Guidelines for Dowel Alignment in Concrete Pavements (2009)

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

22.1 Introduction This section describes the purpose of transverse joints and dowels in concrete pavements, introduces dowel misalignment terminology used in this study, and gives an overview of the dowel misalignment specifications used by various transporta- tion agencies. 2.1.1 Joints and Dowels in Concrete Pavements Joints are introduced in PCC pavements to allow for thermal expansion and contraction, as well as shrinkage after construction. To improve load transfer across the transverse joints (thus minimizing faulting and corner breaks), many transportation agencies place dowels at mid-depth between the pavement slabs using either basket assemblies or an automated dowel bar inserter (DBI) (McGhee, 1995). The load transfer concept is illustrated in Figure 2.1 (PCA, 1991). When a wheel load is applied to an undowelled joint, greater edge and corner deflections (and corresponding stresses) are experienced. Dowel bars reduce the critical deflections and stresses by transferring the load between the slabs. Several studies, including Yu et al. (1998), Khazanovich et al. (1998), and Hoerner et al. (2000), concluded that properly designed and installed dowels greatly reduce transverse joint faulting and corner cracking. Dowels should maximize vertical load transfer, minimize longitudinal restraint, and be durable (Lechner, 2005). 2.1.2 Terminology Ideally, dowel bars should be placed such that the longitu- dinal axis is parallel to both the surface and centerline of the hardened PCC slab, and the geometric center of the dowel bar is directly below the joint. If dowel position in the hardened concrete deviates from this ideal position, it is said to be misaligned. Misalignment may result from misplacement (initially placing the dowels in an incorrect position or saw cutting at the incorrect location), displacement (movement during or following the paving operation), or both. The following major categories of dowel misalignments were identified by Tayabji (1986) (see Figure 2.2): • Longitudinal translation; • Vertical translation; • Horizontal skew; and • Vertical tilt. Vertical translation refers to the deviation of the position of the dowel relative to the reference mid-depth position. How- ever, because concrete cover describes the distance between the dowel and slab surface, it also reflects the vertical positions of the dowel and its vertical translations. 2.1.3 Current Specifications Different states have adopted different requirements for dowel bar tolerances with respect to longitudinal and vertical translation, horizontal skew, and vertical tilt (see Table 2.1). These tolerances can be expressed as absolute maximum measures or as percentages of the length of the dowel or thick- ness of the concrete. Many states have adopted the FHWA- recommended limits for horizontal skew and vertical tilt of 1⁄4 in. over 12 in. (6.3 mm over 305 mm) or 2% (FHWA, 1990). FHWA recommended further studies to determine the validity of this 2% tolerance (FHWA, 1990). The American Concrete Pavement Association (ACPA) recommends limits of 3⁄8 in. over 12 in. (9.5 mm over 305 mm) or 3% based on NCHRP Synthesis 56 (ACPA, 2004; NCHRP, 1979). The data provided in Table 2.1 were obtained from literature review (MCC, 2004; Lechner, 2005) and communications with state department of transportation (DOT) representatives. Table 2.2 gives the dowel bar alignment tolerances permitted in the construction specifications of the Ministry of Trans- C H A P T E R 2 Research Methodology

3Figure 2.1. Effectiveness of load transfer (PCA, 1991). Figure 2.2. Types of dowel misalignment (adopted from Tayabji, 1986). Vertical Tilt Horizontal Skew LongitudinalTranslation Vertical Translation Agency in. per 18 in. in. per 18 in. in. per 18 in. in. Arkansas Connecticut Federal Aviation Administration Hawaii Idaho Kentucky Minnesota Texas Utah Wisconsin Nebraska Iowa Michigan N/A N/A Montana 0.50 0.50 North Dakota Tennessee 0.25 0.25 0.25 0.25 Ontario 0.24 0.24 0.59 0.59 Nevada 0.50 0.50 N/A N/A Missouri 0.50 0.50 0.50 1.00 Kansas 1/10 Pavement Depth Indiana North Carolina 0.38 0.38 N/A N/A Illinois Delaware 0.19 0.19 N/A N/A South Carolina 3.00 0.75 Georgia 0.56 0.56 N/A N/A Germany 0.75 0.75 2.00 N/A Alabama 0.25 0.69 N/A N/A Great Britain 0.39 0.39 N/A N/A New York N/A 0.16 0.25 0.26 Ohio N/A N/A 0.50 0.50 Pennsylvania 0.23 0.23 1.00 1.00 0.25 0.25 0.25 0.25 0.38 0.38 N/A 0.562 0.562 Table 2.1. Specified dowel misalignments limits.

