Data Mining Statewide Department of Transportation Volumetrically Designed Asphalt Mixture Records

Abstract

In recent years, the asphalt paving industry has been strained by numerous factors including increased asphalt binder costs, funding that has not kept up with material costs, increased societal pressure to recycle, and deteriorating pavement networks. Mix design should account for the market in which it is used, which is very different now than when today’s volumetric mix design practices were developed (many of the aforementioned factors were less present). Given this reality, a statewide database of all 1,452 approved mix designs in Mississippi from 2005 to 2018 was compiled and analyzed, and the objective of this paper is to present findings, trends, and unintended consequences of exclusive reliance on volumetrics. With volumetrics-only mix design, asphalt content is primarily controlled by voids in mineral aggregate (VMA), which is influenced by aggregate bulk specific gravity (G_sb). Minor G_sb deviations (i.e., within AASHTO d2s limits), can significantly affect VMA, so much so that 99% of Mississippi’s mixes could be failing VMA while reported VMA passes. This allows mix manipulation and economization, with 0.8% asphalt content reductions possible while still meeting volumetric requirements. Recycled materials can exacerbate this issue, and common approaches to increase asphalt content (decreasing design gyration level or using finer gradations) are ineffective with fixed VMA requirements. Overall, the mix design database analysis agrees with numerous smaller studies but does so with an entire state’s actual practice. This presents a compelling case that volumetrics-only mix design has limitations, and supports ongoing efforts to reintegrate mechanical tests.

Keywords

infrastructure materials asphalt production and use mix design volumetrics

Economic, societal, and performance demands generally compel a field to change over time, and in recent years, asphalt paving has been strained by numerous factors. Figure 1 summarizes a 40-year period and highlights several of the most pressing factors compelling asphalt paving to change. For the past 15 years, Mississippi asphalt binder prices have been roughly triple their price for the 25 years prior, and since the mid-1990s fuel taxes have been essentially the same. Societal pressure to recycle post-consumer materials into pavements has also progressively increased, as has economic pressure to use lower cost recycled materials such as ground tire rubber (GTR), reclaimed asphalt pavement (RAP), reclaimed asphalt shingles (RAS), and re-refined engine oil bottoms (REOB). During this period, traffic levels have also increased.

Figure 1.

Historical overview of four decades of asphalt mix design, recycled materials, fuel tax, and Mississippi asphalt binder index prices (note Mississippi Department of Transportation does not presently use all the recycled materials shown).

Figure 2 shows the negative effects represented by Figure 1 on the Mississippi Department of Transportation (MDOT) highway network. MDOT uses a pavement condition rating (PCR) that includes International Roughness Index (IRI), rutting, and cracking distresses. Current MDOT data collection practices do not record a separate cracking index, though Figure 2 data show that IRI and rutting do not entirely account for the fair and poor ratings. Within MDOT, brittleness and cracking associated with binder contents and properties is believed to be of first-order importance.

Figure 2.

Overview of Mississippi Department of Transportation pavement network condition: note that high pavement condition rating (PCR), low rut depth, and low International Roughness Index (IRI) are desirable.

Mix design, although far from the only factor leading to overall pavement network condition, plays an important role. Mix design should consider the overall market within which the mix is used. As seen in Figure 1, the market is very different now than when Marshall design was the dominant method, and it is also very different than when research occurred to develop Superpave, the volumetrics-based method currently used throughout the United States, including Mississippi. Given market changes represented by Figure 1, it is no surprise that the pavements community is evaluating potential mix design improvements. Superpave was not developed to address present day marketplace factors, nor was Marshall developed for marketplace factors of the 1980s (the following section briefly reviews asphalt mix design methods over time).

Facing the realities of Figures 1 and 2, MDOT initiated an effort to compile a database of all approved mix designs statewide between 2005 and 2018. This paper’s objective is to present findings from this data-mining exercise, evaluate global trends, and explore unintended, and sometimes intuitive, consequences of volumetric mix design. Motivations for this exercise were: (1) to begin the process of assessing how asphalt should be designed within the MDOT network in future years understanding the necessity of recycling; (2) present these findings to other agencies and researchers who may have use for these data or be interested in performing similar investigations within their own network; and (3) assess literature that is primarily composed of modest to small datasets and use this larger dataset to provide context to a worldwide mix design knowledge base.

Note that this paper focuses only on the mix design database as it is believed that much information can be gleaned from such an exercise. Assessing field pavement performance data was outside the scope of this paper, although it is undeniable that mixture properties ultimately influence pavement performance. In a similar manner, mixture properties are dependent on mix design methodology, and it is in this area, volumetric mix design in particular, that this paper seeks to provide insight.

Mix Design Background

Selection of a binder content high enough to resist cracking and brittleness problems but low enough to avoid rutting has conceptually existed for over 100 years, though the relative importance of each extreme has varied over time. Procedures in the early 1900s relied more on visual inspection and general engineering concepts ( 1 ). Procedures originating in the 1920s to 1940s (i.e., Hubbard-Field, Hveem, and Marshall) were driven by a combination of volumetric parameters (e.g., voids in mineral aggregate, or VMA) and mechanical tests (e.g., Hubbard-Field stability, Hveem stability and cohesion, Marshall stability and flow) and are thoroughly summarized elsewhere ( 2 – 6 ).

Marshall mix design was used worldwide and became the most widely implemented method for approximately 50 years. Despite its success, Marshall had known shortcomings from the time of its early development (i.e., stability testing) that were documented in early work of the U.S. Army Corps of Engineers ( 7 ). Stability testing, however, was still adopted by many Departments of Transportation (DOTs), illustrating that some concepts are extensively implemented with less technical backing than their widespread use would suggest. Conceptually, Marshall principles were akin to what is referred to today as balanced mix design (BMD), albeit with technology of generations past (e.g., stability equipment). Marshall methods were referred to as balanced procedures as early as the 1960s by Heithaus and Izatt ( 8 ), and in this case, volumetric principles and mechanical mixture tests were used to select a binder content considering multiple factors.

By the early 1980s, Leahy and McGennis ( 9 ) noted a growing feeling that design methods like Marshall had outlived their usefulness. Incidentally, the Strategic Highway Research Program (SHRP) began in the late 1980s. Superpave was one of the main products of SHRP, where the initial intent was to develop a design method closely tied to performance that could reduce widespread rutting problems.

Early Superpave concepts contained multiple levels of design and mixture performance testing, though performance tests were never implemented ( 10 ). As implemented, Superpave relied almost solely on volumetric properties for which definitions and methods of measurement and calculation were fairly well established in decades prior through key works like Rice ( 11 , 12 ) and McLeod ( 13 , 14 ). Superpave implementation began in 1996, and for reference, MDOT conducted Superpave pilot projects in 1996 with full adoption the following year in a special provision dated April 23, 1997. By 2000, most DOTs had adopted Superpave.

Superpave volumetric design ushered in several positive changes to asphalt mix design but also led to some unintended consequences. During the transition from Marshall to Superpave, increased rutting resistance was generally accomplished by tightening some aggregate requirements and reducing design asphalt content, generally leading to reduced mixture durability. “Dry” mixes, as they are informally called, have, at least to some extent, become associated with Superpave and are currently one of the most prevalent mix design problems facing the asphalt industry ( 5 , 15 , 16 ).

With a volumetrics-only design approach, opportunities exist for economizing mixes by further reducing design asphalt content, as will be discussed in this paper. Given that asphalt binder costs over an order of magnitude more than aggregate and that asphalt paving takes places in a competitive bidding environment, Figure 1 will influence as many facets of mix design as possible. Also, accurate measurement and calculation of volumetrics is more challenging with the increased variety and quantity of recycled materials.

