradation is critical in the quality control of asphalt mixtures during construction. Historically, the most common method has been extraction using any of a number of different sol- vents such as trichloroethylene, methylene chloride, or tri- chloroethane. This method has no need of calibration factors and, in general, the properties of the base material, aggregate, and binder can be accurately measured after extraction. In the 1980s, the use of chlorinated solvents for asphalt extractions began to be questioned because of potential health and safety impacts and difficulty of disposal. The test was also expensive since the solvents became more costly and their disposal after use was even more costly. Biodegradable solvents were evalu- ated by several agencies to replace the more hazardous chlo- rinated solvents. Modifications to the extraction procedure were needed to use these biodegradable solvents, and the test time was longer and the test less accurate than the one using chlorinated solvents (1). Nuclear gauge methods were also used to measure asphalt content. This is an indirect method that uses radiation to determine the amount of asphalt by measuring the amount of hydrogen in the asphalt mixture. This method can measure the asphalt content rapidly and reasonably accurately but does not allow the determination of aggregate gradation and properties of the asphalt binder. Because of the issues with existing tests in the early 1990s, alternative procedures to determine asphalt content were needed. This led to the development of the ignition furnace method that was finalized in the mid-1990s. The test was devel- oped at the National Center for Asphalt Technology (NCAT) and has since been adopted by most laboratories around the United States and many around the world to eliminate the use of hazardous solvents needed for extraction testing methods. The ignition furnace method was shown to be more accurate than the solvent extraction test and also reduced overall cost, especially when the environmental, health, and safety benefits were considered (1, 2). The development of the ignition furnace method at NCAT built on the research work conducted at Clemson University by Antrim and Busching as part of NCHRP Project 10-4 (3). In their study, they showed that by heating the asphalt mixture at a high temperature [approximately 1,550°F (843°C)], com- plete combustion of the asphalt could be achieved. They used a butane burner with the addition of oxygen. They also noted that for certain aggregates, significant errors in asphalt content determination could occur as a result of aggregate mass loss. Aggregate mass loss of granite gneiss was insignificant, but lime- stone loss was significant, which resulted in a deviation from true asphalt content of approximately 1%. The NCAT method incorporated modifications of their procedure that included the use of a muffle furnace and lower temperature [1,000°F (538°C)], which significantly reduced the effect of aggregate type on asphalt content determination and testing time. The ignition furnace procedure has been widely used since the mid-1990s. When the test procedure was first devel- oped, there were two general ways it could be conducted. One method involved weighing a sample of asphalt mixture, placing the weighed sample in a furnace, burning the asphalt binder from the mixture, removing the remaining sample from the furnace, and weighing the sample again after cool- ing (external weighing method). The change in mass before and after burning allowed the asphalt content to be deter- mined. At about the same time, ignition furnace equipment was developed with internal balances so that the test could be conducted automatically to determine the asphalt content. This approach reduced the technician time required and likely provided for a more controlled test procedure. Most laboratories currently use this automatic method. The test procedures for both of these approaches are now described in AASHTO T 308 (4) and ASTM D6307 (5). There are some variations between the two standards, but the differences are minimal. For the purpose of this report, AASHTO T 308 will be referred to as the standard procedure for asphalt content determination by ignition method. C h a p t e r 2 Literature Review

3 The current AASHTO T 308 procedure consists of burn- ing the asphalt from laboratory mix samples in an ignition furnace at a high temperature [1,000°F (538°C)]. The test is complete when the change in mass does not exceed 0.01% for 3 consecutive minutes. The asphalt content is calculated from the mass of the HMA sample prior to ignition and the mass of the aggregate remaining after the ignition, using Equation 1. P M M M b i f i % 100 CF MC Equation 1 ( )( ) = − ×   − − where Pb = measured (corrected) asphalt binder content, percent; Mi = total mass of the HMA sample before ignition, g; Mf = total mass of aggregate remaining after ignition, g; CF = the correction factor as a percent of mass of the HMA sample; and MC = moisture content of the HMA if not dried prior to testing. The procedure to obtain a correction factor involves the prep- aration of two samples of the asphalt mixture at the designed asphalt content. The mixture calibration samples are tested, and the calibration factor is calculated as the difference between the actual and measured asphalt binder contents, expressed as a percentage of the HMA mass. The precision criteria to accept test results for asphalt content for single-operator and multi- laboratory evaluation are presented in Table 1. AASHTO T 308 requires that an asphalt correction fac- tor be determined by testing a set of specimens with known asphalt content for each job mix formula. A correction factor needs to be obtained each time a change of greater than 5% in stockpiled aggregate proportions occurs. The test method also requires that additional testing be conducted at a lower tem- perature (900 ± 8°F) if the measured asphalt correction factor exceeds 1.0%. If no improvement in the correction factor is obtained with this reduction in temperature, the test method allows use of the higher temperature for these problematic aggregates. The test method also indicates that an aggregate correction factor is needed for aggregates that have a proven history of excessive breakdown or those from an unknown source. In order to calculate an aggregate correction factor, the residual aggregate (after ignition testing) is compared to a “blank” sample from each sieve size, and the differences are considered to be the correction factors for each sieve size. Allowable differences for each sieve size are provided, and if the difference of any sieve exceeds the allowable percent, an aggregate gradation correction factor equal to the aver- age difference for each sieve size is applied to the gradation test results. Although the procedure is straightforward, errors due to normal variability of results during the test are still possible. As mentioned previously, loss of some aggregate mass dur- ing ignition and breakdown of the aggregate particles due to high temperatures can cause the measured asphalt content to be greater than the actual content and can also cause changes in the gradation of the recovered aggregate. The procedure establishes minimum mass requirements for the test speci- men that depend on the nominal maximum aggregate size (NMAS) of the mixture. This is important because large spec- imens of fine mixtures may result in incomplete burning of the asphalt binder (4). The minimum mass requirements by NMAS are presented in Table 2. The standard also indicates that the specimen sizes should not be more than 500 g greater than the minimum recommended specimen mass. AASHTO T 308 also includes a procedure to develop aggre- gate correction factors. It is required to perform a gradation analysis on the residual aggregate in accordance with AASHTO T 30, “Mechanical Analysis of Extracted Aggregate.” From the gradation results, the percent passing for each sieve for each sample must be subtracted from the percent passing each sieve of a blank specimen gradation (aggregate-only speci- men batched at the job mix formula gradation). The average difference between the two values is calculated. If the differ- ence for any single sieve exceeds the allowable difference for that sieve, as presented in Table 3, then aggregate gradation correction factors (equal to the resultant average differences) for all sieves must be applied to all acceptance gradation test results determined by AASHTO T 30. If the #200 sieve is the Condition Standard Deviation Acceptable Range of Two Test Results Single-operator precision: asphalt content (%) 0.069 0.196 Multi-laboratory precision: asphalt content (%) 0.117 0.33 Table 1. AASHTO T 308 – precision estimate (4). NMAS (mm) Minimum Mass of Specimen (g) 4.75 1,200 9.5 1,200 12.5 1,500 19.0 2,000 25.0 3,000 37.5 4,000 Table 2. Minimum mass requirements by NMAS (4). Sieve Allowable Difference Size larger than or equal to No. 8 ±5.0% Size larger than No. 200 and smaller than No. 8 ±3.0% Size No. 200 and smaller ±0.5% Table 3. Allowable sieving difference (4).