4portation of Ontario (MTO, 2007). The MTO tolerances are based on research performed in Ontario to determine the extent and effect of dowel misalignment in pavement construction. Recent guidelines developed by the FHWA also are based on the alignment and performance data (FHWA 2007). 2.2 Dowel Misalignment Assessment This section summarizes available information on the state- of-the-art in field and laboratory testing, as well as analytical modeling, for dowel misalignment. 2.2.1 Field Testing There have been a limited number of field studies of dowel misalignment (Tayabji and Okamoto, 1987; Yu et al., 1998). Devices used for identifying dowel misalignment include MIT Scan-2, the Profometer, and ground penetrating radar (GPR) (Khazanovich et al., 2003). In 2005, FHWA identified the MIT Scan-2 as a tool that could potentially improve the assessment of concrete pavements (FHWA, 2005). An assessment con- ducted by the Virginia DOT also identified MIT Scan-2 as a viable technology for construction quality control (Hossain and Elfino, 2006). Inspection of pavements in several states revealed that the misalignment of dowels generally occurs regardless of the placement method. For example, significant dowel misalign- ment was identified in a pavement section constructed using dowel baskets on Highway 115 in Ontario (Leong, 2006) and in a pavement constructed using a DBI on I-16 in Georgia (Fowler and Gulden, 1983). Field studies also have shown variability in dowel position and alignment from one project to another (Yu, 2005). While a majority of dowel bars meet state specifications for alignment on most projects, there are a number of dowel bars that do not meet specifications. The performance of some of these sections indicates that slab cracking and other forms of distress may not always occur as a result of such misalignment. Field studies have shown that the only type of misalignment that clearly had an effect on pavement performance was longitudinal translation (causing low embedment length). An example of the effect of low embedment length was observed on I-35 near Fergus Falls, Minnesota, where significant early faulting occurred when dowel embedment lengths were less than 2.5 in. [63 mm] (Burnham, 1999). 2.2.2 Laboratory Testing Previous laboratory studies generally have been limited to dowel pullout tests that focused on dowel resistance to joint opening by measuring pullout force and by evaluating concrete distresses. These tests include standard pullout tests and slab pullout tests. In the former test, a dowel is pulled away from an anchored concrete slab. In the latter, a moving or “transient” slab is pulled away from an anchored or “stationary” slab to open a dowelled joint to a specified width. Up to five dowels are tested in the joint. Slab tests have been used by Tayabji (1986), Prabhu et al. (2006), and others to model slab expansion. The standard pullout test data typically are presented as a plot of pullout force versus dowel horizontal displacement. The results of such tests have been used to calibrate a finite element model (Khazanovich et al., 2001). This well-controlled test provides valuable information related to dowel-PCC friction. More information on this test and modifications made to the test to better characterize the interaction between a misaligned dowel and the surrounding concrete are presented in Section 2.3.3 and Appendix C. The slab pullout test data can be used to model the effects of several misaligned dowels on joint behavior during joint movement. The following trends have been observed (Prabhu et al., 2006; Tayabji, 1986): • The force required to displace the dowel increased with the increased misalignment. • Nonuniform misalignment had a greater effect on pullout force and distresses than uniform misalignment (non- uniform misalignment refers to dowels oppositely misaligned Misalignment Lower Limit (mm) Upper Limit (mm) Horizontal skew (mm per 450 mm dowel) -15 15 Vertical tilt (mm per 450 mm dowel) -15 15 Longitudinal translation (mm) -50 50 Depth tolerance (for specified slab thickness): 200 mm (mid depth - 6 mm/+6 mm) 94 106 225 mm (mid depth -12 mm/+15 mm) 100 127 250 mm (mid depth -15 mm/+25 mm) 110 150 260 mm (mid depth -15mm/+25 mm) 115 155 1 in. = 2.54 mm Table 2.2. Specification limits for position and alignment of dowel bars (MTO, 2007).

and uniform misalignment refers to dowels all misaligned in the same direction). • Slabs develop cracking only at significant misalignment levels (over 3⁄4 in. [19 mm]) when the alignment of dowels along the joint is nonuniform and when excessive levels of joint opening (over 0.5 in. [13 mm]) are present. • Minor spalling around dowels was found in slabs with uniform and nonuniform significant misalignment (Prabhu et al., 2006). Because rotation of the beam in the direction of the mis- aligned dowel may occur during testing and affect test results, provisions must be made to ensure proper anchoring of the concrete (Tayabji, 1986). Nevertheless, slab testing is not expected to provide detailed information about the inter- action between the dowel bar and the surrounding concrete (Prabhu et al., 2006). 2.2.3 Analytical Models The two main categories of analytical models that can be used for assessing the effects of dowel misalignment are structural response models and performance prediction models. 