To combat dry mixes, the asphalt industry has been working to increase mix durability in recent years. Increased binder content has been pursued volumetrically through adjustments such as decreased design gyrations (N_des) ( 15 , 17 , 18 ), decreased design air voids (V_{a, des}) ( 15 , 18 ), and increased VMA ( 18 , 19 ). The recently developed Superpave5 method ( 20 ) is a purely volumetric method that pursues durability through higher in-place densities brought about by simultaneous changes to multiple volumetric criteria (V_{a, des}, VMA, and N_des). Higher performance materials such as polymer-modified asphalts have also been documented as another means to increase durability ( 18 , 21 , 22 ), though polymer-modified asphalt is more expensive, as shown in Figure 1.

The asphalt industry is also actively pushing to identify, study, and adopt mechanical mix tests that can be used to supplement or even replace volumetrics for mix design ( 22 – 29 ). These tests are generally presented in a BMD context where at least one cracking test and one rutting test would be used in tandem ( 29 – 32 ). Though not commonly identified as such, the BMD thrust is largely a return to Marshall mix design philosophies (albeit with better mechanical tests) in the sense that multiple factors, including mechanical properties, are used to select a design binder content for an acceptable blend of aggregates. A common advertisement for BMD using mechanical tests is that it may help overcome challenges with determining accurate volumetrics for less traditional and recycled materials.

As the push for BMD continues, it is worthwhile to assess from a global perspective, such as a state DOT network, resulting trends and outcomes from volumetric mix design. This type of overall assessment is anticipated to have meaningful value when shaping the direction of future mix design philosophies.

MDOT Mix Design Database

MDOT’s Central Laboratory is responsible for verifying mix design volumetrics and tentatively approving all asphalt mix designs for state projects (final approval is based on field verification). Central Lab currently documents and stores all approved mix designs in their SiteManager software, a practice that began in 2005. Pertinent mix design data from each individual approved mix design were extracted and compiled internally by MDOT for this analysis. The full database includes 2,353 mix designs from 2005 to 2018 with a variety of mix types, nominal maximum aggregate sizes (NMAS), and design gyration levels (N_des) based on traffic level. Seven drainage course and 64 Marshall designs were excluded, as these are outside MDOT’s normal practice.

Many cases existed where essentially the same design was reported for different binder sources or grades (e.g., PG 67-22 or 76-22) or different mix temperatures (e.g., hot mix asphalt [HMA] or warm mix asphalt [WMA]). As this might misrepresent certain analyses and trends for variables of interest, a derivative database, with duplicates removed, was created from the original. This sub-database consisting of only unique mix designs contains 1,452 designs and was used for all analysis in this paper. Further references to the MDOT database here refer to the sub-database containing 1,452 unique designs.

Table 1 provides MDOT database distributions by mix type. The database predominantly consists of dense graded asphalt (DGA) mixes (90%) followed by stone matrix asphalt (SMA) mixes (6%) then other special mixes. The database contains mostly 9.5, 12.5, and 19 mm mixes (475, 403, and 381, respectively) and only a few 4.75 and 25 mm mixes (21 and 28, respectively). There are approximately similar numbers of each gyration level for DGA mixes.

Table 1.

Summary Breakdown of 1,452 Mix Designs in Database

			Number of designs by NMAS
Mix type	Gyration level^b	Category	4.75	9.5	12.5	19	25	All
DGA	50	HMA	6	137	111	116	10	380
		WMA	NA	26	31	29	2	88
		All	6	163	142	145	12	468
	65	HMA	8	133	95	96	7	339
		WMA	1	17	18	18	NA	54
		All	9	150	113	114	7	393
	85	HMA	5	137	121	103	9	375
		WMA	1	25	27	19	NA	72
		All	6	162	148	122	9	447
	All		21	475	403	381	28	1308
SMA	75	HMA	NA	40	44	NA	NA	84
OGFC	50	HMA	NA	34	NA	NA	NA	34
UltraThin^a	50	HMA	NA	22	NA	NA	NA	22
		WMA	NA	4	NA	NA	NA	4
		All	NA	26	NA	NA	NA	26

Note: Binder grade use taken from the original database (2,353 designs) is as follows: PG 67-22, 1,794 mixes (76%); PG 76-22, 547 mixes (23%); PG 82-22, 12 mixes (<1%).

DGA = dense graded asphalt; HMA = hot mox asphalt; OGFC = open graded friction course; NA = not available; NMAS = nominal maximum aggregate size; SMA = stone matrix asphalt; WMA = warm mix asphalt.

UltraThin is MDOT’s specialty thin-lift mix. NMAS is between 4.75 and 9.5 mm. By normal classification, it is a 9.5 mm NMAS, but 6.4 mm NMAS is more fitting.

For DGA mixes, 50 gyr = standard traffic, 65 gyr = medium traffic, 85 gyr = high traffic.

Figure 3 summarizes information extracted from individual mix designs when compiling the database. Some variables apply only to certain mixes. For example, %Fiber only applies to SMA and open graded friction course (OGFC); likewise, RAP binder content (P_b) only applies to mixes containing RAP. Some variables typically reported for MDOT mix designs were not included because of data recording errors or irrelevance to volumetric trends.

Figure 3.

Summary of data compiled in mix design database.

Key Findings

The MDOT mix design database was analyzed to study prominent trends, relationships between various properties, and resulting implications. In several cases, properties from the database were used to calculate other parameters of possible interest that were not contained in the original mix design submittals; for example, gradations were used to calculate surface area (SA) factors. Several interesting trends were identified, and these trends, which generally relate to asphalt content in some manner, are the focus of discussion.

Interesting observations from the MDOT database were summarized into five key findings (KFs), listed below. Each key finding section presents relevant data and analysis followed by discussion of implications and potential approaches to address issues, if applicable.

(KF1) Voids in Mineral Aggregate (VMA)

(KF2) Aggregate Bulk Specific Gravity and Absorption (G_sb and Abs)

(KF3) RAP Content

(KF4) Design Gyration Level (N_des)

(KF5) Coarse versus Fine Gradation

Table 2 lists asphalt mixture specifications relevant to this paper’s KFs from the 14 Southeastern Asphalt User Producer Group (SEAUPG) states. Table 2 provides a discussion platform for evaluating MDOT trends in the following KF subsections and enables a broader discussion of findings based on other state DOT practices.

Table 2.

Summary of Southeastern Asphalt User Producer Group State Department of Transportation Standard Specifications.

										Design
	Aggregate G_sb					RAP			VMA_min (by NMAS)				Production
State	Coarse	Fine	P₂₀₀	Freq.	Tested by	Max.	Property for VMA	N_des levels	19.0	12.5	9.5	V_a	VMA	V_a
MS	T84	T85	W	At mix design	Contractor	30%	G_sb from extraction	50, 65, 85	13	14	15	4	−1.0^d	±1.0
AL	T84	T85	NS	At mix design	Contractor	35%	G_se	60	13.5	14.5	15.5	3.5 & 4	−0.5	±1.25
AR	T84	T85	NS	At mix design	Contractor/DOT A.L.	30%	R35	75, 115, 160, 205	—	14-16	15-17	4 & 4.5	Min. –0.5⁷	3-5
FL	T84	T85	U	1 Per 2 weeks	Contractor	>30%	R35	50, 65, 75, 100	13	14	15	4	NR	±1.2
GA^a	T84	T85	NS	At mix design	DOT or DOT A.L.	30%	G_se	50, 65	14	15	16	4	SD	NR
KY	T84	KM64-605	W	At mix design	Contractor	30%	G_sb from G_se	65	13	14	15	3.5	SD^d	1.5–6.0
LA	T84	T85	W	1 per year	DOT and Contractor	30%	G_sb from G_se	55, 65	12.5	13.5	—	3.5	SD	±1.0
NC	T84	T85	NS	At mix design	Contractor	45%^c	Gse	50, 65, 75, 100	13.5	—	15.5	4 & 5	SD^d	±1.0
OK	T84	T85	U	At mix design	Contractor	25%	G_sb from G_se	50, 65, 80	13.5	14.5	15.5	4	−0.5^f	±1.0
SC^b	T84	T85	NS	At mix design	Contractor	30%	R35	50, 75	13.5	14.5	15.5	2.5 – 9^e	±1.15	±1.15
TN	T84	T85	U	At mix design	Contractor	35%	R35	65 or 75 Blows	13	14	—	3 & 4	SD	2.0-5.5
TX^b	201F	201F	W	At mix design	Contractor	30%	G_sb from Extraction	50 or (35–100)	14	15	16	4	−0.5	±1.0
VA	T84	T85	NS	1 per year	Contractor	35%	G_se	50	14	15	16	2.5, 3.5, & 4	SD^g	1–4 or 2–5
WV	T84	T85	W	At mix design	Contractor	No limit	G_se	50, 65, 80, 100	13.5	14.5	15.5	4	−0.5^d	±1.2