4only sieve outside the limits, the aggregate correction factor must be applied to that sieve only. 2.2 Types of Ignition Furnaces There are essentially two ignition furnace types (based on heating mechanism): convection units and infrared units. Brief descriptions of each unit’s mechanisms are presented in the following paragraphs. 2.2.1 Convection Units In a convection ignition furnace, the furnace chamber is heated using a radiant heat source consisting of an electric heating element enclosed in a refractory ceramic material. The heating element heats the air in the furnace chamber and, hence, heats the sample. The asphalt binder ignites, and a blower pulls air into the chamber to maintain ignition. The released gases are further oxidized while passing out of the main chamber into a secondary chamber at a higher temper- ature of approximately 1,382°F (750°C). This contributes to the reduction of volatiles in the exhaust stream. At this point, the exhaust is cooled by mixing with ambient air. The blower directs the exhaust to the plenum exhaust port, where exhaust tubing leads the exhaust to an outside ventilation system (6). Figure 1 shows an example of a Thermolyne convection unit. The mass loss during ignition is measured using an internal balance. At least one brand does not incorporate an internal balance. For these units, the specimen basket with the asphalt mixture sample needs to be removed from the furnace and weighed a number of times until the change in measured mass of the specimen does not exceed 0.01% of the initial specimen mass. The complete procedure is included in AASHTO T 308 (Test Method B, external balance). 2.2.2 Infrared Units Infrared units use an infrared heating element to heat the sample. This differs from convection heating, where the air in the chamber must be heated first. Infrared heating uses electro- magnetic energy waves to transfer heat energy directly to the sample by stimulating the molecules in the asphalt mixture. The sample then heats the furnace chamber by conduction and convection. Figure 2 shows an example of a Troxler New Technology Oven (NTO) infrared unit. Troxler NTO units have the option to select between three profiles: a default pro- file recommended for most materials; Option 1, designed for soft aggregates such as dolomites (or any mixture with a large correction factor); and Option 2, which is recommended for some mixtures with higher asphalt content, such as stone matrix asphalt and special modified mixtures (7, 8). The pro- files manipulate the heater temperature and fan speed as a Figure 1. Ignition furnace – Thermolyne convection unit with internal balance. Figure 2. Ignition furnace – Troxler NTO infrared unit.

5 function of time. As an example, a typical profile may look like a set of steps of different lengths and amplitudes, where amplitude on the fan would represent speed (air flow), and amplitude on the heater would be the surface temperature of the heater element (personal communication with Troxler personnel, November 2014). These units have an internal bal- ance to monitor the mass loss during ignition. 2.3 Ignition Furnace Studies This section summarizes, in chronological order, past studies related to ignition furnaces for asphalt content determination. Most of these studies were conducted in the mid- to late-1990s and early- to mid-2000s. 2.3.1 Brown, Murphy, Yu, and Mager, 1994 (1) This study documents the development of the ignition fur- nace procedure for the determination of asphalt content. The primary goals were to minimize the overall test time and to produce a test method with acceptable accuracy when com- pared to extraction methods. The authors concluded that the method could be used successfully to determine the asphalt cement (AC) content and aggregate gradation of HMA mix- tures. The authors state that a test temperature of 1,000°F (538°C) was sufficient to remove the asphalt binder but indi- cated that the temperature may have to be lowered for some aggregates to prevent the need for a correction factor. They found that a calibration factor was needed to accurately deter- mine asphalt content for some aggregates and that no cor- rection factor was necessary for aggregate gradation. Using this procedure, they found deviations from the true asphalt content of ±0.3%. Additionally, the effect of variable air flow was investigated in this study. The furnace that was used had the capability to vary the air flow and had scales underneath it to monitor the weight of the sample during the test. Figure 3 shows the effect of two different flow rates, 5 liters per minute (lpm) and 15 lpm, on the test time and mass loss. The tests were con- ducted at 1,000°F (538°C) on mixtures with limestone having 6% asphalt content. The results showed that the time to remove the asphalt binder was approximately 1 h for tests conducted at 5 lpm and approximately 40 to 45 min for tests conducted at 15 lpm. This shows the potential of increased air flow to reduce testing time, which would likely affect the measured correction factor, especially for aggregates with rela- tively high correction factors. The researchers also suggested that more air flow may result in more complete combustion and reduction in generation of smoke. They cautioned that the air flow should not be too high, otherwise it might cause loss of fines from the mixture being tested. From Figure 3 it can be observed that for both samples, approximately 75% of binder was removed after the first 15 min. However, after the first 15 min, the mix with the higher air flow rate lost asphalt binder at a higher rate. This study was the basis of the current test procedure. Figure 3. Effect of air flow rate on removal of asphalt binder by ignition (1).