2.2.3.1 Structural Response Models Several finite element and finite difference models have been used to analyze the effects of dowel misalignment (Khazanovich et al., 2001; Davids, 2003; Leong, 2006; Prabhu et al., 2006). Some of the major findings include: • Dowel misalignment increases PCC-dowel contact stresses. • When embedment length falls below some critical level, bearing stresses increase. • If several consecutive transverse joints are subject to lockup, stresses increase away from the joint, with high stresses developing at the mid-slab location. The analytical models for predicting the effects of dowel misalignment on concrete pavement behavior can be classified according to the degree of detail used for modeling the dowels and their interaction with concrete as Types I and II. Type I models provide detailed modeling of dowels and dowel-PCC interaction. These models include: • ABAQUS 3-D model for a single dowel; • ABAQUS 3-D model for several dowels; and • FLAC-3-D model. Type II models provide simplified modeling of dowel-PCC interaction. These models include: • ISLAB2000; • EVERFE; and • ABAQUS-2D multiple slab model. Type I models are suited for analyzing the effects of dowel misalignment on bearing stresses and joint stiffness, whereas Type II models are suited for multi-slab analysis to predict mid-slab stresses. These models were evaluated based on the following criteria: • Ability to model dowel-PCC slip. • Ability to model stress distribution around misaligned dowel. • Ability to model subgrade and base support. • Ability to model nonuniform misalignment. • Ability to model multiple joints. • Model flexibility. • Input requirements. Based on these criteria and experience, the ABAQUS 3-D models were selected for use and modification in this study. 2.2.3.2 Performance Prediction Models Models for predicting jointed plain concrete pavement (JPCP) cracking, joint faulting, spalling, and roughness were identified and evaluated. The evaluation revealed that none of the performance models consider dowel alignment as an input parameter. However, mechanistic-empirical (ME) pavement performance models can be adapted to account for the effects of dowel misalignment. Several faulting models relate concrete bearing stresses under the critical dowel with the rate of load transfer efficiency (LTE) deterioration and faulting development (Owusu-Antwi et al., 1997; Hoerner et al., 2000; Khazanovich et al., 2004). Higher bearing stress accelerates joint LTE deterioration and causes early faulting. Thus, if higher bearing stresses were observed for reduced dowel diameters and also observed for joints with misaligned dowels, levels of dowel misalignment could be equated to reduced dowel diameters when bearing stress is considered. Cracking models relate PCC pavement longitudinal bending stresses developed at mid-slab with the percentage of cracked slabs. If dowel misalignment causes joint lockup, it may cause additional tensile stresses that should be accounted for in the cracking model. Available performance prediction models that can be used for development of guidelines for dowel alignment were eval- uated based on the accuracy of predictions, simplicity of use, and simplicity of integration with the dowel misalignment analysis. The evaluations indicated the following: • The Mechanistic-Empirical Pavement Design Guide (MEPDG) (AASHTO, 2008) faulting model was appropriate for prediction of the long-term effects of dowel misalign- ment on joint faulting. • The MEPDG cracking model was the most comprehensive model available for cracking prediction. 5

6• None of the identified spalling models could be used for analyzing the effects of dowel misalignment on joint spalling. Based on these evaluations, the MEPDG faulting and crack- ing models were selected for modification in this study and the MEPDG International Roughness Index (IRI) model was adapted for roughness prediction. 2.3 Research Approach The following section describes the approach used in con- ducting field and laboratory tests, analytical modeling, and developing performance predication models for JPCP with misaligned dowels. 2.3.1 Field Testing The field testing included (a) evaluation of typical dowel alignments observed across the United States for a variety of construction projects and (b) identification of short-term and long-term effects of dowel misalignment on pavement performance. 2.3.1.1 Alignment and Performance Database A database of the dowel alignment and pavement perform- ance was assembled from the evaluation of 37 pavement sec- tions and information on 23 additional pavement sections reported in other studies (Yu, 2005). These 60 pavement sections are located in Arizona, California, Colorado, Georgia, Indiana, Illinois, Kansas, Michigan, Minnesota, Missouri, Nevada, North Carolina, Ohio, South Dakota, Virginia, Washington, and Wisconsin. The candidate sections for field data collection were iden- tified with assistance from state DOTs. Another source of projects for field evaluation was the Long Term Pavement Performance (LTPP) database, particularly the Seasonal Mon- itoring Program (SMP) sections because they included infor- mation on joint opening and historical time series on distress, faulting, LTE, etc. Seventeen of the 37 pavement sections surveyed in this study are LTPP test sections. Appendix B lists all pavement sections, summarizes their design features, and describes the testing operations performed on each. These sections represent broad ranges of design, construction, climate and traffic variables: • Climatic region: 8 sections in dry-freeze, 24 sections in dry-nonfreeze, 22 sections in wet-freeze, 6 sections in wet-nonfreeze. • Pavement thickness: 5 sections with thickness ≤ 8 in. [203 mm], 5 sections with thickness between 8 and 9 in. [203 and 229 mm], 10 sections with thickness between 9 and 10 in. [229 and 254 mm], 20 sections with thickness between 10 and 11 in. [254 and 279 mm], and 20 sections with thickness ≥ 11 in. [279 mm]. • Dowel size: 16 sections with 1.25-in. [32 mm] diameter dowels, 42 sections with 1.5-in. [38 mm] diameter dowels, and 2 sections with dowels of other diameters (1 and 1.125 in. [25 and 29 mm]). • Dowel installation procedure: 35 sections were constructed using basket assemblies, 23 sections using DBI, and 2 sections were retrofitted. • Construction year: 4 sections were constructed before 1991, 22 sections between 1991 and 1995, 10 sections between 1995 and 2000, 20 sections between 2000 and 2006, and 4 sections in 2007. • Average daily traffic (ADT): 16 sections had ADT ≤ 15000, 12 sections had ADT between 15000 and 30000, 19 sections had ADT between 30000 and 60000, 12 sections had ADT ≥ 60000, and 1 section (on MnROAD) had 100,000 passes of 80 kip (35.6 kN) truck. 