Note: — NS = Not Specified; NR = Not Required; U = Unwashed; W = Washed; SD = Same as design; DOT A.L. = DOT approved lab; VMA = voids in mineral aggregate; NMAS = nominal maximum aggregate size; RAP = reclaimed asphalt pavement; Max. = maximum; Min. = minimum; G_b = asphalt binder specific gravity; G_mm = mixture maximum specific gravity; G_mb = mixture bulk specific gravity; G_sa = aggregate apparent specific gravity; G_sb = aggregate bulk specific gravity; G_se = aggregate effective specific gravity.

G_se is used for VMA calculations instead of G_sb.

VMA calculated as summation of V_a and V_b.

Value is Percent Recycled Binder Replacement.

Used as a production pay factor.

Total range allowed across all mixture types, actual span ~1% for each type (i.e., 3-4%).

VMA calculated using estimated G_sb equal to G_{se (production)} x G_{sb (mix design)} / G_{se (mix design)}.

VMA calculated using estimated G_sb equal to G_{se (production)} minus mix design G_se-to-G_sb correction.

There are several items of note in Table 2. AASHTO T84 and T85, or equivalent, are fairly standard for aggregate G_sb measurement. There is less consistency in the handling of material finer than 0.075 mm (P₂₀₀), and whether the fines are washed from the test sample or not can have considerable impacts on G_sb. Most interestingly, G_sb measurement is typically required only at mix design; in some cases, such as Mississippi, this mix design is considered valid indefinitely as long as it is produced once per year and meets other production requirements. This practice does not directly account for changes in aggregate properties like G_sb over time.

Maximum allowable RAP content is generally 30%. States differ considerably in relation to how G_sb is handled for RAP. Mississippi requires G_sb be measured on extracted RAP aggregate. Several states substitute G_se instead of G_sb, or some measure G_se, which is generally understood to be a simpler measurement, and then estimate G_sb using an estimated asphalt absorption value.

Most states have consolidated or reduced their N_des levels, or both, relative to AASHTO R35, and several have increased minimum VMA (VMA_min) or adjusted V_{a, des} requirements. In production, states generally require either the same VMA_min from mix design or reduce it slightly to accommodate production factors (e.g., aggregate breakdown). Table 2 footnotes summarize variations in VMA calculations, as not all VMA requirements in Table 2 are defined identically. For example, some states indirectly account for G_sb throughout production by measuring G_se in production and using a G_se-to-G_sb offset from mix design to estimate production G_sb and, in turn, VMA. This is valid if asphalt absorption does not change and asphalt content measurements are accurate and is in contrast to using the mix design G_sb, which could be multiple years old.

Key Finding #1: Voids in Mineral Aggregate (VMA)

VMA’s role and value as a design criterion has been thoroughly debated since its inception ( 33 ). Regardless of its true merit, the way VMA is currently used in Superpave design makes it one of the most important criteria for mix designers today, as it effectively determines an acceptable aggregate blend and design binder content.

Figure 4, a–c, show VMA distributions for MDOT 9.5, 12.5, and 19 mm DGA mixes relative to VMA_min, and Figure 4d shows the deviation from VMA_min (VMA_dev) for all DGA and SMA mixes. Around 45% of all mixes possess VMA values within 0.2% of VMA_min, and around 80% have VMA values within 0.6% of VMA_min. With VMA skewed heavily toward VMA_min, Figure 4 illustrates a trend of optimizing mixes based on VMA_min. Note that deviating too far from VMA_min (e.g., more than 2.0%) can cause other issues, though these are not discussed in this paper. The following subsections further explore VMA trends from two perspectives.

Figure 4.

Mississippi voids in mineral aggregate (VMA) distributions: (a) dense graded asphalt (DGA)—9.5 mm nominal maximum aggregate size (NMAS), (b) DGA—12.5 mm NMAS, (c) DGA—19 mm NMAS, and (d) All DGA and stone matrix asphalt (SMA).

VMA Regression to Investigate Effects of Aggregate Properties

Numerous studies have attempted to understand driving factors behind VMA, typically centered around gradation ( 34 – 36 ). The ability to predict VMA for a given aggregate blend would be of great value, as selecting an aggregate blend to meet VMA_min can be one of the more difficult and time-consuming mix design steps, as noted in Coree and Hislop ( 37 ). Unfortunately, efforts to tie VMA to gradation properties have generally been unfruitful aside from overarching trends ( 34 , 35 ). Knowing the previous challenges, the purpose of the regression analysis discussed here was (1) to leverage the large MDOT dataset and affirm or disaffirm previous findings, and (2) investigate whether other factors aside from gradation (e.g., gravel or RAP content) could be shown to affect VMA, particularly because factors like these may also have economic implications for mix designers.

Following the iterative regression approach of Williams et al. ( 38 ), which included correlation assessments, variable transformations and interactions, and multiple methods of regression model construction, a multivariate regression analysis was performed in SAS. VMA_dev was used as the dependent variable to normalize the influence of NMAS similar to Aschenbrener and MacKean ( 34 ).

Gradation (e.g., percent passing 2.36 mm), aggregate properties (e.g., G_sb, Abs, crushed content), and aggregate proportions (e.g., gravel content) were considered. Several additional variables computed from gradations were also considered. Percent passing deviation from the control point at the primary control sieve (PCS) (P_d(PCS)) was used to quantify coarse and fine gradations. Sums of the distances between the gradation and maximum density lines over a range of standard sieves (e.g., ΣP_{d(9.5 to 2.36)} for 9.5 to 2.36 mm) were computed. Aschenbrener and MacKean ( 34 ) summed the absolute values of distances, whereas Shen and Yu ( 36 ) simply summed the distances. Both were considered in this analysis.

Over a dozen regression models were produced using different subsets of variables; R²_adj values of those ranged from 0.08 at worst to 0.12 at best. Figure 5 shows equality plots of two representative models showing poor relationship, particularly as VMA_dev increases. Behaviors of individual variables in the regression equations were not always logical either, and some factors known to affect VMA such as dust content exhibited no relationship, whereas other less relevant factors like Abs repeatedly showed significant ties to VMA. This type of behavior signifies unreliable regression models and agrees that no variable exhibited any semblance of a one-to-one relationship with VMA. Similar to Huber and Shuler ( 35 ), this analysis affirmed that no specific relationships to VMA exist because of other factors like aggregate angularity and surface texture.

Figure 5.

Evaluating mix parameters that influence voids in mineral aggregate (VMA): (a) VMA_dev regression equation 1 and (b) VMA_dev regression equation 2.

