LTPP Data Analysis: Influence of Design and Construction Features on the Response and Performance of New Flexible and Rigid Pavements (2005)

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.

291 CHAPTER 6 - ANALYSIS RESULTS FOR SPS-2 EXPERIMENT 6.1 INTRODUCTION A summary of findings from all analyses conducted for the SPS-2 experiment is presented in this chapter. The relevant statistical methods and analysis procedures were explained in Chapter 4. Analyses were performed both on performance and response (FWD) data. The performance measures that were analyzed include transverse and longitudinal cracking, faulting, and roughness (initial roughness and change in roughness). The structural response parameters that were analyzed include deflection parameters from mid-slab FWD testing (J1 testing). In this chapter, results from site-level analysis will be followed by results from overall analysis, and results from apparent relationship between response and performance, leading to the summary of findings. In addition, a narrative on the effect of construction on performance of SPS-2 sections and performance of each test section at each site is also presented. A list of all analyses conducted on the SPS-2 data is below: • Site-level analyses (on performance measures): Evaluation of the consistency of the effects of design factors across sites. • Overall analyses (on the performance and response measures): a) Extent of distress by experimental factors b) Analysis of Variance (ANOVA) c) Linear Discriminant Analysis (LDA) d) Binary Logistic Regression (BLR) • An investigation of apparent relationships between response and performance at the site level and for the overall population.

292 6.2 PREVIOUS FINDINGS A review was performed on LTPP studies that have identified factors affecting rigid pavement response and performance. A brief summary of findings from these studies and reports relevant to SPS-2 is presented here. Factors Affecting Rigid Pavement Performance PCC slab thickness • Thicker (279 mm) slabs experience reduced faulting, transverse cracking, spalling, and edge and corner deflections and hence reduced pumping [1-4]. Subgrade type [5-7] • Uniformity of support was identified as a dominant factor in the published literature. It was also concluded that weak subgrade provide non-uniform support that can lead to corner cracking and increased potential for voids. • Pavements resting on very stiff subgrade experience excessive curling and warping. Studies conducted in Chile have shown that pavements resting on stiff subgrade have resulted in cracking of 23% of slabs, on average. The subgrade fineness also has an impact on pavement performance. The finer subgrade is susceptible to pumping, erosion, frost heave and swells. Climate [5-8] • Moisture in LTPP is characterized by the number of wet days and location of site (wet or dry). Temperature in LTPP is characterized by location of site (non-freeze versus freeze), # of freeze -thaw cycles, annual mean temperature and number of days above 32oC. Pavements located in regions with high annual number of freeze thaw cycles, high number of wet days exhibited higher levels of spalling and faulting compared to others. An increase in the number of wet days from 80-130 significantly increased faulting levels. Pavements with un-doweled joints exposed to freeze-thaw cycles less than 70 experienced less faulting than un-doweled pavements exposed to freeze thaw cycles grater than 70. • Curling stresses in combination with heavy axles can increase the potential for occurrence of transverse cracking. These stresses have a substantial influence on performance of slabs

293 resting on stiff bases. Pavement sections located in freeze zones exhibited more roughness than pavements located in non-freeze zones. Another study [9] was conducted using SPS-2 data to identify the factors affecting pavement smoothness. A summary of main findings from this research is given below: • No statistical difference was found between initial roughness of the sections constructed with 200 mm (8”) thick and sections constructed with 275 mm (11”) thick PCC slab. • The highest early-age IRI was obtained for PCC surfaces placed on LCB while the lowest early-age IRI values were obtained for those placed on PATB. • The change in roughness that had occurred over the monitored period at the SPS-2 sections indicated different patterns. Some sections showed very high increase, while some showed a reduction in roughness. • The change in roughness, based on the profile data, could be related to changes in curvature of the PCC slabs. Both temperature-related (curling) and moisture-related (warping) curvatures were identified among the sections. • The time of day (temperature during profiling) is a cause for variation (increase or decrease) in roughness for some sections. • The section (or design) with 200 mm (8”) thick slab with PCC of 3.8 MPa (550-psi) 14-day flexural strength and built on DGAB is more susceptible to changes in curvature than the rest of the designs in SPS-2. Two studies[10, 11] were conducted to investigate the effects of sub-drainage on the performance of asphalt and concrete pavements. The following is a summary of observations from this research based on analysis of the SPS-2 data, for concrete pavements: • Un-drained pavement sections built on DGAB or on LCB may develop roughness, transverse and longitudinal cracking more rapidly than drained sections built on PATB. • The SPS-2 faulting data available through mid-June 2001 were too erratic to support meaningful statistical analysis.