62.3.2 Brown and Mager, 1995 (9) Brown and Mager conducted a round-robin study (RRS) to determine the accuracy and precision values for asphalt con- tent and gradation determination using the recently developed ignition method using a convection furnace (Thermolyne). Thirteen laboratories around the United States participated in the study. The study included four mixtures with different aggregates including gravel, granite, limestone, and trap rock. All of the mixtures had a different percentage of AC-20 asphalt binder and a 9.5 NMAS with different gradations. The asphalt mixtures were prepared and sent to the participant laborato- ries for asphalt content determination and gradation analyses. When this study was conducted, the procedure was not yet a standard; therefore, each laboratory was provided with a test procedure that included a test temperature of 1,000°F (538°C) based on preliminary studies conducted by the authors that found this temperature to be optimal. The samples had to be burned until the measured mass loss did not exceed 0.1 g for 3 consecutive minutes. The study concluded that the ignition method can be used to determine asphalt content and gradation of asphalt mixtures quickly and accurately. The authors developed precision statements for asphalt content and gradation for percent passing 4.75 mm and 75 µm sieves (as presented in Table 4). They also indicated that the within-laboratory and between-laboratory standard deviations for AC con- tent determination by the solvent extraction method were typically 0.21% and 0.22%, respectively; hence, the ignition procedure showed a greater precision than the extraction method. Additionally, the authors concluded that the test was simple to conduct and relatively inexpensive. Another interesting observation from the study was that when con- ducting the ignition test, the temperature inside the furnace increased by approximately 72°F (22°C) once the asphalt ignited. Additional round-robin studies were conducted in Florida, New Mexico, and Texas (10, 11, 12), and similar results were reported that support this study. The Florida study concluded that the asphalt content of an asphalt mixture could be deter- mined with a high degree of accuracy by the ignition method. The New Mexico study concluded that the precision of the ignition furnace was equal to the precision obtained with extraction testing. The Texas study concluded that the igni- tion furnace was a viable option for AC content and aggregate gradation determination but that the furnace had to be cali- brated for each mixture. 2.3.3 Mallick, Brown, and McCauley, 1998 (13) This study evaluated the effect of ignition on the proper- ties of four different types of aggregates (granite, limerock, gravel, and trap rock). The authors reported that for the same test duration, the ignition process had a greater effect on the properties of aggregate when testing an asphalt-aggregate mix instead of when an unbound aggregate sample was tested. The primary reason for this difference between testing the asphalt mixture and the aggregate only was the higher temperature that was generated when burning the asphalt binder from the mixture. 2.3.4 Prowell, 1998 (14) This study evaluated three different methods for asphalt content determination: solvent extraction, nuclear, and ignition furnace. A Thermolyne convection unit was used for the ignition method. A total of seven mix designs from Virginia were used: three corresponded to surface mixes and four were base mixes. Their gradation, coarse aggregate type, and asphalt content are presented in Table 5. A summary of the calibration factors of the aggregate only and of the mixes is shown in Table 6. The mixture calibration results were lower than the corresponding aggregate-only calibra- tion. The author indicated that the mixtures contained 1% hydrated lime and that the difference in calibration factors may be the result of a chemical reaction of the sulfur in the asphalt with the lime. A statistical comparison was con- ducted to evaluate the results on the surface mixes using the three methods. This comparison is presented in Table 7. The standard deviations were calculated on the difference between the measured and actual asphalt content. The author concluded that the best method based on the abso- lute percentage error and mean square error (MSE) was the ignition method. Test Property Standard Deviation Acceptable Range of Two Test Results Within Laboratory Between Laboratories Within Laboratory Between Laboratories Asphalt content 0.04 0.06 0.11 0.17 Percent passing 4.75 mm sieve 0.27 0.37 0.80 1.10 Percent passing 75 m sieve 0.47 0.65 1.30 1.80 Table 4. Precision statement for asphalt content percent by the ignition method – Thermolyne convection furnace (9).

7 and three sets of material for testing, consisting of three types of aggregates and one binder type. The aggregates were gran- ite, limestone, and trap rock. The asphalt content for the three mixtures was 6%, 5%, and 5.5%, respectively. The results of the measured asphalt content and bias of the results are presented in Table 8. Each number represents the average of at least two test results. These results show that the difference between the measured and actual asphalt content ranged from -0.11% to 0.27%. Table 9 summarizes the within-laboratory and between- laboratory standard deviations for the measured asphalt content for each mixture type. The within-laboratory standard devia- tion ranged from 0.09% to 0.10%, and the between-laboratory standard deviation ranged from 0.14% to 0.21%. The overall within-laboratory and between-laboratory standard deviations were reported as 0.09% and 0.17%, respectively. 2.3.6 Prowell and Carter, 2000 (16) This study evaluated the effect on Superpave consensus aggregate properties, aggregate bulk specific gravity (Gsb), and gradation of samples extracted using the ignition furnace Aggregate Basalt Granite/ Gneiss River Gravel Basalt Granite/ Gneiss River Gravel Limestone Mix Surface Surface Surface Base Base Base Base Design Asphalt Content (%) 4.7 5.7 4.9 4.4 4.9 4.4 4.3 Sieve Size, mm Percent Passing 50.0 100 37.5 100 25.0 100 19.0 100 12.5 97 9.50 84 4.80 57 2.36 40 1.18 28 0.60 19 0.30 12 0.15 8 0.075 5.2 100 100 100 100 94 81 56 43 31 21 12 6 4.1 100 100 100 100 99 82 56 42 31 22 14 8 5.5 100 100 95 83 65 58 49 39 29 21 13 8 5.7 100 100 98 90 71 62 49 36 25 17 11 7 4.8 100 100 95 78 65 56 36 28 24 20 14 8 5.0 100 99 81 66 52 47 36 24 14 9 6 5 3.9 Table 5. Mixtures gradation and asphalt content (14). Aggregate Source Correction Factor – Thermolyne Ignition Furnace (%) Mix Aggregate Only Surface Mixes 1 0.09 0.34 2 0.45 0.53 3 0.06 0.3 Base Mixes 1 0.15 NA 2 0.52 NA 3 0.06 NA 4 −0.05 NA Table 6. Ignition furnace calibration factors by aggregate type (14). Test Method Number of Observations Standard Deviation % Asphalt Content Bias % Asphalt Content Absolute Error % MSE Surface Mix Specimens Extraction 23 −0.113 5.16 0.0969 Nuclear gauge 108 0.070 2.51 0.0245 Ignition 24 −0.033 1.50 0.0132 Base Mix Specimens Nuclear gauge 144 0.080 2.98 0.0257 Ignition 46 0.29 0.14 0.11 0.139 0.101 0.027 1.66 0.0109 Table 7. Summary of statistics by test method (14). 2.3.5 Mallick and Brown, 1999 (15) In 1999, NCAT conducted a limited round-robin study to determine the accuracy and precision of a new ignition furnace, an infrared unit developed by Troxler. Five laboratories partici- pated in this study. Each laboratory was provided with a furnace