VMA Sensitivity to G_sb

As shown in Figure 4, a large portion of MDOT’s mixes are near VMA_min, so it is worth considering G_sb’s effects, particularly variability, on VMA calculations (Eq. 1) where G_mb is mixture bulk specific gravity. AASHTO T84 (coarse aggregate) and T85 (fine aggregate) multilaboratory d2s limits (acceptable range for two results) are 0.038 and 0.066, respectively. For aggregate blends, these d2s limits are often combined; a 50/50 blend of coarse and fine aggregate yields a combined d2s limit of 0.052.

VMA = 100 - \frac{G_{mb} (1 - P_{b})}{G_{sb}}

(1)

To put the d2s limit in perspective, a G_sb error of only 0.006 would change VMA by 0.2%. Likewise, a G_sb error of 0.018, only about one-third the d2s limit, would change VMA by 0.6%. In reference to Figure 4, a 0.018 G_sb error equal to only one-third the allowable d2s limit would cause roughly 80% of MDOT’s mixes to calculate as failing VMA. Taken to the extreme, a 0.052 G_sb error equal to the d2s limit would cause 99% of MDOT’s mixes to fail VMA_min. This illustrates how vulnerable volumetric mix designs are to G_sb errors.

Following the logic in the preceding paragraph, almost all of MDOT’s currently approved mixes could actually be failing VMA according to their “true” G_sb but appear to pass based on VMA calculated using G_sb values on the higher end of those that could still be “acceptable” by d2s standards. It is unlikely that G_sb is consistently 0.052 g/cm³ too high; however, even with a modest inflation of 0.018, it is plausible that about 80% of MDOT’s mixes do not meet VMA requirements though they appear to do so.

Figure 6 illustrates that inflating G_sb increases the calculated VMA and reduces calculated P_ba, which will translate to lower P_{b, des} in practice as mix designers will likely adjust the mix slightly to bring calculated VMA back near VMA_min (actual VMA would then be lower than VMA_min). Although a 0.052 inflation could be used to reduce P_{b, des} approximately 0.8%, even the more modest 0.018 inflation could be used to drop P_{b, des} approximately 0.3%. Coupling this with trends from Figure 4, it becomes clear how lower asphalt contents than intended by volumetric design could easily overtake a large market. This issue is perpetuated by the infrequent G_sb testing rates typical of most states as discussed with Table 2; most only require G_sb measurement at mix design.

Figure 6.

Phase diagrams illustrating G_sb effects on voids in mineral aggregate (VMA) and P_ba.

Key Finding #2: Aggregate Bulk Specific Gravity and Absorption (G_sb and Abs)

Given G_sb’s cascading effects on other key properties, namely effective binder volume (V_be), this section discusses G_sb alongside water absorption (Abs). Figure 7 plots G_sb and Abs for Mississippi’s most common aggregates: crushed gravel, limestone, sand, and RAP. These aggregates are contained in 87, 77, 87, and 88% of mixes, respectively. By contrast, only 2% (generally “premium” mixes like SMA and OGFC) contain granite. For mixes that contain RAP, average percentages of gravel, limestone, sand, and RAP are 46, 23, 9, and 18%, respectively. Note that Figure 7 shows all data available for each aggregate type; in some cases, two aggregates of the same type (e.g., #67 and #11 limestone) may be used in a single mix design, meaning n in Figure 7 may exceed the 1,452 total number of mix designs in the database.

Figure 7.

G_sb and Abs distributions for most prevalent aggregate types: (a) crushed gravel G_sb, (b) crushed gravel Abs, (c) limestone G_sb, (d) limestone Abs, (e) Sand G_sb (includes coarse and man. sand), (f) Sand Abs (includes coarse and man. sand), (g) reclaimed asphalt pavement (RAP) G_sb, and (h) RAP Abs.

Overall spread in G_sb is similar for gravel and limestone, though gravel G_sb values are much lower overall. Gravel and limestone Abs are notably different, with some gravel values in excess of 4%. RAP G_sb and Abs fall approximately in the middle of the other three aggregates, which is logical as RAP represents older mixes likely containing gravel, limestone, and sand.

One caveat worth noting is that Figure 7 contains duplicate data, as multiple producers use the same aggregate sources for multiple mix designs. The assumption is that any duplication would occur uniformly across the state and that Figure 7 is representative. To support this assumption, Figure 8 provides a more systematic look at Mississippi crushed gravel data from Doyle et al. ( 39 ) where 77 gravels from 35 pits across the state are shown. This constitutes a more representative distribution of gravel statewide as duplication is minimized. Despite less resolution in Figure 8 as a result of fewer replicates, G_sb and Abs averages and overall distributions appear similar to Figure 7, a and b , suggesting Figure 7 reasonably represents Mississippi. The following two subsections further evaluate G_sb in more detail.

Figure 8.

Mississippi crushed gravel G_sb and Abs from Doyle et al. ( 39 ): (a) crushed gravel G_sb and (b) crushed gravel Abs.

Evaluating G_sb with Abs

Given the wide range of G_sb values observed in the MDOT database, it would be difficult to simply look at a G_sb value and decide whether or not it seems reasonable. However, as G_sb is vitally important to VMA and, thus, V_be, a method for evaluating reasonableness of G_sb would be valuable. This has its challenges. Huber and Pine ( 5 ) show that air voids (V_a) and VMA may look reasonable and P_{b, des} will seem out of place only if it is particularly low. Evaluating P_ba is not reliable either, as it will be unknown whether P_ba is low because of low-absorption aggregate or inflated G_sb.

It has been suggested that G_sb can be assessed by comparing P_ba with Abs using the estimate that P_ba will be approximately 50 to 80% of Abs depending on the Abs value (i.e., the P_ba-to-Abs relationship is nonlinear) ( 5 , 33 , 40 ). Along these lines, Huber and Pine ( 5 ) presented data for 28 mixes with Abs from 1.0 to 2.2% (regenerated here as Figure 9a) and suggested a systematic relationship. An initial assessment would seem to indicate that is so, but expanding this concept to the MDOT database provides greater clarity.

Figure 9.

Comparison of various P_ba and Abs relationships: (a) P_ba/Abs versus P_ba (5), (b) P_ba/Abs versus P_ba (1,452 Mississippi Department of Transportation [MDOT] mixes), (c) P_ba/Abs versus Abs (5), (d) P_ba/Abs versus Abs (1,452 MDOT mixes), (e) P_ba versus Abs (5), and (f) P_ba versus Abs (1,452 MDOT mixes).

Whereas Figure 9a appears defined, the same plot for all 1,452 MDOT mixes (Figure 9b) shows significant scatter (P_ba/Abs could range from 36 to 205 times P_ba). The relationship almost completely disappears when evaluating P_ba/Abs as a function of Abs in Figure 9, c and d , which would be the more appropriate comparison as Abs is a more appropriate independent variable than P_ba, as shown in Figure 9, a and b . Figure 9, e and f , show that a general relationship between P_ba and Abs does in fact exist, but scatter around the trendline prevents fruitful analysis with the P_ba/Abs parameter in Figure 9, a–d.

Interestingly, Figure 9f Abs values span the entire range of all three rules of thumb in Figure 9e, yet the P_ba is consistently about 33% of Abs regardless of Abs level. Not only is the relationship linear, but it is meaningfully lower than even the lowest 50% rule of thumb. Exact reasoning for these differences is not fully understood; however, Figure 9f is believed to be representative of Mississippi mixes based on an earlier assessment ( 39 ) of 568 MDOT mixes in which the P_ba to Abs relationship was also 33%.