294 • With respect to IRI change, larger mean differences were detected for the PATB sections with “poor” drainage than for PATB sections with “good” drainage, when un-drained and drained sections were compared. The quality of drainage is not a significant factor in the differences observed in IRI increase. • In the analyses of transverse and longitudinal cracking in drained versus un-drained SPS-2 sections, larger mean differences were detected for PATB sections with “good” drainage than for those with “poor” drainage. 6.3 EFFECT OF CONSTRUCTION ON PAVEMENT PERFORMANCE As mentioned in section 5.3 in Chapter 5, any abnormality in early performance was used as an indicator to identify sections exhibiting premature “failure”. The performance of all the sections, over time, was observed for this purpose and those sections that had premature “failure” (in first few years of service life) were identified. In order to further investigate the construction-related performance issues, the performance, with respect to each performance measure, for all pavement sections in the SPS-2 experiment was examined over time. This analysis helped minimize the bias, if any, in the results. The analysis is discussed next with illustrations. A brief discussion of construction-related performance issues, for each performance measure is presented in this portion of the report. Based on the time-series plots for all distress measures it was found that cracking (transverse and/or longitudinal) was the predominant premature “failure” for most of the pavement sections. Transverse and Longitudinal Cracking Figure 6-1 and Figure 6 - 3 show cracking in all the SPS-2 test sections, over time. It can be observed that some sections have conspicuously high initial cracking. It was found that most of the sections with this abnormal performance are from NV (32). A wide range of construction issues (material-related) that were reported in the construction report for the site is believed to be the cause (see site summary of NV in Appendix B1 for details). Figure 6 - 2 and Figure 6 - 4 show transverse and longitudinal cracking in all sections except those from NV (32). In light of the unusual behavior of test sections at NV (32), data from these sections was excluded from all statistical analyses.

295 0 50 100 150 200 250 0 1 2 3 4 5 6 7 8 9 10 11 12 Age (years) Tr an sv er se C ra ck in g (N o. ) Figure 6 - 1 Transverse cracking with time - All sections 0 50 100 150 200 250 0 2 4 6 8 10 12 Age (years) Tr an sv er se C ra ck in g (N o. ) Figure 6 - 2 Transverse cracking with time - Selected sections (without Nevada)

296 0 50 100 150 200 250 300 0 1 2 3 4 5 6 7 8 9 10 11 12 Age (years) Lo ng itu di na l C ra ck in g (m ) Figure 6 - 3 Longitudinal cracking with time - All sections 0 50 100 150 200 250 300 0 2 4 6 8 10 12 Age (years) Lo ng itu di na l C ra ck in g (m ) Figure 6 - 4 Longitudinal cracking with time - Selected sections (without Nevada)

297 Roughness and Joint Faulting Figure 6 - 5 shows the progression of roughness over time in all the SPS-2 sections except NV (32). It can be observed that only a few sections have exhibited an unusual performance. Exclusion of data from these sections was not considered necessary, as their inclusion will not impact the results considerably. Therefore, all the pavement sections [except those from NV (32)] were included in analyses regarding roughness. Figure 6 - 6 shows faulting growth over time in selected SPS-2 sections (i.e. without NV). As in the case of roughness, only a few sections have exhibited abnormal performance and exclusion of data from these sections was not considered necessary. Hence, all the pavement sections [except those from NV (32)] were included in analyses regarding faulting. In the above section of the report, issues related to the performance of the pavement sections were highlighted. Some construction and/or maintenance related issues with respect to the in-pavement drainage were identified in previous research [10, 11]. The in-pavement drainage for the rigid pavement sections was found to have some deviations from design.

