1411

Effects of surface preparation, thickness, and material on asphalt pavement overlay transverse crack propagation Feng Hong and Dar-Hao Chen

Abstract: Asphalt overlay has been widely used in pavement rehabilitation. The most frequently observed distress in an asphalt overlay is transverse cracking. In this study, the Texas long-term pavement performance (LTPP) specific pavement study 5 (SPS-5) test sections are highlighted. Three key factors affecting overlay cracking performance are investigated: (i) surface preparation, (ii) overlay thickness, and (iii) material. A deterioration model is developed to evaluate the effects of these factors. The deterioration process is well captured by incorporating both engineering principles and statistical modeling techniques, and the effects of the three key factors are thoroughly evaluated. The results suggest that (i) milling of existing pavement does not help reduce overlay transverse cracking if existing cracking is not completely removed, (ii) thicker overlay contributes to transverse crack resistance, (iii) Texas type C asphalt mixture is more effective in resisting transverse cracks than type B asphalt mixture (with coarser aggregate and less binder than type C), and (iv) overlays that incorporate reclaimed asphalt pavement (RAP) are more prone to transverse cracking than virgin asphalt. Furthermore, these effects are quantified based on the model estimation results. Key words: deterioration modeling, pavement, asphalt overlay, cracking, RAP, LTPP. Re´sume´ : Le reveˆtement bitumineux a e´te´ largement utilise´ lors de la re´fection des chausse´es. Les de´gaˆts les plus souvent observe´s dans le reveˆtement bitumineux est la fissuration transversale. Dans la pre´sente e´tude, les sections d’essai de l’e´tude 5 (SPS-5Le) au Texas sur le rendement des chausse´es a` long terme (« LTPP ») sont souligne´es. Trois parame`tres affectant le rendement du reveˆtement contre la fissuration sont examine´s : (i) la pre´paration de la surface, (ii) l’e´paisseur du reveˆtement et (iii) les mate´riaux. Un mode`le de de´te´rioration est mis sur pied afin d’e´valuer les effets de ces parame`tres. Le processus de de´te´rioration est bien capte´ graˆce a` l’incorporation des principes d’inge´nierie et des techniques de mode´lisation statistique; les effets de ces trois facteurs cle´s sont e´tudie´s en de´tails. Les re´sultats sugge`rent que (i) le fraisage de la chausse´e existant n’aide pas a` re´duire la fissuration transversale du reveˆtement si la fissuration existante n’est pas comple`tement e´limine´e, (ii) un reveˆtement plus e´pais contribue a` la re´sistance contre la fissuration transversale, (iii) le me´lange bitumineux Texas type C est plus efficace pour re´sister a` la fissuration transversale que le me´lange bitumineux de type B (pre´sentant des agre´gats plus grossiers et moins de liant que le type C) et (iv) les reveˆtements incorporant de l’asphalte re´cupe´re´e (« RAP ») sont plus sujets a` la fissuration transversale que l’asphalte neuve. De plus, ces effets sont quantifie´s en se basant sur les re´sultats estimatifs du mode`le. Mots-cle´s : mode`le de de´te´rioration, chausse´e, reveˆtement bitumineux, fissuration, RAP, LTPP. [Traduit par la Re´daction]

Introduction Pavement is the most common element of the transportation infrastructure and is built to provide a safe and comfortable ride for the public. To maintain a pavement system with an acceptable ride quality, state highway agencies (SHA) invest significant funds in maintenance and rehabilitation (M&R) activities annually (Haas et al. 1994; Tayabji Received 28 December 2008. Revision accepted 8 May 2009. Published on the NRC Research Press Web site at cjce.nrc.ca on 11 September 2009. F. Hong1 and D. Chen. F. Hong1 and D.-H. Chen. Texas Department of Transportation, 4203 Bull Creek No. 39, Austin, TX 78731, USA. Written discussion of this article is welcomed and will be received by the Editor until 31 January 2010. 1Corresponding

author (e-mail: [email protected]).