2.3.1.2 Data Collection Every joint in each pavement section was tested using MIT Scan-2, and MagnoProof (MIT Scan-2 PC software) was used to quantify dowel alignment and position in the pave- ment section. In addition to the dowel alignment and position data, fault- ing of each transverse joint was measured using a faultmeter reading to the nearest 0.01 in. (0.25 mm). Readings were taken in the outer wheel path (approximately 18 in. [450 mm] from the edge of the lane) and at the slab corner. Thus, for a 500-ft. (150-m) section with 15-ft. (4.6-m) joint spacing, over 30 faulting measurements were made. A complete distress survey of each pavement section also was conducted in accordance with the LTPP Distress Identification Manual (Miller and Bellinger, 2003); the extent and severity of cracking, spalling, corner breaks, and so on were noted. At each transverse joint, the overall extent of joint deterioration was noted and the severity was rated as None, Low, Medium, or High. The condition of the joint seal (if present) also was noted. Digital photographs were taken to document the overall condition of each test section and typical distresses (if any), as well as the site conditions. On some sections, joint LTE was measured using the Falling Weight Deflectometer (FWD). 2.3.2 Laboratory Testing Laboratory testing was conducted on pavement slabs in a controlled environment to determine the effects of dowel misalignment. Performance parameters such as maximum required pullout force, dowel shear stiffness, and ultimate

dowel shear capacity were measured. The standard pullout test was modified to eliminate the influence of beam rotation. Also, shear pull tests were conducted to address the effect of misalignment on shear performance. 2.3.2.1 Laboratory Setup Each specimen consisted of a 4-ft. wide [1.2 m], 8-in. thick [203 mm], and 18-in. [457 mm] tall concrete beam containing four 1.25- or 1.5-in. [32 or 38 mm] diameter dowels placed 12 in. [305 mm] apart, with the end dowels 6 in. [152 mm] away from the edge (Figure 2.3). Test specimen dimensions were selected with consideration to the capabilities of the available test apparatus and the planned finite element modeling. Beam thickness was based on a general design of a thin (8-in. [200-mm] thick), doweled PCC pavement. The width was chosen as the most efficient width for casting and testing the dowels using the available testing apparatus and the height was selected to ensure that the test beam adequately represents a “long” PCC slab (i.e., the specimen has sufficient length in the direction of dowel embedment such that boundary con- ditions do not significantly influence test results). The finite element simulation of the modified pullout test indicated that increasing the beam height beyond 18 in. [457 mm] would not provide any advantage but would increase specimen weights and make it difficult to handle. The 1.25- and 1.5-in. [33- and 38-mm] dowel diameters were chosen because they are commonly used in the United States. The distance between the dowels was selected to ensure that (1) the specimen clamps could be placed on the beam at sufficient distances from each dowel being tested to avoid influencing the test results and (2) the damage of the beam after a pullout test on one dowel would not affect the adjacent dowels. To ensure precision in installing the dowels with the intended misalignment, a dowel jig was fabricated and a procedure was used for setting dowels with precisely the desired type and amount of misalignment. Each jig featured two holes that were offset to provide the desired misalignment (see Figure 2.4). The top end of each dowel was tapped to allow a dowel exten- sion to be screwed into place. The extended dowel was inserted through the jig holes and set at the proper embedment length and angle. After the concrete had been placed and cured suf- ficiently, the dowel extension and jig were removed. The mold was stripped from the specimen after the con- crete was cured sufficiently (a minimum of 24 hours) to avoid damage. Each beam was then cured under water for 6 days before testing. One ungreased 6-in. [153-mm] dowel was included and tested in each beam to provide a reference between beams. A compressive strength test was conducted 7 days after beam casting. The MinneALF structure was modified to accommodate the modified pullout and shear pull tests (Khazanovich et al., 2005). These modifications included adjustment of the actuator positions and installation of the beam clamping mechanism. 2.3.2.2 Test Procedure The dowel pullout testing was conducted after the test beams had been water-cured for 7 days. Because concrete pavement can experience contraction and shrinkage within several hours after concrete setting, the 7-day curing time was selected to ensure uniformity of the test beams. Dowels with various levels of misalignments were tested as follows: 1. Each dowel was tested individually by pulling it vertically with respect to the concrete beam, along the ideal axial direction of a properly aligned dowel in a displacement- controlled mode at a rate of 0.003 in./sec (0.076 mm/sec) until the dowel had translated (pulled out) 0.25 in. (6 mm) relative to the concrete. Pullout force and displacement were recorded continuously. 2. A post-test examination was conducted to evaluate the con- crete surrounding the dowels (visible damage was recorded).Figure 2.3. Test specimen. Dowel extension Dowel inserter Tapped end of the dowel Figure 2.4. Dowel extension and alignment jig. 7