Last, and most importantly, the concept of using P_ba and Abs to evaluate reasonableness operates under the premise that Abs is accurate. The pitfall of this approach is that G_sb and Abs are related; therefore, if G_sb is incorrectly inflated, Abs should be incorrectly deflated. Figure 10 illustrates this using a simple AASHTO T85 experiment in which a typical Mississippi crushed gravel was tested at varying moisture conditions from normal saturated surface dry (SSD) to wet (W) and very wet (VW) to dry (D) and very dry (VD) (moisture conditions were subjective).

Figure 10.

G_sb and Abs trends at varying moisture conditions.

Relative to the SSD condition, Figure 10 shows the VD condition results in a G_sb increase of 0.032 (within d2s limits) and an Abs decrease of 0.5% from 1.7 to 1.2%. Using the P_ba to Abs rules of thumb in Figure 9e, estimated P_ba for this gravel would drop from 1.1% (65% of 1.7) to 0.6% (50% of 1.2). Referring to Figure 6, the incorrect P_ba that would be calculated using a G_sb inflated by 0.032 would also be approximately 0.5% lower than the actual P_ba and match that estimated from Abs. The two errors would match and effectively cancel out if that Abs was determined at the same time as the incorrect G_sb value. The low (and incorrect) reported P_ba would go unnoticed. Though there are overarching trends, relying on Abs to validate G_sb as reasonable in any one specific case could be wildly misleading.

G_sb of RAP

Determining G_sb of RAP is not as straightforward as for virgin aggregate, and the issue has been well studied. RAP handling practices in Table 2 vary by state; in Mississippi, G_sb is determined on extracted aggregates, the effects of which have also been studied. Within the MDOT database, there is no direct means of comparing G_sb or Abs of extracted RAP aggregates to that of the original aggregate to determine extraction effects. However, the database size is sufficient to provide a global perspective by comparing RAP aggregate with combined aggregate blends.

The assumption behind this approach is that aggregate blends in the database are representative of that contained in the RAP, which would have been produced in years prior. Only mixes that contain RAP were considered, as virgin mixes tend to also contain less common aggregate types like steel slag or granite that would cause unrepresentative differences.

Figure 11 compares G_sb and Abs distributions between RAP aggregate and combined aggregate blends. RAP G_sb is, on average, 0.011 lower than that of the blend (statistically significant at 5% significance level). RAP Abs, at 1.42% on average, is slightly lower than blend Abs, and there appear to be more Abs values above around 2.0% for blends than for RAP. Again, Figure 11 is not a direct measure of the effects of extraction and recovery on G_sb and Abs, but results do agree with similar studies that have investigated this issue. For example, G_sb drop after ignition extraction has been reported as 0.005 to 0.066 (0.036 average) ( 41 ), 0.024 and 0.039 (fine and coarse aggregate, respectively) ( 42 ), 0.008 on average ( 43 ), and 0.036 to 0.087 ( 44 ). Ignition-extracted G_sb is generally lower than solvent-extracted G_sb (e.g., 0.040 on average) ( 44 ).

Figure 11.

Comparing reclaimed asphalt pavement (RAP) and blend G_sb and Abs distributions: (a) RAP G_sb versus blend G_sb (for >0% RAP mixes) and (b) RAP Abs versus blend Abs (for >0% RAP mixes).

The impact of RAP G_sb being slightly lower as a result of extraction is minimal on the overall blend. Given that the average RAP G_sb is 2.546 and the average RAP content is 18% for Mississippi RAP mixes, the G_sb for the remaining 82% virgin aggregate portion would need to be 2.558 to achieve a blend G_sb of 2.556 as in Figure 11a. This suggests that slight errors in RAP G_sb, at least on average, may not have a major bearing on overall blend G_sb and, ultimately, VMA. Error would increase as RAP content increases, but these errors (i.e., lower G_sb values) would actually benefit VMA during the design process. As noted in previous sections, this discussion must be considered in light of testing variability, either intentional or unintentional, that can significantly affect VMA.

Figure 12 shows the relationship between G_se and G_sb. The difference between the two is less at high gravities (i.e., low absorptions), but in the range of most Mississippi gravels, differences can be large. This has implications with respect to the states in Table 2 that use RAP G_se instead of determining G_sb. The error will be diluted as only a portion of the mix will be RAP (average maximum allowable RAP content in Table 2 is around 30%), and G_sb would be used for the virgin aggregate portion.

Figure 12.

Comparing G_se with G_sb.

If RAP G_sb and G_se are taken to be 2.545 and 2.592 (average values in MDOT database), respectively, and the same volumetric properties from Figure 6 are used, using G_se over G_sb increases calculated VMA by 0.16% for every 10% increase in RAP content. This shows that there can be a meaningful increase in calculated VMA when G_se is used in favor of G_sb and intuitively that difference grows as higher RAP contents are used (at 30% RAP, a 0.2% P_{b, des} reduction would be possible). As most mixes are designed as close to VMA_min as possible, the use of G_se for RAP could easily be a deciding factor in whether a mixture meets the required VMA, especially if the effect is compounded by other G_sb inflation errors.

Key Finding #3: RAP Content

Based on discussion surrounding Figure 11 in the previous section, RAP, on average, appears to have minimal effects on VMA with respect to G_sb. As a result, P_ba(mix), P_be, and P_b,des, all other factors being equal, would remain relatively unaffected even though P_b,Virgin will drop as a result of replacement with P_b,RAP. Figure 13, a and b , align with expectations in that P_ba(mix) remains relatively constant and P_b,Virgin drops with RAP content (about 0.9% per 10% RAP). Less intuitively, Figure 13, c and d , show that total asphalt content, as represented by V_be (normalized across NMAS to 12.5 mm NMAS, or 10% minimum V_be) and VMA_dev, drops with increasing RAP, meaning all other factors are not equal.

Figure 13.

Asphalt contents and VMA_dev versus reclaimed asphalt pavement (RAP) content: (a) P_ba(mix), (b) P_{b, Virgin}, (c) V_be, and (d) VMA_dev.

One example of unequal factors is binder stiffness. Along the lines of viscosity–temperature charts, the stiffer composite binder of a RAP mix would require higher mixing and compaction temperatures to maintain consistent volumetric properties. If mixing and compaction temperatures were not increased, then P_b,des should increase slightly. Zhou et al. ( 45 ) affirmed this with volumetrically designed 0 to 40% RAP mixes which required 0.5 to 1.0% higher P_b,des at 35, 40, and 15(5)% RAP(RAS) contents. DGA mixes in the MDOT database do not indicate higher mixing temperatures (average of 154°C for RAP mixes compared with 155°C for virgin mixes; range for either is identical at 138 to 171°C). This implies P_b,des might be expected to increase slightly with RAP content, if anything; however, this is not observed in Figure 13.

Although there is much scatter because of other variables not considered here, the overall trend in Figure 13c is that a 30% RAP mix contains 0.45% less V_be, or approximately 0.2% P_be, as a result of lower VMA. Similarly, Yu et al. ( 46 ) reported that P_{b, des} determined by Superpave volumetric design dropped 0.45 and 0.42% for the 20 and 40% RAP mixes in their study, respectively.

In practice, P_b,des decreasing with increasing RAP is an unintended consequence of volumetric-based design that is concerning. Stiffer RAP binder and the level of blending achieved between RAP and virgin binders (Mississippi assumes complete blending) are already areas of concern. In addition, a third of Table 2 states use RAP G_se for VMA calculations instead of G_sb, which already provides opportunity to reduce P_b,des (note that these states have generally increased VMA_min requirements slightly to help offset use of RAP G_se). Though any one factor by itself may be manageable, these factors considered collectively need to be addressed; volumetrics-only mix design, as it is used in practice, may not be optimal for addressing these factors, especially when considering the growing push to utilize an increasing variety and quantity of recycled materials.