Can. J. Civ. Eng. 36: 1411–1420 (2009)

et al. 2000). Asphalt overlays have been widely used in pavement M&R to extend the service life or restore both structural and functional performance of existing pavements (Arudi et al. 1996; Wen et al. 2006; Loria et al. 2008). As an added structural layer in a pavement, the performance of an overlay is related to a series of factors, which include the existing pavement conditions, traffic loading, environmental effects, overlay material properties, and others. As a result, different types of distress may occur in the overlay during its service life, and cracking is the most frequently observed. Asphalt pavement cracking occurs in three typical categories: transverse, longitudinal, and alligator. As the most widely observed type of cracking in asphalt overlays, transverse cracking is investigated in this study, with an example presented in Fig. 1. A transverse crack can be a reflection of a preexisting crack in the underlying layers and can also be caused by temperature-related contraction of the overlay material (Huang 2003). For an asphalt overlay on existing asphalt pavement, these two mechanisms jointly contribute to

doi:10.1139/L09-080

Published by NRC Research Press

1412 Fig. 1. Transverse cracking observed in asphalt overlay.

the formation of transverse cracks, which are often randomly distributed in the overlay. Thus, it is difficult to tell from surface distress alone whether a transverse crack is primarily reflective or temperature-specific. This research investigates the propagation (by length) of overall transverse cracking, although it is believed that the majority of transverse cracking falls into the reflection cracking category at the locations under study. Penetration of water through the cracks into pavement can result in a loss of bond between the overlay and underlying layer, stripping in asphalt mixtures, and softening of the base and subgrade layers (AASHTO 1993). In engineering practice, different measures have been suggested to mitigate crack propagation in the overlay. These measures include pre-overlay repair, increasing overlay thickness, using softer asphalt binder, and adopting synthetic fabrics. However, the economic effectiveness of these measures remains to be verified (FHWA 2003). With the advancement of in-place recycling and reclamation technology, recycled asphalt material has been increasingly used in overlay projects (Tayabji et al. 2000). The effectiveness of these materials has yet to be determined. A better understanding of the cracking mechanism is imperative before solutions to these issues can be found. Because of the complexity of the propagation of transverse cracking in asphalt concrete (AC) overlays, most prior studies have relied on either laboratory testing or engineering experience in the field. Laboratory testing can only partially predict real-world results, since the tests are under controlled and simplified conditions, unlike those of the inservice environment and traffic. Although engineering experience can provide awareness of a cracking mechanism in the field, it cannot be used to accurately quantify pavement performance in relation to the influencing factors. In this regard, in-service pavement data such as those collected in this study are preferred. The paper is organized in sections as follows: (i) the data source is presented, with a focus on experimental design and pavement cracking distress; (ii) a description of the methodology is presented, with a focus on the development of a deterioration model for transverse crack propagation; (iii) the model estimation results are presented and their implications are discussed; (iv) a pavement performance and life comparison is conducted based on the established model; and (v) the key findings from the study are summarized and conclusions are presented.

Can. J. Civ. Eng. Vol. 36, 2009

Data source, experimental design, and pavement cracking performance Introduction to the long-term pavement performance (LTPP) program To facilitate understanding of the in-service pavement performance, the long-term pavement performance (LTPP) program was launched in 1987 through the strategic highway research program (SHRP). After about 20 years of data collection, the LTPP program has established the largest inservice pavement database in the world. According to the LTPP program, there are two types of in-service pavement studies, namely the general pavement study (GPS) and the specific pavement study (SPS) (FHWA 2003). The GPS includes nine experimental categories (GPS-1 to GPS-9), which aim to develop a comprehensive national pavement performance database. The SPS also includes nine experimental categories (SPS-1 to SPS-9), which focus on specially constructed, maintained, or rehabilitated pavement sections for a more detailed and complete study of pavement performance. For the purpose of this study, SPS experiment category 5 (SPS-5, entitled ‘‘Rehabilitation of AC Pavements’’) is highlighted herein. SPS-5 test sections in Texas The Texas SPS-5 experiment covers a segment of U.S. Highway 175 and is located about 25 miles (40 km) southeast of Dallas. The existing pavement structure (before overlay) and new structure after overlay are presented in Fig. 2. These test sections are subject to real traffic and thus reflect pavement deterioration under real-world conditions. This SPS experiment was initiated in 1992 and was divided into eight sections, each with a specific asphalt overlay. Following the standard LTPP test configuration, each section is 500 ft. (152.4 m) long and 12 ft. (4 m) wide. Two representative sections are shown in Fig. 3. A 100 ft. (30.3 m) transition area was constructed between adjacent sections, and thus it is reasonable to assume that these sections are independent of each other from a construction viewpoint. Pavement performance in each section has been continuously monitored since construction of the overlay. After quality assurance (QC) inspection, the dataset was published through the Federal Highway Administration (FHWA) LTPP standard data release and Web site (www.datapave.com) (FHWA 2008). Experimental design Three key factors encountered in AC overlay work were intentionally included in the SPS-5 study, namely surface preparation for overlay, overlay thickness, and overlay material, as summarized in Table 1 and explained as follows (SHRP 1989): (1) Surface preparation for overlay — It is believed that overlay performance is related to the surface preparation of the overlay (AASHTO 1993). For example, removal of existing distress can contribute to better performance of the overlay. In this study, the surface preparation is focused on milling the existing pavement before placing the overlay. To compare the effect of milling on overlay Published by NRC Research Press

Hong and Chen

1413

Fig. 2. Existing pavement structure before and after rehabilitation: (a) no milling; (b) milling. Note that the layer thicknesses are not to scale.