8for typical Virginia aggregates. A total of 10 Superpave mix designs from nine aggregate sources were used and included 12.5, 19.0, 25.0, and 37.5 mm nominal maximum size aggre- gate blends. Table 10 presents comparisons between the number of significant differences and correction factors for the different aggregates under evaluation. For six of 10 cases, testing in the ignition furnace caused significant differences between the Gsb before and after the test. The authors also reported that aggregate recovered using the ignition furnace appeared to be unsuitable for sand equivalency testing. The results of fine aggregate angularity testing were significantly different between the virgin and burnt samples in three of 10 cases. The authors also noted that accurate results may be obtained for gradation analysis and flat and elongated par- ticle measurements performed on aggregates recovered from the ignition furnace test. 2.3.7 Prowell and Youtcheff, 2000 (17 ) Prowell and Youtcheff conducted a study in Virginia to evaluate mixture components that might affect the igni- tion furnace correction factor. Four experiments were designed to investigate the effects of the amounts of lime, sulfur, cal- cium carboxylates, and fines. A 9.5 mm NMAS Superpave mix and asphalt binders with different chemistries were included in each experiment. The authors found that the variation in the percentage of hydrated lime added to the mixture had a significant effect on the ignition furnace correction factor. The results are summarized in Table 11. The correction fac- tor varied from 0.64% with no hydrated lime to 0.13% with 2% hydrated lime. The variation reported was large enough to cause noncompliance with quality control tests according to VDOT’s specifications. The authors indicated that the lime appears to react with the sulfur dioxide (SO2) produced from the combustion of organic sulfur to generate calcium sulfate. They also reported that the amount of sulfur present in the asphalt binder significantly affected the mass loss occurring during the ignition process, but to a lesser degree than the lime content. The other two components, calcium carboxylates and fines, did not have a significant effect on the ignition furnace correction fac- tors, but the authors recommended that additional testing be Laboratory Aggregate Type Measured Asphalt Content (%) Actual Asphalt Content (%) Bias Lab 1 Granite 5.91 6.0 Limestone 4.99 5.0 Trap rock 5.39 5.5 Lab 2 Granite 6.23 6.0 Limestone 5.02 5.0 Trap rock 5.66 5.5 Lab 3 Granite 6.14 6.0 Limestone 5.15 5.0 Trap rock 5.77 5.5 Lab 4 Granite 5.97 6.0 Limestone 5.24 5.0 Trap rock 5.64 5.5 Lab 5 Granite 6.00 6.0 Limestone 5.03 5.0 −0.09 −0.01 −0.11 0.23 0.01 0.16 0.14 0.15 0.27 −0.03 0.24 0.14 0.00 0.03 Trap rock – 5.5 – Table 8. Average measured and actual asphalt content and bias for each laboratory and aggregate type (15). Material Standard Deviation Within Laboratory Between Laboratories Granite 0.0987 0.151 Limestone 0.0894 0.143 Trap rock 0.100 0.218 Table 9. Within- and between-laboratory standard deviation for asphalt content for different aggregates using Troxler unit (15). Aggregate Type Fi ne A gg re ga te A ng u la ri ty Sa nd Eq ui va le nt Fi ne A gg re ga te (G sb ) Fl at a nd El on ga te d 5: 1 C oa rs e A gg re ga te (G sb ) N om in al M ax im um 4. 75 m m 0. 07 5 m m To ta l Si gn ifi ca nt D iff er en ce s Fu rn ac e C or re ct io n Fa ct or (% ) Siltstone 12.5 mm – – S – – – – – 1 0.09 Quartzite 12.5 mm S S S – S – – – 4 0.79 Granite 12.5 mm S S S – S – – – 6 0.30 Diabase mix 12.5 mm – S – – – S – S 3 0.43 River gravel 12.5 mm – S – – S – – S 2 0.10 Limestone 19.0 mm – – S S – – – – 2 0.28 Granite/gravel 25.0 mm S S S – – – – S 3 0.30 Siltstone 25.0 mm – – – – S – – – 2 0.09 Diabase 25.0 mm – – – – S – – – 1 0.14 Granite 37.5 mm – – – – S – – – 1 2.02 Note: S = significant. Table 10. Comparison of significant differences between samples means properties by aggregate type and correction factor (16).