After the pullout test was completed, the beam was rotated 90 degrees so that it was lying on its side (Figure 2.5). The beam then was clamped to the test stand and a shear load was applied to selected dowels in a direction perpendicular to the plane of the slab surface. The dowel was pulled in a displacement-controlled mode until failure. Dowels with various levels of misalignments were tested as follows: 1. Each dowel was tested individually by pulling it vertically with respect to the concrete beam, in a displacement- controlled mode until the concrete surrounding the dowel failed. During the testing, the shear pull force and displacements of the dowel and concrete were recorded continuously. 2. After testing each beam, an examination was conducted. The concrete surrounding the dowels was evaluated and the failure mode was recorded. The effect of misalignment also was evaluated in repeated shear load tests with the following parameters: • Magnitude of loading: 3 kips [13.3 kN]. • Load frequency: 2 Hz. • Rest period: 0.5 seconds. To reduce the residual effect (“bouncing”), a static “seating” load of 500 lb [2.2 kN] was present between loading cycles. • Measurement frequency: every 0.1 seconds. • Number of load cycles: at least 10,000. The laboratory process and setup for the repeated load test was the same as that used for the static test, except that a repeated 3-kip [13.3 kN] load was applied for at least 10,000 cycles. In this manner, the effect of a one-time load to failure on strength and stiffness could be related to the effect of a repeated load fatigue test. 2.3.2.3 Testing Factorial The levels of misalignment used in the tests were selected based on a review of previous misalignment investigations, which showed that dowel rotational misalignments of up to 1 in. [25 mm] cause only small changes in distress and resistance to joint opening (Tayabji, 1986; Prabhu et al., 2006). Tests of dowels with up to 4 inches [102 mm] in vertical tilt showed no significant differences in behavior for the various levels of misalignment (Tayabji, 1986). Other tests found no distresses for levels of skew and tilt of up to 1 in. [25 mm] for vertical and horizontal misalignment and up to 3⁄4 in. [19 mm] for combined tilt and skew (Prabhu et al., 2006). In this study, tests were conducted on dowels with rotation ranging from 0 in. (i.e., properly aligned) to 4 in. per 18 in. [102 mm per 457 mm], embedment length ranging from 2 to 9 in. [51 to 229 mm], and concrete cover ranging from 1.25 to 3.375 in. [32 to 86 mm]. Because a typical dowel is 18 in. [229 mm] long, all dowel rotations are expressed as the ver- tical or horizontal displacement of one end of the dowel per 18 inches [229 mm] in length. Pullout tests were conducted on specimens, each containing four dowels with various types and levels of misalignment. After the test, the two dowels on the outside of the beam were tested in shear or repeated shear (Appendix C presents the dowel alignments for each dowel tested in this study). Shear testing of the first two beams revealed difficulties in testing four dowels within a single test beam due to the occur- rence of horizontal cracks. It was concluded that only the outside dowels could be tested to measure the ultimate shear load capacity of a given dowel because testing of the interior dowels would result in horizontal cracks that would influence the shear pull test results of adjacent dowels. Therefore, all of the dowels were tested in pullout but only the outside dowels were tested in shear, as noted in Figure 2.6. 8 Figure 2.5. Setup for vertical shear test (also for repeated shear testing).

2.3.2.4 Results Interpretation Modified Pullout Tests. During pullout testing, the pullout force and displacements were recorded continuously. Figure 2.7 shows an example of the two types of curves recorded. Throughout the pullout testing, a majority of the dowels showed a monotonically increasing force-displacement curve (i.e., the pullout force increased with dowel displacement) and others showed discontinuous force-displacement curves, simi- lar to those shown in Figure 2.7. The behavior illustrated by the properly aligned curves is characterized as “static-slip” because it appears that the dowel “slips” slightly each time enough force is generated to exceed the static friction conditions. In these tests, maximum pullout force for a dowel with a given combination of misalignments was increased and then used to evaluate the effects of dowel misalignment on dowel resistance to joint opening and joint lockup. It should be noted that the measured pullout forces in this study were higher than the pullout forces reported from foren- sic studies in which the doweled joints were extracted from pavements and tested in the lab. The difference is probably because the laboratory-prepared specimens were cured for 7 days before testing whereas the in situ dowels were subjected to joint movements at a much earlier age. Shear-Pull Tests. Figure 2.8 shows the locations at which vertical displacement measurements were recorded. The first linear variable differential transformer (LVDT1) measures the displacement of the metal angle above the dowel, giving a measure of absolute dowel displacement. LVDT2 measures the displacement at the edge of the concrete beam closest to the dowel, LVDT3 measures the beam displacement 2 in.[51 mm] from the edge, and LVDT4 measures the beam displacement 4 in. [102 mm] from the edge. All four LVDTs are located in the vertical plane of the dowel (i.e., directly above the dowel). The shear-pull force and displacements of the dowel at the joint face (LVDT1) and at the three locations on the beam (LVDTs 2 through 4) were recorded continuously during testing. Figure 2.9 shows an example of the recorded displace- ments and shear force for a 1.5-in. [38-mm] diameter dowel that was vertically tilted by 2 in. per 18 in. [51 mm per 457 mm] of dowel length. 9 Pull, shear Pull, shearPullPull Concrete beam Figure 2.6. Tests conducted on individual dowels in a beam. 0 1000 2000 3000 4000 5000 6000 7000 8000 0 0.05 0.1 0.15 0.2 0.25 Pu llo ut F or ce , l bs . Dowel Displacement with Respect to the Concrete, in. monotonically increasing static slip Figure 2.7. Examples of the typical pullout test results. Shear Pull Direction LVDT 1 LVDT 2 LVDT 3 LVDT 4 Clamping Angles Figure 2.8. Locations of displacement measurements. 0 2000 4000 6000 8000 0 0.05 0.1 0.15 0.2 Sh ea r Fo rc e, lb s. Relative Dowel Displacement, in. dowel displacement (LVDT1) edge displacement (LVDT2) 2 in. displacement (LVDT3) 4 in. displacement (LVDT4) Figure 2.9. Force versus displacement for a 2 in. vertically tilted dowel.