Fig. 3. Two representative pavement sections under study.

performance, both milled and nonmilled sections were included in the experimental design. On four of the eight sections, 2 in. (50.8 mm) of the existing surface AC was milled. Before the overlay placement, the milled thickness was replaced by the same material as that used in the overlay (a mill-and-fill operation). For the milled sections, the design thickness does not include the replaced asphalt mixture. For the remaining sections without milling, overlay material was placed directly on the existing pavement. (2) Overlay thickness — According to the SHRP SPS-5 experimental plan, two overlay design thicknesses were

used, namely thin (2 in. or 50.8 mm) and thick (5 in. or 127.0 mm) (SHRP 1989). From a structural viewpoint, the replaced 2 in. (50.8 mm) thickness needs to be accounted for in the total overlay thickness. The overlay is constructed with either one course (surface course) or two courses (surface and binder courses). Based on the traffic level at the test site, the surface course is composed of type C (TYC) AC, and the binder course is composed of type B (TYB) AC (Texas Department of Transportation 2004). Both types of AC belong to dense-graded hot mixture asphalt and differ in aggregate size and asphalt content. The aggregate in TYB AC is Published by NRC Research Press

1414

Can. J. Civ. Eng. Vol. 36, 2009 Table 1. Experimental design factors at Texas SPS-5 sections. Section ID A502 A503 A504 A505 A506 A507 A508 A509

Construction preparation No mill No mill No mill No mill Mill Mill Mill Mill

Overlay thickness 2.2 in. (55.9 mm) TYC 2.1 in. (53.3 mm) TYC + 2.2 in. (55.9 mm) TYC + 2.0 in. (50.8 mm) TYC 4.3 in. (109.2 mm) TYC 2.0 in. (50.8 mm) TYC + 2.1 in. (53.3 mm) TYC + 2.2 in. (55.9 mm) TYC +

Fig. 4. Sigmoid curve to capture crack length propagation with time.

3.2 in. (81.3 mm) TYB 3.0 in. (76.2 mm) TYB

5.0 in. (127.0 mm) TYB 5.2 in. (132.1 mm) TYB 2.1 in. (53.3 mm) TYB

AC material 35% RAP 35% RAP Virgin Virgin Virgin Virgin 35% RAP 35% RAP

Pavement performance in transverse cracking There are two units of pavement distress to quantify transverse cracking for each individual section, namely number of cracks and total crack length. Two concerns may arise in counting the number of cracks: (i) this method of counting does not differentiate between long and short cracks; and (ii) the joining of adjacent cracks may reduce the number of cracks, which would be recorded as an ‘‘improvement’’ of pavement performance when in fact the pavement has deteriorated. Therefore, the total transverse cracking length is used in this study. Since the overlay construction in 1992, data collection was conducted 11 times until 2007. Thus, the sample size is 88 for the eight sections involved in this study.

Model specification

coarser than that in TYC (with a maximum sieve size of 1 in. or 25.4 mm for TYB versus 0.75 in. or 19.1 mm for TYC). The asphalt content in the former is usually lower than that in the latter, e.g., around 0.6% lower in TYB than in TYC AC in this study. (3) Overlay material — Environmental pressure, the depletion of natural resources, and the rising cost of asphaltic materials emphasize the necessity of exploring the use of reclaimed asphalt pavement (RAP). SHRP foresaw this trend and incorporated RAP as one of the major experimental factors in the SPS-5 study. In engineering practice involving RAP, used and new asphalt concrete are mixed, in certain percentages, to produce a ‘‘combined’’ mixture for pavement construction. Among the eight sections, four included RAP. Based on current pavement construction practice, the maximum RAP content (i.e., 35% recycled AC) was adopted in the study. The overlay AC material is referred to as ‘‘virgin’’ in the other four sections not containing RAP. Color can be used to identify whether a pavement overlay is composed of virgin material or RAP. The virgin material is usually dark (test section A504, Fig. 3), whereas the RAP material is relatively light (test section A508, Fig. 1). A preliminary study was conducted by Chen and Daleiden (2005) on the overlay performance of RAP after 10 years of service. They suggested that overlay performance with RAP can meet engineering requirements. This study will fine-tune the previous work through more detailed analysis based on a longer span of service history.