9 conducted to assess the effect of fines’ variability with other aggregates, particularly dolomites. 2.3.8 Williams and Hall, 2001 (18) This study compared the standard convection furnace (orig- inal ignition furnace developed at NCAT) to infrared furnaces by determining asphalt content and corresponding correction factors. The study included 12 Superpave mixtures with aggre- gates from Arkansas with two NMAS (12 and 25 mm) and two performance-grade (PG) binders (PG 64-22 and PG 70-22). The primary aggregate types included sandstone, limestone, syenite, pit gravel, and river gravel. The study concluded that the correction factors for the two furnaces were different and recommended that each blend of aggregate be calibrated for each furnace used. The results for the 12 mixtures are pre- sented in Figure 4. Regarding the asphalt content, a paired t-test showed that the results for both furnaces were similar and had similar levels of accuracy. Additionally, the study reported that the peak temperature achieved by the infrared furnace was often higher than the peak temperature achieved by the convection furnace and that both furnaces had similar testing durations—approximately 40 to 45 min. 2.3.9 Prowell, 2002 (19) Similar to the Williams and Hall study, this report evaluated the accuracy of infrared ignition furnaces compared to convec- tion units. Additionally, the degradation of aggregate produced in both units was compared. The study used two nominal max- imum aggregate sizes (9.5 and 19 mm), four aggregate types (granite, crushed gravel, limestone, and dolomite), and two asphalt contents (optimum and optimum + 0.5%) for a total of 48 samples in each furnace. A PG 64-22 asphalt was used in all samples. For each mixture–furnace combination, correc- tion factors for aggregate loss were obtained by burning three samples with optimum asphalt contents and averaging the difference between the total loss and the known asphalt con- tent. From this analysis, it was found that each mixture had a unique correction factor for aggregate loss and that these fac- tors were significantly different for both furnaces. The results are summarized in Table 12. The accuracy and variability of measured asphalt contents were also evaluated using the aggregate correction factors deter- mined for mixtures at optimum asphalt content. The accuracy of the asphalt content was measured in terms of the bias (differ- ence between the measured and true asphalt content of the Description Measured Asphalt Content Average Standard Deviation Correction Factor (%) 1 2 3 Control 5.84 5.80 5.89 5.84 0.045 0.64 +0.5% hydrated lime 5.63 5.65 NA 5.64 NA 0.44 +1% hydrated lime 5.47 5.50 5.45 5.47 0.025 0.27 +1% hydrated lime aged 2 h 5.46 5.59 5.44 5.50 0.081 0.30 +2% hydrated lime 5.32 5.31 5.35 5.33 0.021 0.13 Table 11. Effect of hydrated lime content on ignition furnace factors (17). Figure 4. Comparison of correction factors for infrared and standard (convection) furnaces (18).

10 samples), and the variability was measured in terms of the standard deviation of the results. The results are presented in Table 13. An analysis of variance was conducted to evaluate the results. The analysis showed that the biases were not significantly different. The author concluded that when properly calibrated, both units could produce accurate results regardless of the mixture evaluated. The standard deviations of the biases were not sta- tistically different. The author concluded that the variability data were not sufficient to use for comparison of the precision statement for the two types of units and recommended a round- robin study to confirm that the precisions of both units were similar. Finally, a comparison of the aggregate gradations recov- ered from both units showed no significant difference. These comparisons were conducted with the NMAS, 4.75, 2.36, and 0.075 mm sieves. 2.3.10 Sholar, Page, and Musselman, 2002 (20) In 2002, a round-robin study was conducted in Florida to determine precision values for asphalt content and gradation for plant-produced mixtures. The study indicated that plant- produced mixtures have different possible sources of vari- ability that include differences in an asphalt mixture within the truck bed, collection of samples from the truck, splitting of the mixture into samples for testing, differences in equip- ment, and operator differences. The study included 12 labo- ratories and nine different mixtures: six Superpave mixtures (four coarse mixtures and two fine mixtures) and three open- graded friction courses. The major aggregate types corre- spond to limestone and granite. Reclaimed asphalt pavement (RAP) was included in some of the mixtures, but no informa- tion about the percentages used was reported. Each mixture was sampled at the truck bed at the asphalt plant. The study recommended new precision values (presented in Table 14) to be used for plant-produced mixtures only. These values are currently included in the Florida Method FM-5-563 for the determination of asphalt content of asphalt mixtures by the ignition method (21). 2.3.11 Prowell and Hurley, 2005 (22) During the development of the ignition equipment, it was found that for both methods (internal and external weigh- Mixture Type Asphalt Correction Factor (%) Convection Infrared 9.5 mm granite 0.07 0.01 9.5 mm crushed gravel 0.11 0.03 9.5 mm limestone 0.24 0.14 9.5 mm dolomite 0.66 0.51 19.0 mm granite −0.03 −0.13 19.0 mm crushed gravel −0.02 −0.04 19.0 mm limestone 0.19 0.16 19.0 mm dolomite 0.55 0.40 Table 12. Average correction factors for aggregate loss (19). Aggregate NMAS ActualPb Measured Pb Bias (Measured Actual) Average Bias Standard Deviation 1 2 3 1 2 3 Convection Furnace Dolomite 5 5.05 4.51 5.07 −0.12 0.3177 Dolomite 4.3 4.24 4.22 4.24 −0.07 0.0115 Granite 5.8 5.84 5.75 5.74 −0.02 0.0551 Granite 4.7 4.78 4.77 4.73 0.06 0.0265 Gravel 5.2 5.14 5.15 5.22 −0.03 0.0436 Gravel 4.7 4.76 4.81 4.7 0.06 0.0551 Limestone 4.2 4.17 4.09 4.17 −0.06 0.0462 Limestone 4.2 4.15 4.21 4.09 −0.05 0.06 Avg. −0.0292 0.077 Infrared Furnace Dolomite 5 4.98 4.95 4.97 −0.03 0.0153 Dolomite 4.3 4.29 4.3 4.27 −0.01 0.0153 Granite 5.8 5.78 5.77 5.7 −0.05 0.0436 Granite 4.7 4.77 4.83 4.72 0.07 0.0551 Gravel 5.2 5.19 5.22 4.93 −0.08 0.1595 Gravel 4.7 4.68 4.7 4.5 −0.07 0.1102 Limestone 4.2 4.06 4.15 4.14 −0.08 0.0493 Limestone 9.5 19 9.5 19 9.5 19 9.5 19 9.5 19 9.5 19 9.5 19 9.5 19 4.2 4.02 4.16 4.02 0.05 −0.06 0.04 0.08 −0.06 0.06 −0.03 −0.05 −0.02 −0.01 −0.02 0.07 −0.01 −0.02 −0.14 −0.18 −0.49 −0.08 −0.05 0.07 −0.05 0.11 −0.11 0.01 −0.05 0 −0.03 0.13 0.02 0 −0.05 −0.04 − 0.07 −0.06 −0.06 0.03 0.02 0.00 −0.03 −0.11 −0.03 −0.03 −0.1 0.02 −0.27 −0.2 −0.06 0.18 −0.13 0.0808 Avg. −0.0479 0.0661 Table 13. Measured asphalt content, biases, and standard deviation by furnaces (19).