To conduct analyses of the effects of dowel misalignment on the stiffness of the dowel-concrete interaction, the relative vertical displacement of the dowel end with respect to the surrounding concrete was estimated by subtracting the cal- culated dowel displacement due to the rigid body rotation of the beam from the actual dowel displacement. The calculated dowel displacement was determined from the displacements of LVDT2 and LVDT4 (Figure 2.8) and the specimen geometry. These displacements were used to calculate the slope of the rigid body motion of the beam, Δrb as follows: where l24 = the distance between LVDT2 and LVDT4; ∂2 = the vertical displacement at the edge of the beam (measured by LVDT2); and ∂4 = the vertical displacement 4 inches [102 mm] from the edge (measured by LVDT4). The slope of the rigid body could then be used to calculate the position of the dowel, assuming rigid body motion (i.e., that the beam rotates under load without bending) as follows: where l14 = the distance between LVDT1 and LVDT4; and ∂calc = calculated dowel displacement assuming rigid body motion. To check the accuracy of the rigid body motion calculation, the displacements at LVDT3 were calculated similarly and compared to the actual LVDT3 measurements. Figure 2.10 shows a typical example of the calculated versus measured dis- placements at the 2 in. [51 mm] location for the same dowel. ∂ = ∂ +calc rbl4 14 2Δ ( ) Δrb l = ∂ − ∂2 4 24 1( ) This plot confirms appropriateness of the rigid body assump- tion for the rotation of the concrete beam. Figure 2.11 shows the calculated values for LVDT1 and LVDT3 for the example illustrated in Figure 2.9. The plot shows that, while the beam surface displacements can be described as rigid body motion (i.e., the measured and calcu- lated data points at 2 in. [51 mm] from the joint face are sim- ilar), the dowel exhibits additional displacements with respect to the concrete beam surface. The relative dowel displacement can be computed as: where ∂rel = the dowel displacement due to the compression of the concrete around dowel; ∂calc = the calculated rigid body displacement; and ∂meas = the dowel displacement measured by LVDT1. Figure 2.12 shows a plot of applied shear force versus relative dowel displacement. The relative dowel displacement was used ∂ = ∂ − ∂rel meas calc ( )3 10 0 0.01 0.02 0.03 0.04 0.05 0.06 0 0.01 0.02 0.03 0.04 0.05 0.06 Ca lc ul at ed D isp la ce m en t ( 2 i n. fr om e dg e), in . Measured Displacement (2 in. from edge), in. 0 4000 8000 12000 0 0.01 0.02 0.03 0.04 0.05 0.06 Sh ea r Fo rc e, lb s. Relative Dowel Displacement, in. 9 in. embedment Failure of Concrete Surrounding the Dowel 0 1000 2000 3000 4000 5000 6000 7000 8000 0 0.05 0.1 0.15 0.2 Sh ea r Fo rc e, lb s. Relative Dowel Displacement, in. dowel displacement (LVDT1) dowel calculated displacement edge displacement(LVDT2) 2 in. calculated displacement 2 in. displacement(LVDT3) 4 in. displacement (LVDT4) Figure 2.10. Verification of the rigid body slope. Figure 2.11. Measured and calculated force versus displacement for a 2 in. vertically tilted dowel. Figure 2.12. Example of shear force versus relative displacement.

to estimate the degree of deformation of the concrete during testing. The shear force that caused failure of the concrete surrounding the dowel was considered the ultimate shear strength of the dowel, which indicates the ability to sustain overloading and maintain stiffness under a large number of repeated loads. In addition, the slope of the curve character- izes the stiffness of the tested dowel-concrete interaction. The ultimate shear and stiffness associated with each dowel indicate how the load transfer efficiency might be affected by dowel misalignment. 2.3.3 Analytical Modeling This section provides a brief overview of the finite element modeling conducted in this study (a more detailed description is provided in Appendix D). 2.3.3.1 Finite Element Models To model the effects of dowel misalignment on concrete pavement behavior, the following 3-D ABAQUS models were developed using the approach developed by Khazanovich et al. (2001): • Beam model replicating the laboratory test with individual dowels. • Slab model with four dowels for analysis of the effect of nonuniform dowel rotation on joint load transfer efficiency. The beam model for dowel-concrete interaction was cali- brated using results of the laboratory tests. The calibrated model then was used to investigate misalignment cases and magnitudes that were not tested. To analyze the effect of multiple dowels, a slab model was built using this dowel- concrete interaction model. Based on the results of the finite element modeling, the concept of an equivalent dowel diameter was developed. The effects of dowel misalignment on long-term pavement per- formance were then estimated using this concept and the MEPDG performance prediction models. Beam Model. A single finite element model was used to simulate both modified pullout and shear pull laboratory tests. In the lab, shear testing always was conducted after the pullout testing to model the effect of joint opening prior to wheel loading. In a similar manner, the finite element beam model was set up to apply the pullout test prior to applying the shear pull test, therefore accounting for damage in the concrete beam. Thus, it was necessary to add or remove the clamping mech- anism when changing the simulation from pullout to shear testing. This was accomplished by modeling the horizontal and vertical clamping mechanisms with a temperature-sensitive stiffness. The beam model consists of 8700 elements of type C3D8R (8-node, reduced-integration 3-D linear brick element). A finer mesh was modeled for the dowels and concrete around the dowel to allow for a more detailed analysis of the critical sections surrounding the dowel. Although more computational time is needed for the finer mesh, the higher mesh density allows for more accurate analysis of the strains, stresses, and deflections at the most relevant points. A coarser mesh was assigned to the concrete not surrounding the dowel because less precision was needed in this area and to reduce the computational time without significantly decreasing the accuracy of analysis. Two