Empirical or statistical techniques are commonly used to model highway infrastructure deterioration (Haas et al. 1994; Prozzi and Madanat 2003; Hong et al. 2008). The statistical deterioration modeling is composed of two hierarchical procedures, namely model specification development and model estimation. Although different combinations of mathematical terms can be used to fit the observed data, a specification properly describing the deterioration mechanism is preferred. The selection of the appropriate model specification relies heavily on engineering judgment on the problem of interest. Basic model As usual, deterioration modeling starts with a basic model depicting the general trend of the deterioration process for a particular distress. Sigmoid or S-shaped models have been widely used in predicting pavement performance due to their flexibility in fitting almost-linear, concave, convex, and mixed pavement deterioration processes (Garcia-Diaz and Riggins 1984; Haas et al. 1994; Zaghloul et al. 2008). Based on field observations and engineering experience, sigmoid or S-shaped curves are found capable of capturing the development of asphalt pavement cracking (NCHRP 2004). The current Mechanistic–empirical pavement design guide (MEPDG) available from the U.S. Transportation Research Board adopts sigmoid functions to model pavement cracking propagation for different types of cracking, including topdown cracking, bottom-up cracking, and reflective cracking. It is noted that the effect of key factors, including surface preparation, thickness, and materials, is not systematically and sufficiently addressed in these models. Published by NRC Research Press

Hong and Chen

1415 Table 2. Model estimation results. Variable Constant Constant Constant ML RP THB THC

Parameter b1 b2 b3 b4 b5 b6 b7

Mean 1.16102 5.32 –9.6510–1 –1.1510–1 –4.4710–1 1.3310–1 2.1010–1

t-statistic 19.93 9.38 –8.18 –3.43 –7.19 7.70 6.12

Note: The goodness-of-fit (R2) statistic is 0.92.

Fig. 5. Observed transverse cracking versus that predicted by the model: (a) all sections; (b) example (section A502).

By adopting the sigmoid function as given in the following equation, it is implied that with time the quantity or extent of cracks, e.g., number of cracks or total crack length of a given pavement section, increases until a stable asymptote is reached: ½1

f ðxÞ ¼

b1 1 þ expðb2 þ b3 xÞ

where x is a variable including relevant pavement information, and b1–b3 are parameters as explained later in the paper. The maximum value is the numerator term b1. The exponential index in the denominator affects the steepness of the curve. A larger value of this term indicates a slower rate of transverse crack development Fig. 4 illustrates a typical sigmoid function in the context of asphalt pavement crack propagation.

Published by NRC Research Press

1416

Can. J. Civ. Eng. Vol. 36, 2009

Fig. 6. Pavement core from test section A509 indicating transverse cracking in existing and overlay structures.

Full model specification development With the basic model format established, the subsequent task is to fill in the basic model with the variables associated with relevant pavement information based on engineering judgment. Regarding the numerator term b1, it is termed as follows: ½2

b1 ¼ b1

where b1 is a parameter to be estimated. It is implied that the maximum total length of transverse cracking approaches b1 . Regarding the denominator term, the exponential index in eq. [1] remains to be modeled particularly. Based on engineering judgment, a nonlinear function is adopted to incorporate different factors as explanatory variables in the model, as is shown in the following term:

½3

b2 þ b3 x ¼ b2 þ ½b3 þ b4 MLi þ expðb5 RPi Þðb6 THBi þ b7 THCi ÞTi

where ML is a dummy variable representing mill (ML = 1) or no mill (ML = 0), RP is a dummy variable representing RAP material (RP = 1) and virgin material (RP = 0), THB represents the thickness of type B AC in the overlay (in. or mm), THC represents the thickness of type C AC in the overlay (in. or mm), T is the time in years after overlay construction, and b2–b9 are parameters to be estimated. The term exp(b5RPi)(b6THBi + b7THCi) deserves further discussion. This term is modeled in this way for two purposes. First, it serves to calculate the relative effectiveness of different AC mixtures (type B versus type C) in terms of their ability to resist transverse cracking. It can be shown that the relative effectiveness can be represented as the ratio of their coefficients, i.e., b6/b7. It is expected that both parameters will be larger than zero in the context of the full model, since thicker pavement is believed to be more capable of resisting transverse cracking. Second, the term exp(b5RPi) serves to quantify the effect of RAP: a positive b5 (implying an exponential term greater than one) indicates that the RAP helps resist transverse cracking, and a negative b5 (implying an exponential term less than one) indicates that the RAP contributes to more transverse cracking than virgin AC. From field experience, a negative b5 will be expected, since RAP is more brittle than virgin AC and is thus more prone to cracking. Based on statistical inference from the model estimation results, conclusions can be drawn regarding whether and the underlying factors significantly affect overlay performance and how much the effect will be. By integrating the previous terms into the basic model, the final model specification is obtained as follows: ½4

yi;t ¼

b1 þ 3i;t 1 þ expfb2 þ ½b3 þb4 MLi þexpðb5 RPi Þðb6 THBi þb7 THCi ÞTi;t g

where yi,t is the transverse cracking length (ft. or m) in section i (each 500 ft. or 152.4 m in length) at time point t; ML is a dummy variable for construction preparation (ML = 1 for mill, ML = 0 for no mill); RP is a dummy variable for overlay material (RP = 1 for RAP, RP = 0 for virgin); THB is a variable for type B overlay thickness (in. or mm); THC is a variable for type C overlay thickness (in. or mm); T is the time after overlay construction (years); b1–b7 are parameters to be estimated; and 3 is an error term, customarily assumed to follow a normal distribution.

Model estimation and results The parameter means are estimated using the least-squares technique in nonlinear model estimation. An iterative process is used to obtain the parameter estimates by minimizing the sum of squared errors: ½5

X i;t

3i;t

2

 ¼ yi;t 

b1 1 þ expfb2 þ ½b3 þb4 MLi þexpðb5 RPi Þðb6 THBi þb7 THCi ÞTi;t g

2

b 1b b 7 and error b3 i;t , the inferences, particularly the t-statistics for all parameters, Based on the estimated parameter means b are obtained (Greene 2007). These results are presented in Table 2. The goodness-of-fit (R2) statistic is also given in Table 2. The R2 value of 0.92 suggests that the model provides a good fit to the observations. Fig. 5 illustrates the relationship between observed and predicted transverse cracking length for all test sections and provides an example from test section A502. Figure 5 shows that (i) the data points for the observation–prediction pairs are closely scattered around the line of equality, and (ii) the proposed S-curve captures the transverse crack development very well. Published by NRC Research Press

Hong and Chen

1417

Fig. 7. Calculating pavement performance and life from established deterioration model (an example on no-mill 2 in. TYC AC RAP).

Fig. 8. Predicted pavement performance comparison under different conditions.

As emphasized in most statistical deterioration modeling, the statistics in the parameter estimates deliver valuable information and thus deserve a special discussion. As is customary, a confidence level of 95% or a significance level of 5% is adopted in this study. Table 2 shows that all parameter estimates are statistically significant at a 95% confidence level (based on the criterion of an absolute t-statistic larger than 1.96). In particular, this means that all critical factors in the underlying experiment, namely the surface preparation, overlay thickness, and nature of the overlay material, have a significant impact on transverse crack propagation. The details of the findings are discussed as follows. For the effect of surface preparation on transverse crack propagation, the negative sign and statistical significance of b 4 imply that (with everything else the same) the 2 in. b (50.8 mm) milling results in greater overall length of transverse cracking. It can be shown in eq. [4] that, when compared with the no-mill condition (ML = 0), the milled b 4 leads to a smaller condition (ML = 1) with the negative b denominator in the performance model, and thus a larger value of yi,t. This finding suggests that the 2 in. (50.8 mm) surface mill (which did not completely remove existing cracking; to be discussed subsequently) worsens the anticracking capability of an overlay. This can be explained from an overlay structural design perspective. The milled 2 in. (50.8 mm) surface layer reduces the effective structural

number (as a linear function of the thickness of the pavement layers) of the existing pavement and thus leads to a lower load bearing capacity of the pavement (AASHTO 1993). There have been some discussions regarding the effectiveness of milling on AC overlay performance in terms of cracking. For example, a previous study by Wen et al. (2006) on longitudinal crack propagation showed that a higher ratio of overlay thickness to milling depth could result in a lower rate of crack development. This implies that deeper milling could not contribute to mitigating overlay cracking in some cases, which is consistent with the finding in this study. Loria et al. (2008) reported that milling the overlay is an effective treatment for reflective cracking. However, neither study clearly indicated if the existing cracking was completely removed during milling. Existing cracking can occur only in the asphalt layer or also in base or subbase layers and subgrade. It should be noted that if the existing cracking is completely milled, it may be helpful in reducing cracks in the overlay. In the underlying pavement in this study, the cracking was not completely removed, as is shown in Fig. 6 from the field coring survey. The two half-pieces of existing AC indicate that transverse cracking still existed in the underlying pavement after milling and then propagated upward into the AC overlay. With respect to the effect of overlay thickness on the transverse cracking propagation, the positive sign and statisPublished by NRC Research Press