11 ing), most aggregates would lose some mass during the burning process; hence, a correction factor had to be deter- mined and used to account for this aggregate mass loss when calculating the asphalt content. As the equipment began to be more widely used, it was learned that some aggregates (for example, dolomites) lose a significant amount of mass during the test. As a result, while the correction factor was typically 0.1% to 0.3% for most aggregates, it could be well over 1% for others. Further studies were conducted by Prowell and Hurley to refine the ignition method so that it could be used effec- tively for HMA containing aggregates having significant mass loss during tests. A secondary objective was to compare the aggregates, which were selected from various locations throughout the United States represented by different types of aggregates that were known to experience high mass loss dur- ing ignition tests. Six mix designs were included in this study with high-loss aggregates: four contained dolomites, one a basalt aggregate, and one a serpentinite/chlorite aggregate. The design gradations and optimum asphalt contents are summarized in Table 15. Six NMASs were used: three 19 mm NMASs, two 12.5 mm NMASs, and one 9.5 mm NMAS. A PG 64-22 binder was used for five mixtures, and a PG 64-28 was used for one mix design (AZ). The materials, aggregate, and binder were shipped to the participants after all mixtures were prepared at NCAT’s laboratory. A further objective was to compare the four ignition furnace and asphalt content determination methods: Thermolyne, Troxler NTO Infrared, Tempyrox Pyro-Clean (used for glassware and metals clean- ing), and Ontario. The Tempyrox Pyro-Clean system is used to clean organic matter from precision laboratory glassware or metal parts. The temperature is carefully controlled in these units, pre- venting the aggregates from being heated to higher tem- peratures than required. The furnace operates on the basis of pyrolysis oxidation, which allows organic matter to be removed from a material in the absence of oxygen until only carbon ash remains. Stainless-steel shelves inside the fur- nace hold the glassware. The temperature rises inside the chamber and starts to drive off hydrocarbons, and organic matter smoke reacts with a catalyst inside the chamber and reduces the oxygen to less than 15%, where no ignition can occur. This allows for better control of the temperature to exact values. The Ontario method in this study corresponded to the Ontario Ministry of Transportation Test Method LS-292. Although a convection or infrared unit can be used, a con- vection unit was used for this particular study. The method specifies that the sample has reached the end point when the change in sample mass is less than 1 g for 3 consecutive minutes. This is in contrast with AASHTO T 308, which spec- ifies that the end point occurs when the sample mass does not change by more than 0.01% for 3 consecutive minutes. The correction samples are burned at 430, 480, and 540°C, Condition Standard Deviation Acceptable Range of Two Test Results Within laboratory: asphalt content (%) 0.1138 0.32 Between laboratories: asphalt content (%) 0.1563 0.44 Table 14. Precision statement for asphalt content for plant-produced mixtures – Florida (20). Sieve Size Alabama (Dolomite) Indiana (Dolomite) Missouri (Dolomite) Wisconsin (Dolomite) Maryland (Serpentinite/ Chlorite) Arizona (Basalt) 25 19 12.5 9.5 4.75 2.36 1.18 0.6 0.3 0.15 0.075 Asphalt content (%) 100 95.1 80.7 69.3 44.5 27.7 17.1 12.0 9.2 7.6 3.9 4.5 100 100 100 94.8 57.4 35.8 28.0 20.0 10.0 5.6 4.4 5.6 100 100 98.3 89.8 54.1 38.0 25.6 17.7 10.5 6.1 3.3 5.3 100 97.1 84.1 72.2 52.0 34.5 21.7 13.1 6.7 4.4 3.7 4.6 100 100 98.8 88.3 43.5 25.5 19.1 16.0 12.9 5.3 4.2 5.0 100 100 84.0 75.3 58.3 42.9 29.7 19.3 10.4 6.0 4.2 4.7 Table 15. Aggregate gradation and asphalt content (22).

12 and the highest temperature that gives a correction factor of less than 1% is selected. For the four types of furnaces investigated, significant varia- tion was found in the asphalt correction factors for these aggregates with high mass loss. The results are summarized in Table 16 and presented in Figure 5. The Thermolyne equip- ment always produced the highest aggregate mass loss, and the Tempyrox always produced the lowest mass loss. The standard deviation of the corrected asphalt contents for these high-mass-loss aggregates was higher than the within- laboratory standard deviation reported for AASHTO T 308, with the exception of the aggregate from Alabama. A sum- mary of asphalt content correction factor bias found in this study is presented in Figure 6. As part of this study, the researchers investigated aggregate breakdown by sieve size. Critical sieve sizes for HMA con- trol were identified as the NMAS and the 4.75 mm sieves. The authors’ statistical analysis showed that there was not a significant effect from the furnace or method for the per- centages of aggregate breakdown on either the NMAS or the 4.75 mm sieve size. They concluded that no method improved or reduced this loss during ignition testing. 2.3.12 Kowalski, McDaniel, Olek, and Shah, 2010 (23) States like Indiana have reported problems with the use of ignition furnaces with some types of aggregates, such as dolomites. This study, conducted with dolomites, indicated that the high temperature required during ignition testing produced a decomposition or a chemical change in the aggregates that caused mass loss to continue after the binder was burned off; as a result, the mass remaining at the end of the test could be variable and even higher than at the start. As part of this study, the research team investigated differ- ent variables that could influence the mass loss during the Source Thermolyne Troxler Tempyrox Ontario Method Correction Factor (%) Temp. (°C) AL 0.74 0.19 0.14 0.46 540 IN 2.65 1.25 0.59 0.85 450 AZ 2.53 1.82 1.38 1.89 450 MO 3.26 1.10 0.32 0.78 450 WI 1.62 0.48 0.08 0.75 480 MD 1.62 0.76 0.42 0.69 480 Table 16. Average correction factors by method (22). Figure 5. Comparison of furnace correction factors by method (22).