Key Finding #4: Design Gyration Level (N_des)

In theory, decreasing N_des will increase design binder content. A 1% VMA increase can be achieved by decreasing N_des 25 to 30 gyrations, which would increase P_b,des approximately 0.4% ( 47 , 48 ). Buchanan ( 15 ) discusses an internal company survey in which 42% of the 30 DOTs surveyed lowered N_des, namely in an effort to increase binder content. Relative to the N_des levels in AASHTO M323 and R35, Table 2 shows nearly every SEAUPG state has reduced or consolidated N_des levels, or both.

In practice, decreasing N_des will do little to increase P_{b, des} because the concept is only valid if the aggregate blend remains unchanged. There is nothing that prevents a mix designer from using alternate aggregates and/or adjusting gradations, and they are extremely likely to do so to be competitive; this fact is often overlooked as noted by several others ( 15 , 47 , 48 ). Note that traffic volume, stiffness, and compactibility are other factors of changing N_des levels; although acknowledged, these are not discussed here in favor of a discussion on asphalt content as it relates to durability.

Figure 14a plots V_be for 9.5 to 19 mm MDOT DGA mixes by gyration level. By plotting V_be, the dependence of asphalt content on VMA becomes quickly evident; the lower end of all box plots is just slightly above the minimum V_be (i.e., VMA_min minus 4% V_a,des). V_be does increase slightly with each drop in N_des but only 0.11 to 0.15% on average; this equates to P_{b, des} increases of only 0.05 to 0.08%, which is minuscule compared with the 0.4% increase noted in literature for similar gyration reductions. This affirms what many have already stated—that economic factors in a competitive environment will result in N_des having little effect on asphalt content.

Figure 14.

N_des trends: (a) box plot of V_be by nominal maximum aggregate size (NMAS) and N_des, (b) gradations by N_des—9.5 mm NMAS, (c) gradations by N_des—12.5 mm NMAS, and (d) gradations by N_des—19 mm NMAS.

Figure 14, b–d, illustrate to some extent why V_be is not increasing as N_des decreases. For each NMAS, the average of all gradations shifts slightly toward the maximum density line (MDL) with decreasing N_des. Note that this is not an absolute observation; this is not a controlled data set, the gradation shifts are slight, and the rule of thumb that moving toward the maximum density line will decrease VMA is generally, but not always, true ( 35 , 49 , 50 ). However, Figure 14 is useful in showing that actual aggregate blends are not fixed but vary between gyration levels. Not only are slight gradation changes apparent, but VMA increases are marginal and nowhere near the 1% per 25–30 gyration estimate; together these affirm aggregate blends are being adjusted to design economical mixes just above VMA_min.

Key Finding #5: Coarse versus Fine Gradation

A common misconception exists that finer gradations will yield greater asphalt content and durability because they have a higher binder demand as a result of increased surface area. Several other studies have pointed out that with VMA_min criteria, finer gradations will not have higher asphalt contents because VMA is blind to surface area ( 5 , 51 , 52 ). The MDOT database provides a raw assessment of this issue because, as stated previously, the database is composed of mixes designed in a competitive, economics-driven environment.

Figure 15a shows the distribution of MDOT’s 9.5 to 19 mm DGA mixes with respect to coarse and fine gradations (determined by the PCS control point). The database primarily contains coarse-graded mixes but has a fair number fine-graded mixes as well. In all cases (Figure 15, b–d), average coarse and fine gradations are within 0.1% with respect to VMA, and thus V_be as well. There is effectively no change in design asphalt content as VMA_min is fixed.

Figure 15.

Coarse versus fine gradation trends: (a) distribution of fine and coarse gradations, (b) gradations by type—9.5 mm nominal maximum aggregate size (NMAS), (c) gradations by type—12.5 mm NMAS, and (d) gradations by type—19 mm NMAS.

Note that asphalt film thickness (FT) decreases slightly for fine gradations compared with coarse. This is logical, as surface area increases while binder volume is constant. From the early days of asphalt mix design, the relative merits of VMA over FT, or vice versa, have been debated. Campen et al. ( 51 ) and Kandhal et al. ( 52 ) favored FT though recommendations varied (e.g., 6 to 8 microns, 8 microns minimum, etc.); in particular, they argued that VMA may penalize an economical coarse gradation that does not meet minimum VMA but has sufficient film thickness for durability. Christensen and Bonaquist ( 53 ) recommend FT not be used for specifying or controlling mixes despite some indirect relationships with rutting. Similarly, Huber and Pine ( 5 ) presented bending beam fatigue data that indicated fatigue resistance was more directly related to V_be than FT. Data in the MDOT database is insufficient to support one view or the other; however, it is useful in quelling any lingering misconceptions that finer gradations will yield higher P_b,des. This will not occur in a volumetric mix design system where VMA_min is based solely off NMAS.

Discussion of Findings

The MDOT database affirmed several trends already documented in literature but with a more comprehensive dataset. Observing multiple trends simultaneously in a statewide dataset is believed to provide a strong validation of smaller datasets in literature consisting of a few mixes where perhaps only one or two issues are investigated within each dataset. The following paragraphs are a very high-level summary of findings affirmed by the MDOT database.

Within acceptable G_sb variability, 99% of Mississippi’s mixes could be manipulated to reduce asphalt content as much as 0.8% while still meeting calculated volumetric requirements. Other concepts to increase asphalt content (decreasing N_des or using finer gradations) were affirmed as ineffective with fixed VMA requirements. The MDOT database also showed that higher RAP content generally yields lower total asphalt content (approximately 0.2% lower for 30% RAP), which is perhaps less intuitive and not as well documented in literature.

Global characteristics observed in the MDOT database are a product, albeit unintended, of exclusive reliance on volumetrics. Given the realities illustrated by Figures 1 and 2, these issues will likely continue to be difficult to address exclusively with volumetrics. Mechanical tests are badly needed, and many practitioners and agencies are already working to this end with tests such as the Illinois Flexibility Index Test (I-FIT) ( 23 , 24 ), Disc-Shaped Compact Tension (DCT) ( 30 ), Cantabro Mass Loss (CML) ( 22 , 54 – 57 ), and others as described in Howard et al. ( 16 ).

Conclusions

Asphalt paving marketplace factors are very different now than when today’s volumetric mix design practices were developed. To investigate the implications of using volumetric mix design in today’s market, a database of 1,452 MDOT approved mix designs from 2005 to 2018 was compiled and analyzed, providing a holistic assessment of an entire state’s practice.

This paper discussed five key findings from the MDOT database that relate to volumetric mix design and dry mixes, many of which are, and will continue to be, difficult to address exclusively with volumetrics. In addition to numerous other smaller datasets in literature, this holistic statewide assessment of volumetric mix design in practice presents a compelling argument that mechanical tests are needed more now than when they were sought during SHRP.

MDOT has initiated a multi-year funded project denoted State Study 321 that aims to comprehensively evaluate the manner in which asphalt is designed, supplied, produced, placed, and accepted for MDOT projects. The goal of the study is to improve durability of the asphalt placed for MDOT while being economically mindful. A variety of potential enhancements are to be evaluated, including use of CML during design and construction (other mechanical testes could be evaluated as well). Once this multi-year study concludes, it is anticipated that MDOT will be in a position to determine their desire to incorporate additional mechanical testing into their standard practices.

Footnotes

Acknowledgements

Mississippi Department of Transportation supported this effort in a variety of manners. Permission to publish was granted by the Director, Geotechnical and Structures Laboratory, Engineer Research and Development Center.