298 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 1 2 3 4 5 6 7 8 9 10 11 12 Age (years) IR I ( m /k m ) Figure 6 - 5 IRI with time - Selected sections (without Nevada) 0 5 10 15 20 25 30 0 2 4 6 8 10 12 Age, years N o. o f j oi nt s w ith fa ul tin g > 1. 0m m Figure 6 - 6 Joint Faulting with time - Selected sections (without Nevada)

299 Drainage Issues All the drained sections of the SPS-2 experiment were video taped to assess the condition of the drainage in the project NCHRP 1-34C [11]. A subjective assessment of the quality of the drainage functioning as “good” or “poor” was reported for each section. The ratings assigned to each section of the experiment in Table 6-1. As shown in the table, some of the sections that were supposed to be un-drained, according to the experiment design, were constructed with drainage. A “poor” rating is an indication of; (i) buried lateral outlet, (ii) outlet fully blocked with silt, gravel or other debris (iii) longitudinal drains being fully blocked, or (iv) a considerable amount of standing water in the longitudinal drain. A “good” rating was given to drainage if a reasonably sufficient flow of water was evident even if some amount material was present in the drains. Hall et al [11] conducted preliminary analysis of the performance of SPS-2 test sections in light of their assessment of drainage, and a brief (paraphrased) summary of their findings are presented below: • Undrained pavement sections built on DGAB or on LCB may develop roughness, transverse and longitudinal cracking more rapidly than drained sections built on PATB. • The SPS-2 faulting data available through mid-June 2001 were too erratic to support meaningful statistical analysis. • With respect to IRI change, larger mean differences were detected for the PATB sections with “poor” drainage than for PATB sections with “good” drainage, when un-drained and drained sections were compared. The quality of drainage is not a significant factor in the differences observed in IRI increase. • In the analyses of transverse and longitudinal cracking in drained versus un-drained SPS-2 sections, larger mean differences were detected for PATB sections with “good” drainage functioning than for those with “poor” drainage functioning. However, the above trends were based only on the average performance and in no case, were the differences detected statistically significant. These findings regarding the functioning of drainage and the effect of drainage may be helpful during the interpretation of the results (from this study), regarding the effect of drainage.

300 Table 6- 1 Subjective ratings of drainage functioning for SPS-2 test sections based on video inspection results (Hall et al [11]) Test Section ID 0201 0213 0202 0214 0203 0215 0204 0216 0205 0217 0206 0218 0207 0219 0208 0220 0209 0221 0210 0222 0211 0223 0212 0224 Base Type Dense-graded aggregate base Lean concrete base Permeable asphalt-treated base over aggregate State Un-drained Drained AZ (4) G G G G AR (5) P P P P P P CA (6) P G G G P CO (8) G P P P DE (10) P G P G IA (19) P P ? ? KS (20) G G G G MI (26) P P P ? P ? NV (32) G G G ? NC (37) P P P P P ND (38) G G G G OH (39) ?* ?* ?* ?* ?* ?* ?* ?* G P G P WA (53) G G G G WI (55) ? ? ? ? 1G= Drainage function rated as good 2? = Drainage outlet not found 3P = Drainage function rated as poor 4?*= Camera could not be inserted

301 6.4 SITE-WIDE PERFORMANCE SUMMARIES This section summarizes the performance trends for each site within the SPS-2 experiment based on the latest available data (Release 17 of DataPave) at the time of writing this report. This is intended to help the reader gain an understanding of performance of test sections at each site. The performance measures discussed here include transverse and longitudinal cracking, wheelpath joint faulting, and roughness. Additional details about each of the sites can be found in site-level summaries presented in Appendix B1. A summary of performance of the test sections with “noticeable” distresses is in Appendix B2. A section is said to be exhibiting “noticeable” distress when a crack (transverse or longitudinal) or a wheelpath joint faulting of 2.0 mm or more is exhibited. Arizona, AZ (4) This site is located in the Dry No Freeze zone and built on coarse-grained subgrade soils. The site was opened to traffic in October 1993. The ‘proposed’ traffic volume is 1092 KESAL/ year. Any “noticeable” distress did not occur on section 214. In addition, no cracking (transverse or longitudinal) was observed in sections 215, 216 and 223. Transverse cracking was observed in sections 217 through 220. About 45% of the slabs are cracked in sections 217 and 218, whereas, less than 12% of slabs are cracked in sections 219 and 220. About half of the transverse cracks for section 217 are of medium severity. Longitudinal cracking of 52 m and 80 m occurred on sections 0213 and 0217, respectively, while cracking in sections 218, 221, 222 and 224 is less than 12.5 m. In section 0213, about 30% of longitudinal cracking is of medium or high severity. More than 40% of the joints in 6 out of the 12 sections have at least 1.0 mm of wheelpath joint faulting. Less than 3 joints in sections 215 through 219 and section 223 have faulted in excess of 2.0 mm. The initial IRI of the sections ranged from 1.0 to 1.4 m/km. The latest IRI measurement indicates that after about 11 years of service the IRI ranges from 1.1 to 1.9 m/km. Arkansas, AR (5) This test site is located in the Wet No Freeze zone and was opened to traffic in November 1995. All sections, except sections 222 and 223, were built on coarse-grained soils. The ‘proposed’