13 tests. The following paragraphs summarize some of their key findings. 2.3.12.1 Test Temperature Effect The influence of test temperature on mass loss during the ignition test was investigated using three temperatures: 1,000, 900, and 800°F (538, 482, and 427°C). A 25 mm NMAS mix with a PG 64-22 binder was used in this evaluation, and the design binder content was 4.6%. The study found that the observed mass losses were a function of the test tempera- ture, with higher losses as temperature increased. It reported mass losses of approximately 8.5% at 1,000°F (538°C), 6.5% at 900°F (482°C), and 5.8% at 800°F (427°C). The authors concluded that since the asphalt content of this mixture was 4.6%, the difference in mass losses must represent the mass loss due to thermal decomposition of the dolomite; these increases were 3.9%, 1.5%, and 1.2%, respectively. A strong relationship between test temperature and sample mass loss was found, shown in Figure 7. They also found that the furnace temperature exceeded the target temperature in all cases and that the higher the test temperature, the faster the temperature was exceeded. This indicates that by decreas- ing the temperature, a significant effect on the mass loss and rate of mass loss can be achieved. 2.3.12.2 Test Time Effect The influence of test time on mass loss was also inves- tigated as part of this study. In order to accomplish this, ignition tests of between 90 and 240 min were conducted at 1,000°F (538°C). The results are presented in Figure 8. This linear relationship shows that the mass loss was 0.2% for each 10-min period. Ignition tests were also conducted on the dolomite aggregate blend with no binder, and the results showed that the mass loss was 0.1% for every 10 min. This indicates that at high temperature, test time does affect mass loss, which also confirms that ignition of the binder in the mixture increases the mass loss rate. 2.3.12.3 Mixture Type Effect The effect of mixture type on ignition results was also investigated using three laboratory-produced mixtures. A 25 mm NMAS with 4.6% binder (HM-1), a 9.5 mm mixture with 5.4% binder (HM-2), and a 19 mm mixture with 4.5% binder (HM-3) were investigated. Tests were conducted at Figure 6. Summary of asphalt content correction factor bias by method (22). Figure 7. Mass loss as a function of test temperature (23).

14 the standard test temperature of 1,000°F (538°C). Figure 9 shows the temperatures and mass loss plots for each test as a function of time. The highest temperatures recorded were 1,076, 1,085, and 1,112°F (580, 585, and 600°C) for each mix- ture, respectively. This indicates a temperature increase of 108 to 144°F (42 to 62°C) with respect to the preset furnace temperature. Higher binder content produced higher mass loss, as expected. From the plot, it can also be observed that higher binder contents seem to cause faster increases in fur- nace temperature. 2.3.12.4 Temperature Distribution Effect The temperature inside the ignition furnace was moni- tored with the furnace thermistor and two additional ther- mocouples installed in the top (TC1) and bottom (TC2) of the sample baskets. Figure 10 summarizes the results of the tests, which were conducted using a mixture containing dolo- mite (HM-1) and a dolomite aggregate sample (DBL-1 with- Figure 8. Mass loss as a function of test time (23). Figure 9. Temperatures and mass loss plots for each test as a function of time for HM-1 (4.6% AC), HM-2 (5.4% AC), and HM-3 (4.5% AC) (23). out binder) at 900°F (482°C) [similar trends were obtained at 800 and 1,000°F (427 and 538°C)]. This figure shows that the temperature inside the furnace was not constant and that the temperature in the top basket was higher than that in the lower basket. When the temperature was stabilized, the dif- ferences became relatively constant. The temperature differ- ence between the bottom and top baskets was 136°F (58°C), between the top basket and furnace was 95°F (35°C), and between the furnace and bottom basket was 75°F (24°C). These differences were similar at different target tempera- tures, as shown in Figure 11. The authors concluded that these temperature differences might cause an increase in mass loss and decomposition of the aggregate and that plac- ing the mix sample only in the bottom basket would result in a more uniform temperature, hence limiting the decomposi- tion of high-loss aggregates. 2.3.12.5 Aggregate Properties Before and After Ignition Tests A 25 mm NMAS with 4.6% binder was used to com- pare the results of bulk and apparent specific gravities and absorption of the aggregate before and after ignition. The authors reported that the ignition test affected the results significantly. After ignition, specific gravities decreased and absorption increased. The results are summarized in Table 17. Additionally, the aggregate gradations of different sizes of dolomites were measured before and after the ignition test at 1,000°F (538°C) with samples without binder. The greatest change in aggregate gradation was observed for the coarse aggregate (Gradation #5) with differences of as high as 10%. No significant differences were found for fine aggre- gates (Gradations #8 and #12). These results are presented in Figure 12.

15 2.3.12.6 Thermogravimetric Analysis Thermogravimetric analysis was also conducted as part of this study. The results indicated that the sample dolo- mites that were exposed to high temperatures in the ignition furnace had decomposed partially; hence, less mass loss in the thermo gravimetric analysis was recorded. On the other hand, samples of dolomites recovered from the mix samples after ignition decomposed more than samples of aggregate only; this was attributed to the higher temperatures in the ignition furnace when the binder ignited, as presented in Figure 10. 2.3.12.7 Modification of Ignition Furnace Test Procedure Based on the findings of this study, modifications to the ignition furnace test procedure were suggested for the high- mass-loss materials in Indiana. In order to reduce the mass loss, it was recommended to reduce the total mass by one-half, use only the bottom basket, and run the test at 800°F (427°C). The modified protocol was recommended for problematic aggregates, aggregates with correction factors of greater than or equal to 1%, or when the test was not completed in less than 90 min. Results from the modified procedure were verified using six plant-produced mixtures with problematic aggre- gates. Comparing these ignition tests to solvent extraction Figure 10. Temperature distribution inside ignition furnace, TC1, TC2, and furnace Thermistor (HMA mixture and dolomite aggregate sample) (23). Figure 11. Relationship between furnace and thermocouples in the top and bottom basket when temperature stabilized [target temperatures, 800, 900, 1,000çF (427, 483, and 538çC)] (23). Mixture at 1,000°F (538°C) Before Ignition After Ignition Bulk specific gravity 2.710 2.423 Apparent specific gravity 2.773 2.642 Percent absorption 0.8 3.4 Table 17. Specific gravities and absorption of dolomite before and after ignition (23).