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: BC, GS, IH, JE, AM; data collection: GS, AM, IH; analysis and interpretation of results: BC, GS, IH, JE, AM; draft manuscript preparation: BC, JE, IH. All authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Ben C. Cox

Jonathan Easterling

W. Griffin Sullivan

Isaac L. Howard

References

Richardson

The Modern Asphalt Pavement. John Wiley & Sons, London, 1905.

Goetz

W. H.

STP1041. The Evolution of Asphalt Concrete Mix Design. In Asphalt Concrete Mix Design: Development of More Rational Approaches ( Gartner

ed.), ASTM International, West Conshohocken, PA, 1989, pp. 5–14.

Huber

History of Asphalt Mix Design in North America, Part 1: From Hubbard to Marshall. Asphalt, Vol. 28, No. 1, 2013, pp. 15–20. http://asphaltmagazine.com/history-of-asphalt-mix-design-in-north-america-part-1/. Accessed July 25, 2020.

Huber

History of Asphalt Mix Design in North America, Part 2: Superpave. Asphalt, Vol. 28, No. 2, 2013, pp. 25–29. http://asphaltmagazine.com/history-of-asphalt-mix-design-in-north-america-part-2/. Accessed July 25, 2020.

Huber

Pine

Measurement of Effective Asphalt Content: Understanding How Much is Present. Proc., 61st Annual Conference of the Canadian Technical Asphalt Association (CTAA): Banff, Alberta, Canadian Technical Asphalt Association, Canada, 2016, pp. 315–344.

Roberts

F. L.

Mohammad

L. N.

Wang

L. B.

History of Hot Mix Asphalt Mixture Design in the United States. Journal of Materials in Civil Engineering, Vol. 14, No. 4, 2002, pp. 279–293.

Foster

C. R.

Symposium on Asphalt Paving Mixtures: Asphalt Stability Test Section. Research Report No. 7-B. Highway Research Board, National Research Council, Washington, D.C., 1949.

Heithaus

J. J.

Izatt

J. O.

Control of Marshall Stability of Asphalt Paving Mixes. Journal of the Association of Asphalt Paving Technologists, Vol. 30, 1961, pp. 223–237.

Leahy

R. B.

McGennis

R. B.

Asphalt Mixes: Materials, Design, and Characterization. Journal of the Association of Asphalt Paving Technologists, Vol. 68A, 1999, pp. 70–127.

10.

Cominsky

R. J.

The Superpave Mix Design Manual for New Construction and Overlays. Publication SHRP-A-407, Strategic Highway Research Program, National Research Council, Washington, D.C., 1994.

11.

Rice

J. M.

Volumetric Methods for Measuring Asphalt Content and Effective Gravity of Aggregates in Bituminous Mixtures. Journal of the Association of Asphalt Paving Technologists, Vol. 22, 1953, pp. 284–301.

12.

Rice

J. M.

Volumetric Specific Gravity of Bituminous Mixtures by Vacuum Saturation Procedure. ASTM International STP191-EB. ASTM International, 1956, pp. 43–61.

13.

McLeod

N. W.

STP252-EB. Void Requirements for Dense-Graded Bituminous Paving Mixtures. In Bituminous Paving Materials ( Rader

ed.), ASTM International, West Conshohocken, PA, 1959, pp. 81–112.

14.

McLeod

N. W.

Designing Standard Asphalt Paving Mixtures for Greater Durability. Proc., 16th Annual Conference of the Canadian Technical Asphalt Association (CTAA), Canadian Technical Asphalt Association, 1971, pp. 53–93.

15.

Buchanan

Optimized Mix Design Approach – Contractor’s Perspective. Journal of the Association of Asphalt Paving Technologists, Vol. 85, 2016, pp. 819–837.

16.

Howard

I. L.

Baumgardner

G. L.

Monismith

C. L.

Asphalt Mix Cracking State of the Art: Materials, Design, and Testing. Journal of the Association of Asphalt Paving Technologists, Vol. 85, 2016, pp. 777–806.

17.

Gibson

N. H.

Xinjun

Measured and Predicted Performance Impacts From Reduced Design Gyrations. Journal of the Association of Asphalt Paving Technologists, Vol. 81, 2012, pp. 337–367.

18.

Tran

Huber

Leiva

Pine

Yin

Mix Design Strategies for Improving Asphalt Mixture Performance. NCAT Report 19-08, National Center for Asphalt Technology, Auburn, AL, 2019.

19.

Lynn

James

R. S.

P. Y.

Jared

D. M.

Effect of Aggregate Degradation on Volumetric Properties of Georgia’s Hot-Mix Asphalt. Transportation Research Record: Journal of the Transportation Research Board, 2007. 1998: 123–131.

20.

Hekmatfar

Shah

Huber

McDaniel

Haddock

J. E.

Modifying Laboratory Mixture Design to Improve Field Compaction. Road Materials and Pavement Design, Vol. 16, Supplement 2, 2015, pp. 149–167.

21.

Bonaquist

Impact of Mix Design on Asphalt Pavement Durability. Transportation Research Circular E-C186. Transportation Research Board, Washington, D.C., 2014, pp. 1–17.

22.

Cox

B. C.

Smith

B. T.

Howard

I. L.

James

R. S.

State of Knowledge for Cantabro Testing of Dense Graded Asphalt. Journal of Materials in Civil Engineering, Vol. 29, No. 10, 2017, p. 04017174.

23.

Ozer

Al-Qadi

I. L.

Lambros

El-Khatib

Singhvi

Doll

Development of a Fracture-Based Flexibility Index for Asphalt Concrete Cracking Potential Using Modified Semi-Circle Bending Test Parameters. Construction and Building Materials, Vol. 115, 2016, pp. 390–401.

24.

Ozer

Al-Qadi

I. L.

Singhvi

Khan

Rivera-Perez

El-Khatib

Fracture Characterization of Asphalt Mixtures with High Recycled Content Using Illinois Semicircular Bending Test Method and Flexibility Index. Transportation Research Record: Journal of the Transportation Research Board, 2016. 2575: 130–137.

25.

Roque

Birgisson

Drakos

Dietrich

Development and Field Evaluation of Energy-Based Criteria for Top-Down Cracking Performance of Hot Mix Asphalt. Journal of the Association of Asphalt Paving Technologists, Vol. 73, 2004, pp. 229–260.

26.

West

Van Winkle

Maghsoodloo

Dixon

Relationships Between Simple Asphalt Mixture Cracking Tests Using N_design Specimens and Fatigue Cracking at FHWA’s Accelerated Loading Facility. Road Materials and Pavement Design, Vol. 18, No. 4, 2017, pp. 428–446.

27.

West

Rodezno

Leiva

Yin

Development of a Framework for Balanced Mix Design. Final Report for NCHRP Project 20-07/Task 406. Transportation Research Board, Washington, D.C., 2018.

28.

Zhou

Sun

Scullion

Development of an IDEAL Cracking Test for Asphalt Mix Design and QC/QA. Road Materials and Pavement Design, Vol. 18, No. 4, 2017, pp. 405–427.

29.

Zhou

Crockford

Zhang

Epps

Sun

Development and Validation of an Ideal Shear Rutting Test for Asphalt Mix Design and QC/QA. Journal of the Association of Asphalt Paving Technologists, Vol. 88, 2019, pp. 719–750.

30.

Buttlar

W. G.

Hill

B. C.

Wang

Mogawer

Performance-Space Diagram for the Evaluation of High and Low Temperature Asphalt Mixture Performance. Journal of the Association of Asphalt Paving Technologists, Vol. 85, 2016, pp. 519–547.

31.

Cooper

S.B.

III Mohammad

L. N.

Kabir

King

Jr.

Balanced Asphalt Mixture Design through Specification Modification: Louisiana’s Experience. Transportation Research Record: Journal of the Transportation Research Board, 2014. 2447: 92–100.