separate material models were employed to model the concrete. Following Prabhu et al. (2006), the concrete surrounding the dowel was modeled using the “concrete damaged plasticity” option available in ABAQUS. The inelastic behavior of concrete was modeled using the concept of isotropic damaged elasticity in combination with isotropic tensile and compressive plasticity (ABAQUS, 2007). This model accounts for the loss of elastic stiffness due to plastic straining of the concrete in tension and compression. The concrete away from the dowel was modeled as a linear elastic material. The dowel was modeled as an elastic isotropic material, with 20 elements along its length and 20 elements in the cross- section. A finer mesh in the dowel compared to the surround- ing concrete was necessary to improve the stability of the dowel-concrete interaction that was modeled as a surface-to- surface contact defined between two deformable bodies. Initially for the pullout testing, the clamping mechanism was set to be very stiff, and the shear-pull clamp was set at a very low stiffness. A stable friction contact between the dowel and the surrounding concrete was ensured using the procedure developed by Khazanovich et al. (2001). This was followed by the application of the prescribed pullout displacement at the end of the dowel to simulate the displacement-controlled mode of the laboratory testing. After the dowel reached the maximum prescribed displacement, the displacement at the end of the dowel was deactivated to simulate the removal of the test load. After the pullout test simulation was completed, the prop- erties of both clamping mechanisms were changed to simulate stiff shear-pull clamping fixtures and negligible stiffness of the pullout fixtures. This was followed by the application of the prescribed shear-pull displacement at the end of the dowel. Model Validation. To validate the finite element dowel- concrete interaction model, the simulated deflections from the beam model were analyzed in the same manner as the deflections measured in the tests. The relative vertical dis- placements of the dowel end (with respect to the surrounding concrete) were estimated. These laboratory-measured relative displacements and shear force data were used to validate the 11

finite element model. The displacement measurement loca- tions used in the laboratory also were used in the finite element model calculations so that the relative dowel displacement data obtained in the lab could be compared directly to the finite element results. By comparing the results of the shear-pull tests to those of the ABAQUS model, rational parameters were established for the dowel-concrete interaction model. Figure 2.13 shows the shear force versus relative displace- ments for an aligned dowel tested in the laboratory and for a simulated dowel using ABAQUS. The figure shows some agreement between the model and laboratory results. Similar observations were made for the other alignment conditions. The shear capacity was used to compare the performance of the dowels in the ABAQUS simulations to that obtained from laboratory measurements. Figure 2.14 illustrates the shear force versus relative displacement for a reduced embedment length of 6 in. [152 mm] and the most extreme case of reduced embedment length of 2 in. [51 mm] for both laboratory and analytical estimates. The figure indicates similar shear stiffness for the model and laboratory data with respect to the shear capacity. The shear force at failure was used in the labora- tory analysis while the shear force required to cause 0.05 in. [1.27 mm] of relative displacement was used in the finite element analysis. Table 2.3 compares the shear capacities obtained from the lab testing to those estimated from the calibrated analytical model for all of the embedment lengths. There is a consistent agreement between the laboratory testing and analytical model results with the largest difference being less than 1 kip [4.45 kN]. Thus, the analytical model can be applied to those cases for which lab testing was not feasible. Slab Model. Although the beam finite element model is an effective tool for analyzing the effects of longitudinal and vertical translations on the behavior of individual dowels, previous laboratory and analytical studies indicated that the effects of dowel rotational misalignments (in the form of horizontal skew and vertical tilt) are affected by the mis- alignments of other dowels in the joint (Tayabji, 1986; Khazanovich et al., 2001; Prabhu et al., 2006). To investigate this phenomenon, the beam model was expanded to simu- late a slab with multiple dowels in the joint and to consider temperature expansion and contraction, as well as wheel loading at the joint. The slab model consists of two slabs connected by four dowels at the joint and resting on an elastic Winkler foundation. Each slab is 60 in. [1524 mm] wide and 90 in. [2286 mm] long. The symmetrical boundary conditions along one of the longi- tudinal slab edges make the effective slab width 120 in. [3 m]. These boundary conditions also reduce the model run time by more than half (with respect to a full-scale model) without affecting accuracy. To limit the effect of the reduced slab length in the longitudinal direction, the ends of the slab along the outside longitudinal edges were constrained by springs. A comparison of this model with a full-scale, four-slab ISLAB2000 12 0 4,000 8,000 12,000 0 0.01 0.02 0.03 0.04 0.05 0.06 Sh ea r Fo rc e, lb s. Relative Dowel Displacement, in. Analytical Laboratory 0 4,000 8,000 12,000 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 Sh ea r Fo rc e, lb s. Relative Displacement, in. Laboratory (6 in. embedment) Analytical (6 in. embedment) Laboratory (2 in. embedment) Analytical (2 in. embedment) Figure 2.13. Laboratory and analytical shear versus displacement for an aligned dowel. Figure 2.14. Laboratory and analytical sheer versus displacement for misaligned dowels. Shear Capacity (lb) Embedment Length (in.) ABAQUS Calculated Laboratory Measured 2 4870 5360 3 6590 5930 4 7950 7050 5 9070 N/A 6 9890 9770 9 10,370 9570 Table 2.3. ABAQUS-computed versus laboratory-measured shear capacities for various dowel embedment lengths.