1418

Can. J. Civ. Eng. Vol. 36, 2009

Fig. 9. Comparison of (a) predicted transverse cracking length and (b) service life between mill and non-mill and RAP and virgin material conditions.

overlay containing RAP material exhibits more overall transverse cracking than an overlay with only virgin asphalt. This can be explained from the viewpoint of an asphalt material property. It is known that asphalt binder in RAP usually demonstrates higher viscosity due to the aging effect. Thus, the overlay with RAP is less resistant to cracking than that with virgin asphalt. Furthermore, the relative effectiveness of AC with 35% RAP over virgin AC in terms of the capability to resist transverse cracking can be obtained b 5 RPÞ = 0.64 when RP = 1 quantitatively. The value of expðb implies that, with everything else being the same, an overlay with 35% RAP material is 0.64 times as effective as that with virgin AC in resisting transverse cracking. The implication of this finding in pavement overlay structural design is appealing because it can serve as a shift factor when RAP is adopted to replace part of the virgin AC. It is noted that an overlay with 35% RAP material represents the highest percentage of reclaimed AC used in current practice. With a lower percentage of RAP employed, the shift factor will be closer to 1. In this sense, a balance can be sought to maximize the benefit to cost ratio (b/c) through jointly considering RAP percentage and cracking performance. This deserves further study as more data concerning pavement performance and construction costs at different percentages of RAP adopted are available.

Performance–life comparison based on the developed model

b 6 and b b 7 indicate that a thicker overlay tical significance of b is less prone to transverse cracking than a thin overlay. This finding supports the fact that thicker pavement reduces both bending and vertical shear, and thus is more resistant to b 6 = 1.6 suggests cracking (AASHTO 1993). The ratio of b b 7 /b that (with everything else the same) a unit thickness of type C AC has a greater capacity (by a factor of 1.6) than that of type B AC to resist transverse cracking. This is consistent with the idea that AC with finer aggregate and a higher asphalt content demonstrates lower stiffness and is more crack resistant. Note that type C AC is composed of finer aggregate than type B AC (the maximum sieve sizes for types B and C AC are 1 in. and 0.75 in., respectively); type C contains a higher asphalt binder content than type B AC (the binder content in type C AC is 0.4%–0.7% higher than that in type B AC in this study). Lastly, for the effect of different materials on the transverse crack propagation, the negative sign and statistical sigb 5 suggest that, with all else being equal, an nificance of b

½6

Eðyt¼T jXÞ ¼

With the deterioration model established, it is possible to compare pavement performance (or life in terms of transverse cracking) under different conditions. Two criteria must be established to carry out the comparison, namely minimum service life for the overlay and maximum allowable transverse crack length. Eight years was adopted as the minimum service life, since that is usually the minimum service time required by state highway agencies for pavement rehabilitation work (Liu and Scullion 2001; Zaghloul et al. 2008). There is very little information available to establish pavement life in terms of total transverse cracking length. Fortunately, there is a length-based cracking criterion for pavement failure in terms of surface-down fatigue cracking and thermal cracking. According to the MEPDG, the maximum allowable length for both cracking types is 1000 ft. (333.3 m) per mile (NCHRP 2004). By the same ratio, for a standard LTPP test section of 500 ft. (152.4 m), the failure criterion is 100 ft. (33.3 m) per section. This criterion is adopted in this study for determining pavement failure in terms of total transverse crack length. However, if another limit is desired, the same method applies in predicting transverse cracking related pavement life. The expected total length of transverse cracking at time T (in years) can be obtained as follows:

b b1 b 2 þ ½b b3 þ b b 4 ML þ expðb b 5 RPÞðb b 6 THB þ b b 7 THCÞTg 1 þ expfb

Published by NRC Research Press

Hong and Chen

where X = [ML, RP, THB, THC, T]’, including a given pavement condition; and all the parameters are the estimated means from model estimation results (see Table 2). Comparison of pavement performance Transverse crack development under different conditions can be calculated and compared using eq. [6]. As an example, Fig. 7 illustrates the performance curve for the no-mill condition, with 2 in. (50.8 mm) type C AC and a RAP overlay. Following the broken lines and arrows, transverse crack length at time point T = 8 years is 195 ft. (65 m) in this example. By the same method, the total transverse crack length in no-mill pavement with 2 in. type C AC and only virgin asphalt can be calculated as 96 ft. (32 m). This indicates that for a thin overlay of 2 in., the expected overall transverse crack length in the pavement using 35% RAP is about two times that in virgin AC. For an overall comparison, transverse crack length development under different combinations of mill–non-mill and RAP–virgin (with an overlay structure of 2 in. or 50.8 mm type B + 2.6 in. or 66 mm type C) is obtained, as shown in Fig. 8. Pavement performance in terms of transverse crack length with virgin AC and non-mill is better than that with 35% RAP and 2 in. or 50.4 mm mill. The difference in the material performance (RAP versus virgin AC) is more pronounced than the difference in surface preparation (2 in. or 50.8 mm mill versus non-mill). At the end of 8 years (required minimum service life), it was found that the average crack length ratios were 2.4 and 8.3 between milled and non–milled and 35% RAP and virgin AC conditions, respectively, as is shown in Fig. 9a. Comparison of pavement life A back-calculation process is required to determine pavement life for a given crack length limit (i.e., 100 ft. or 33.3 m per 500 ft. or 152.4 m section). Pavement life is obtained by solving eq. [6] when Eðyt¼T jXÞ = 100 ft. (33.3 m). For example, following the solid lines in Fig. 7, pavement life is 6.3 years for pavement with no milling, 2 in. (50.8 mm) type C AC, and 35% RAP in the overlay. Similarly, the expected life in terms of total transverse crack length in a pavement section with no milling, 2 in. type C virgin AC overlay can be calculated as 8.1 years, which is about 35% longer than that with RAP. The average life ratios are 0.6 and 0.4 between milled and non-milled and 35% RAP and virgin AC conditions, respectively, as is shown in Fig. 9b. In summary, the performance or life comparison capacity of these models serves to better evaluate pavement performance in terms of transverse cracking under different design factors and can be employed to facilitate rehabilitation schemes in a life-cycle cost context.

Conclusions This study investigates the effect of three frequently encountered factors in flexible pavement rehabilitation projects on the transverse crack length propagation in an overlay structure, namely surface preparation, thickness, and materials. Based on the data collected from the in-service pavement sections of long-term pavement performance (LTPP) specific pavement study 5 (SPS-5) located in Texas, a dete-

1419

rioration model is established to characterize transverse crack length development. The model is capable of accurately capturing the deterioration process. By integrating statistical analysis results and engineering knowledge, the key findings are as follows: (1) Surface preparation with 2 in. (50.8 mm) milling on existing pavement leads to longer overall transverse crack length if the milling does not completely remove existing cracks. On average, after 8 years of service, the transverse crack length in milled sections is 2.4 times that in sections without milling, or pavement life in milled sections is 0.6 times that without milling. (2) In terms of the capability to resist transverse cracking, the average effectiveness of the asphalt concrete (AC) with 35% reclaimed asphalt pavement (RAP) is about 64% that of virgin AC. The transverse crack length is 8.3 times as long, on average, after 8 years of service, and pavement life is 0.4 times as long with RAP versus virgin AC. (3) The thickness of the overlay contributes to the resistance to transverse cracking: the thicker the overlay structure, the greater the reduction in transverse cracking in the overlay. (4) The capability to resist transverse cracking of an overlay composed of Texas type C AC is around 1.6 times that of an overlay composed of type B AC. It is noted that the latter contains coarser aggregate and a lower asphalt content, which contribute to higher transverse cracking resistance in the AC overlay. In summary, this study quantifies the effect of three key factors in flexible pavement overlay work. The deterioration model and findings can be used to update pavement rehabilitation design and help facilitate pavement rehabilitation management.

Acknowledgements The authors would like to express their sincere thanks to Abbas Mehdibeigi and Chris Johnson from the Texas Department of Transportation, Jerry Daleiden, Mark Gardner, and Mainey James from Fugro BRE, Inc., and Fujie Zhou from the Texas Transportation Institute for their kind support in the field survey and instrumental discussion.