16 results showed very similar results. The current Indiana DOT test method to determine asphalt content by ignition allows the test to be conducted at 800°F (427°C) if dolomite is used in the mixture (24). 2.4 RAP and RAS Asphalt Content Determination Using Ignition Furnace The price of HMA has significantly increased in recent years, primarily due to the increased cost of asphalt binder. As a result, contractors and state DOTs have looked for ways to reduce the cost of HMA without sacrificing performance. Two of the most common methods used to reduce costs are higher RAP content and the use of reclaimed asphalt shingles (RAS) in asphalt mixtures. These materials provide a portion of the asphalt binder in the recycled HMA, thus reducing the overall cost of the mixture. One of the issues when using higher RAP or RAS is the need to adjust the procedures for measuring asphalt content with the ignition test. For example, the AASHTO and ASTM ignition test procedures do not provide adequate guid- ance when using high RAP or RAS. Hence, modifications to the test procedure to make it more appropriate for high RAP and RAS are needed. One problem with using the ignition test to determine the asphalt content for RAP and RAS mixes is determining the correction factor for asphalt content or aggregate gradation. The procedure for virgin mixtures is to mix a known grada- tion with known asphalt content and conduct the ignition test. The difference between the actual gradation and actual asphalt content and between the measured gradation and measured asphalt content is identified as the correction factor. This procedure works well for virgin mixtures but becomes more difficult when RAP or RAS is used. If the amount of RAP or RAS is relatively small, the correction factor of the mixture can be determined without significant issues. How- ever, as the amount of these recycled materials is increased, the need to have a more specific procedure for these mixtures is increased. Hence, improved guidance is needed for HMA containing RAP or RAS. One approach for handling RAP or RAS is to determine the asphalt content and gradation of these materials before they are mixed with virgin materials (25, 26). These test results for the RAP and RAS, along with the known asphalt content and gradation of the virgin materials, can be used to calculate the actual asphalt content and gradation of the recycled mix- ture. The correction factor can then be established by conduct- ing the ignition test on the recycled mixture and subtracting the measured asphalt content and gradation from the actual asphalt content and gradation. When using this process, the asphalt content and gradation of the RAP or RAS are typically determined using the solvent extraction test or the ignition test. The solvent extraction test may not be able to remove all of the asphalt binder, especially in post-consumer RAS, and this can lead to some error (27). Also, when the ignition test is used to determine the components of RAP and RAS, it may result in some error since the correction factor is not known for the RAP and RAS. Shingles are generally manufactured using high-quality aggregates that tend to have very little mass loss or change in particle size during the ignition test (25). In this case, the assumption that the ignition test for the RAS materials has a correction factor of 0% should not introduce much error (28). However, it would be good to conduct the solvent extrac- tion test if there is any concern that the correction factor for the shingles is relatively high. The same is true for relatively high RAP content mixtures. Another approach for RAS or RAP is to use a correction factor that has been established based on past test results (27). Some agencies require that the asphalt binder from the recycled mixture be recovered and tested, and in this case the solvent extraction test will have to be used. One caution that should be exercised when determining the asphalt content of RAS is the sample size used for testing (25). Figure 12. Effect of ignition on aggregate gradation (23).

17 RAS typically contains 20% to 30% asphalt binder, which is much greater than the 5% to 6% that normally exists in HMA. If the sample size is too large, the equipment may be damaged as combustion begins due to the large amount of asphalt binder. Also, all of the asphalt binder may not be removed during combustion if the sample size is too large (28). It is recommended that the sample size of asphalt shingles being tested not exceed 500 to 700 g due to the high asphalt content. 2.5 Summary of Findings from the Literature Review Most of the research studies on ignition furnaces for asphalt content determination were conducted in the mid-1990s to mid-2000s. The main focus was to evaluate the effectiveness and accuracy of the new method with respect to the extraction method and to compare the different units/brands available. The studies mainly focused on evaluating variables related to the mixture components. The evaluation of issues related to the influence of other factors, such as installation and maintenance of the equipment, is very limited. Some of the general findings from the literature supported by at least one study are summa- rized in the following. • Convection and infrared heating mechanisms are differ- ent; both can be used to obtain aggregate and asphalt cor- rection factors with the accuracy required in the AASHTO standard, but the correction factors are different. Correc- tion factors for each furnace should be used. • The addition of lime in the mixture decreases the asphalt correction factors. • The source of aggregate has a significant impact on the cor- rection factors. • One study suggested that when a convection unit is used, the temperature inside the furnace is not constant and is not distributed uniformly; also, the temperature in the top basket is higher than that in the lower basket (23). • The same study suggested that for high-mass-loss aggregates, the test temperature and test time have a significant impact on the correction factors; higher test temperatures and test time seem to produce higher mass losses (23). Also, peak temperatures inside the furnace during ignition can be sig- nificantly higher than the specified test temperature. • One study recommended that plant-produced mixtures should use a different precision statement than laboratory- produced mixtures (20). The authors indicated that differ- ences in an asphalt mixture within the truck bed, collection of samples from the truck, splitting of the mixture into samples for testing, differences in equipment, and operator differences are additional variables not included affecting laboratory mixes. • Changes in gradation after ignition seem to occur for some coarser aggregates. • Higher binder contents seem to cause faster increases in furnace temperature.