model (Khazanovich et al., 2000) found that the difference in LTE at the slab corner was within one percent. As in the beam model, a finer element mesh was used for the dowels and the concrete surrounding the dowels; a total of 30,464 type C3D8R elements (8-node, reduced-integration 3-D linear brick, continuum element) were used in this model. The model design was selected to optimize the computational time without significantly influencing the accuracy. However, each individual slab simulation required 4 hours or more of run time on a supercomputer. The concrete and dowel materials, as well as the interface between the dowel and concrete, were assumed to have the same properties as those used in the beam model. Each dowel could be rotated about the vertical or horizontal axes to simulate uniform and different types of nonuniform misalignments. To initiate the dowel-concrete contact, the same procedure used in the beam model was used. After that, prescribed longi- tudinal displacements were applied at opposite transverse edges of the modeled slab to simulate temperature contraction of the concrete slabs. The resulting joint opening induced damage in the concrete surrounding the misaligned dowels. The analysis showed that misalignment of 2 in. per 18 in. [51 mm per 457 mm] of vertical tilt causes higher stresses in the concrete surrounding the dowels than the aligned dowels. After the joint opening was simulated, the prescribed dis- placements were deactivated, and the simulated wheel load was then applied at the corner of the slab causing displacements of the system. The ratio of displacements of the loaded and unloaded slabs provides a measure of the LTE for different levels of dowel rotation. 2.3.3.2 Performance Prediction Models The MEPDG (AASHTO, 2008)does not include an input parameter for dowel misalignment. However, the laboratory and field data and analysis conducted in this study provide a means for incorporating the effect of dowel misalignment in the MEPDG. The following section discusses approaches for accounting for the effects of dowel misalignment in the models used in the MEPDG for JPCP joint faulting, transverse crack- ing, joint spalling, and IRI models. Faulting Model. Dowel diameter is one of the most impor- tant parameters of the faulting model. Dowel design has a major effect on the LTE of JPCP joints. If all other parameters are equal, a joint with a greater dowel diameter will have a higher LTE, which according to the MEPDG faulting model should reduce the rate of faulting development. The laboratory and analytical studies showed that dowel misalignment may affect dowel shear capacity and cause accelerated development of joint faulting. In the absence of an input parameter for dowel misalign- ment in the faulting model, the use of the equivalent dowel diameter concept to account for the effects of dowel misalign- ment is proposed. The equivalent dowel diameter is the dowel diameter that will yield the same dowel shear capacity of a mis- aligned dowel. This equivalent dowel diameter can then be used to investigate the effect of dowel misalignment on the long- term pavement performance using the MEPDG procedures. The equivalent dowel diameter concept postulates that, with regard to joint faulting, a joint with misaligned dowels behaves as a joint with aligned dowels with a diameter, deq, as defined by the following equation: where remb = the adjustment factor for a reduction in embedment length; rcc = the adjustment factor for a reduction in concrete cover; rvt = the adjustment factor for vertical tilt; rhs = the adjustment factor for horizontal skew; and d0 = the nominal dowel diameter. The adjustment factors can have values ranging from 0 to 1, where the value is inversely related to the level of misalignment. For a perfectly aligned dowel, all adjustment factors are equal to 1 and the corresponding equivalent dowel diameter is the same as the design dowel diameter. Conversely, dowels that are extremely misaligned in some way will have adjustment factors that approach zero. For example, if the dowel embed- ment length is equal to zero, then the adjustment factor remb is zero, making the equivalent dowel diameter zero, and the MEPDG faulting model would treat it as an undowelled pave- ment. Derivations of each individual dowel adjustment factor are presented in Chapter 3. Transverse Cracking Model. The field, laboratory, and finite element investigations conducted in this study could not link dowel misalignment with transverse cracking. Spalling Model. The MEPDG spalling model accounts only for the damage due to combinations of concrete degra- dation (controlled by the inputs of air content, water/cement ratio, and climate), and ability of incompressibles to penetrate the joint (controlled by the inputs of sealant type). The MEPDG spalling model does not contain any parameters related to dowel diameter or misalignment in either of these factors. The model addresses only spalling due to concrete degradation and age. Since dowel misalignment due to reduced concrete cover does not greatly influence the aging or wearing of joints, d r r r r deq emb cc vt hs= × × × × 0 4( ) 13

incorporating dowel misalignment into the original MEPDG spalling model could not be rationally suggested. On the other hand, dowel misalignment that reduces the concrete cover to the extent that the dowels are exposed at the surface is a road- way safety concern and is therefore considered in this study (recommendations for concrete cover to minimizing the risk of such critical spalling are presented in Chapter 3). IRI Model. The MEPDG IRI model considers changes in ride quality over time as a degradation of the initial smoothness due to transverse cracking, spalling, faulting, and pavement site conditions. The general equation for the IRI model is presented below to illustrate the compound effect of predicted transverse cracking, spalling, and faulting on prediction of pavement ride quality (AASHTO, 2008): IRI IRI C CRK C SPALL C TFAULT C SFI= + + + +1 2 3 4 5( ) where C1 through C4 = weighting factors or coefficients and IRI = predicted IRI, in./mi; IRII = initial IRI, in./mi; CRK = % slabs with transverse cracks (all sever- ities); SPALL = % joints with spalling (medium and high severities); TFAULT = total joint faulting cumulated per mi, in.; and SF = site factor. If each of these components is affected by dowel misalign- ment in some fashion, then the IRI model will account for the effects of dowel misalignment on IRI. Therefore, the MEPDG IRI model can be adopted in its current form to account for the effects of dowel misalignment. 14