32.

Zhou

Scullion

Development and Implementation of Balanced Mix Design in Texas. Journal of the Association of Asphalt Paving Technologists, Vol. 85, 2016, pp. 839–852.

33.

Hinrichsen

Heggen

Minimum VMA in HMA Based on Gradation and Volumetric Properties. Transportation Research Record: Journal of the Transportation Research Board, 1996. 1545: 75–79.

34.

Aschenbrener

Mac Kean

Factors That Affect the Voids in the Mineral Aggregate of Hot-Mix Asphalt. Transportation Research Record: Journal of the Transportation Research Board, 1994. 1469: 1–8.

35.

Huber

G. A.

Shuler

T. S.

STP1147. Providing Sufficient Void Space for Asphalt Cement: Relationship of Mineral Aggregate Voids and Aggregate Gradation. In Effects of Aggregates and Mineral Fillers on Asphalt Mixture Performance ( R.

Meininger

ed.), ASTM International, West Conshohocken, PA, 1992, pp. 225–251.

36.

Shen

Analysis of Aggregate Gradation and Packing for Easy Estimation of Hot-Mix-Asphalt Voids in Mineral Aggregate. Journal of Materials in Civil Engineering, Vol. 23, No. 5, 2011, pp. 664–672.

37.

Coree

B. J.

Hislop

W. P.

Difficult Nature of Minimum Voids in the Mineral Aggregate: Historical Perspective. Transportation Research Record: Journal of the Transportation Research Board, 1999. 1681: 148–156.

38.

Williams

K. L.

Cox

B. C.

Howard

I. L.

Cooley

L. A.

Jr.

Models of Asphalt Concrete Field Compactibility with Focus on Lift Thickness. Transportation Research Record: Journal of the Transportation Research Board, 2015. 2504: 135–147.

39.

Doyle

J. D.

Howard

I. L.

Robinson

W. J.

Prediction of Absorbed, Inert, and Effective Bituminous Quantities in Reclaimed Asphalt Pavement. Journal of Materials in Civil Engineering, Vol. 24, No. 1, 2012, pp. 102–112.

40.

Copeland

Reclaimed Asphalt Pavement in Asphalt Mixtures. Publication FHWA-HRT-11-021, Federal Highway Administration, McLean, VA, 2011.

41.

Hall

K. D.

Williams

S. G.

Effects of the Ignition Method on Aggregate Properties. Journal of the Association of Asphalt Paving Technologists, Vol. 68, 1999, pp. 574–588.

42.

Prowell

B. D.

Carter

C. B.

Effect of the Ignition Furnace on Aggregate Properties of Recycled Asphalt Pavement. Proc., 11th AAPA International Flexible Pavements Conference, Sydney, Australia, 2000.

43.

Kvasnak

West

Michael

Loria

Hajj

E. Y.

Tran

Bulk Specific Gravity of Reclaimed Asphalt Pavement Aggregate: Evaluating the Effect on Voids in Mineral Aggregate. Transportation Research Record: Journal of the Transportation Research Board, 2010. 2180: 30–35.

44.

Xie

Taylor

A. J.

Tran

Kim

Evaluation of Different VMA Estimation Methods for Asphalt Mixtures Containing RAP. Journal of Materials in Civil Engineering, Vol. 32, No. 7, 2020, p. 04020162.

45.

Zhou

Das

Lee

Scullion

Claros

Successful High RAP Mixes Designed with Balanced Rutting and Cracking Requirements. Journal of the Association of Asphalt Paving Technologists, Vol. 81, 2012, pp. 477–505.

46.

Shen

Zhou

Effect of Partial Blending on High Content Reclaimed Asphalt Pavement (RAP) Mix Design and Mixture Properties. Transportation Research Record: Journal of the Transportation Research Board, 2018. 2672: 79–87.

47.

Huber

G. A.

Anderson

R. M.

Superpave Design Compaction Effort: Validity of Using Density at the End of Service Life as Parameter to Define N-Design. Journal of the Association of Asphalt Paving Technologists, Vol. 73, 2004, pp. 807–835.

48.

Prowell

B. D.

Brown

E. R.

Verification of the N Design Levels. Journal of the Association of Asphalt Paving Technologists, Vol. 76, 2007, pp. 771–821.

49.

Anderson

R. M.

Bentsen

R. A.

Influence of Voids in the Mineral Aggregate (VMA) on the Mechanical Properties of Coarse and Fine Asphalt Mixtures. Journal of the Association of Asphalt Paving Technologists, Vol. 70, 2001, pp. 1–37.

50.

Goode

J. F.

Lufsey

L. A.

A New Graphical Chart for Evaluating Aggregate Gradations. Journal of the Association of Asphalt Paving Technologists, Vol. 31, 1962, pp. 176–207.

51.

Campen

W. H.

Smith

J. R.

Erickson

L. G.

Mertz

L. R.

The Relationships Between Voids, Surface Area, Film Thickness, and Stability in Bituminous Paving Mixtures. Journal of the Association of Asphalt Paving Technologists, Vol. 28, 1959, pp. 149–178.

52.

Kandhal

P. S.

Foo

K. Y.

Mallick

R. B.

Critical Review of Voids in Mineral Aggregate Requirements in Superpave. Transportation Research Record: Journal of the Transportation Research Board, 1998. 1609: 21–27.

53.

Christensen

D. W.

Jr. Bonaquist

R. F.

NCHRP Report 567: Volumetric Requirements for Superpave Mix Design. Transportation Research Board, Washington, D.C., 2006.

54.

Smith

B. T.

Howard

I. L.

Short- and Longer-Term Effects of Short-Term Aging on Asphalt Mixture Properties. Journal of Materials in Civil Engineering, Vol. 30, No. 3, 2018, p. 04018005.

55.

Bazuhair

R. W.

Pittman

C. V.

Howard

I. L.

Jordan

W. S.

III Hemsley

J. M.

Baumgardner

G. L.

Conditioning and Testing Protocol Combinations to Detect Asphalt Mixture Damage. Transportation Research Record: Journal of the Transportation Research Board, 2018. 2672: 10–21.

56.

Bazuhair

R. W.

Easterling

Cox

B. C.

Howard

I. L.

Assessing the Micro-Deval Test for Compacted Dense-Graded Asphalt. Advances in Civil Engineering Materials, Vol. 8, No. 1, 2019, pp. 670–687.

57.

Bazuhair

R. W.

Howard

I. L.

Middleton

Jordan

W. S.

III Cox

B. C.

Combined Effects of Oxidation, Moisture, and Freeze-Thaw on Asphalt Mixtures. Transportation Research Record: Journal of the Transportation Research Board, 2020. 2674: 409–424.

Data Mining Statewide Department of Transportation Volumetrically Designed Asphalt Mixture Records

Abstract

Keywords

Mix Design Background

MDOT Mix Design Database

Key Findings

Key Finding #1: Voids in Mineral Aggregate (VMA)

VMA Regression to Investigate Effects of Aggregate Properties

VMA Sensitivity to Gsb

Key Finding #2: Aggregate Bulk Specific Gravity and Absorption (Gsb and Abs)

Evaluating Gsb with Abs

Gsb of RAP

Key Finding #3: RAP Content

Key Finding #4: Design Gyration Level (Ndes)

Key Finding #5: Coarse versus Fine Gradation

Discussion of Findings

Conclusions

Footnotes

Acknowledgements

Author Contributions

Declaration of Conflicting Interests

Funding

ORCID iDs

References

VMA Sensitivity to G_sb

Key Finding #2: Aggregate Bulk Specific Gravity and Absorption (G_sb and Abs)

Evaluating G_sb with Abs

G_sb of RAP

Key Finding #4: Design Gyration Level (N_des)