References AASHTO. 1993. AASHTO guide for design of pavement structures. American Association of State Highway and Transportation Officials (AASHTO), Washington., D.C. Arudi, R.S., Minkarah, I., Kandula, K., and Gosain, A. 1996. Performance evaluation of asphalt overlays on broken and seated concrete pavements. In Design and construction of asphalt overlays and hot mix asphalt construction practices. Transportation Research Record 1543, Transportation Research Board, Washington, D.C. pp. 55–62. Chen, D.-H., and Daleiden, J. 2005. Lessons learned from longterm pavement performance program and several recycled sections in Texas. In Proceedings of the 1st National Conference on Pavement Preservation, Kansas City, Mo., 31 October – 1 November 2005. Circular E-C078, Transportation Research Board, Washington, D.C. pp. 70–84. FHWA. 2003. Long-term pavement performance information manPublished by NRC Research Press

1420 agement system pavement performance database user guide. Federal Highway Administration Report FHWA-RD-03–088, Federal Highway Administration (FHWA), Department of Transportation, Washington, D.C. FHWA. 2008. Long term pavement performance program. Federal Highway Administration (FHWA), Department of Transportation, Washington, D.C. Available from www.datapave.com [cited October 2008]. Garcia-Diaz, A., and Riggins, M. 1984. Serviceability and distress methodology for predicting pavement performance. In Pavement management activities. Transportation Research Record 997, Transportation Research Board, Washington, D.C. pp. 56–61. Greene, W.H. 2007. Limdep, version 9.0. Econometric Software, Inc., Plainview, N.Y. Haas, R., Hudson, W.R., and Zaniewski, J. 1994. Modern pavement management. Krieger Publishing Company, Malabar, Fla. Hong, F., Rosales-Herrera, V.I., and Prozzi, J.A. 2008. Development of asphalt pavement transverse crack initiation models based on LTPP data. In Proceedings of the 87th Annual Meeting of the Transportation Research Board, Washington, D.C., 13–17 January 2008. CD-ROM. National Research Council, Washington, D.C. Paper 08–2543. Huang, Y.H. 2003. Pavement analysis and design. Prentice-Hall, Inc., Upper Saddle River, N.J. Liu, W., and Scullion, T. 2001. Flexible pavement design system FPS19W: user’s manual. Texas Transportation Institute, College Station, Tex. Loria, L., Sebaaly, P.E., and Hajj, E.Y. 2008. Long-term performance of reflective cracking mitigation techniques in Nevada. In Infrastructure maintenance. Transportation Research Record 2044, Transportation Research Board, Washington, D.C. pp. 86–95.

Can. J. Civ. Eng. Vol. 36, 2009 NCHRP. 2004. Mechanistic–empirical design of new and rehabilitated pavement structures. National Cooperative Highway Research Program Design Guide 1–37A, National Cooperative Highway Research Program (NCHRP), Transportation Research Board, Washington, D.C. Prozzi, J.A., and Madanat, S.M. 2003. Incremental nonlinear model for predicting pavement serviceability. Journal of Transportation Engineering, ASCE, 129(6): 635–641. doi:10.1061/(ASCE)0733947X(2003)129:6(635). SHRP. 1989. Specific pavement studies, experimental design and research plan for experiment SPS-5, rehabilitation of asphalt concrete pavements. Strategic Highway Research Program (SHRP), National Research Council, Washington, D.C. Tayabji, S.D., Brown, J., Mack, J.W., Hearne, T.M., Jr., Anderson, J., Murrell, S., and Noureldin, A.S. 2000. Pavement rehabilitation. Transportation Research Board, Washington, D.C. Available from http://onlinepubs.trb.org/Onlinepubs/millennium/ 00086.pdf [cited October 2008]. Texas Department of Transportation. 2004. Standard specifications for construction and maintenance of highways, streets, and bridges. Texas Department of Transportation, Austin, Tex. Wen, H., Titi, H., and Berry, D. 2006. Study of best practices for pre-overlay repair and asphalt overlay. In Airfield and highway pavement: meeting today’s challenges with emerging technologies. Edited by I.L. Al-Qadi. ASCE, Reston, Va. pp. 815–823. Zaghloul, S., Khaled, H., and Bekheet, W. 2006. Development and implementation of Arizona Department of Transportation (ADOT) pavement management system (PMS). Arizona Department of Transportation, Phoenix, Ariz. Available from http:// www.azdot.gov/TPD/ATRC/publications/project_reports/PDF/ AZ494/Text.pdf [cited October 2008].

Published by NRC Research Press