302 traffic volume is 1903 KESAL/ year. No “noticeable” distress was observed in sections 220, 221 and 223. Cracking (transverse or longitudinal) occurred only in sections 213, 217, 218 and 219. Longitudinal cracking of range 108 m to 160 m occurred on sections 213, 217 and 218. In sections 213 and 218, 25% and 40% of cracking, respectively, is of medium severity. About 88% of slabs have transverse cracks in section 218, and sections 213 and 217 have less than 10 % of slabs cracked. About 70% of cracking in section 218 is of medium severity. More than 50% of the joints have measurable faulting (> 1.0 mm) in sections 213 through 216 and sections 221, 222, and 224. In sections 213, 214, 215, 222, and 224 more than 20% of joints faulted at least 2.0 mm. The initial roughness of the sections at this site ranged from 0.9 to 1.6 m/ km. After about 9 years of service, the roughness (IRI) ranges from 1.1 to 2.3 m/ km. California, CA (6) This test site is located in Dry No Freeze zone and was built on coarse-grained soils. This is the youngest site of the experiment and was opened to traffic in October 2000. The ‘proposed’ traffic volume is 2405 KESAL/ year. No “noticeable” distress was observed in sections 204 and sections 209 through 212. No cracking (transverse or longitudinal) was observed in section 214 and in sections 209 through 212. Transverse Cracking occurred in sections from 201 through 208 except 204. In sections 201, 202, 205, and 206, 30% to 42% of the slabs exhibited transverse cracks. Less than 3 cracks were observed in sections 203, 207, and 208. In sections 201, 202, and 206, 60 to 70% of the cracking is of medium or high severity. Longitudinal cracking occurred only in sections 205 and 208 with about 29 m of cracking in 208 and less than 1.0 m of cracking in 205. All cracking is of low severity. Measurable faulting occurred in less than 3 joints in sections 201, 202, 205, and 212. The initial roughness of the sections at this site ranged from 0.9 to 1.7 m/km. The latest IRI measurement indicates that after about 4 years of service, the roughness (IRI) ranges from 1.3 to 2.0 m/km.

303 Colorado, CO (8) This test site is located in Dry Freeze zone and was opened to traffic in November 1993. Sections 214, 216, 219, 223, and 224 were built on coarse-grained soils while other sections were built on fine-grained soils. The ‘proposed’ traffic volume is 400 KESAL/year. No “noticeable” distress was observed in sections 214, 215, 216 and 221. One transverse crack each occurred in sections 218 and 222. Longitudinal cracking occurred only on sections 213, 217, and 222. Longitudinal cracking of about 21.0 m was observed in section 217 whereas less than 1.5 m of cracking occurred in sections 213 and 222. In section 217, about 38% of cracking is of medium severity. No measurable wheelpath joint faulting occurred in 215 and 222. In other sections, measurable faulting (1.0 mm or more) occurred at 9% to 42% of joints. The initial roughness of the sections at this site ranged from 1.0 to 1.8 m/ km. According to the latest data, the roughness (IRI) ranges from 1.2 to 1.8 m/ km. Delaware, DE (10) This test site is located in Wet Freeze zone and was built on coarse-grained soils. The site was opened to traffic in May 1996. The ‘proposed’ traffic volume is 380 KESAL/year. Transverse cracking (9 cracks) occurred only in section 205. No “noticeable” distress was observed in sections 202, 203, 204, 208 and 212. Longitudinal cracking occurred in sections 207 and 209 with magnitudes of 41 m and 3 m, respectively. In section 207 all cracking is of medium severity. Measurable faulting (1.0 mm and more) occurred in all sections except 203 and 208. At least 15% of the joints exhibited measurable faulting in sections 201, 202, 205, 206, and 209. Faulting more than 1.0 mm was observed at 2 to 5 joints in sections 205, 209, and 210. The initial roughness of the sections at this site ranged from 0.8 to 1.6 m/km. After about 8 years of service, the roughness (IRI) ranges from 0.8 to 1.9 m/km. Iowa, IA (19) This test site is located in Wet Freeze zone and was built on fine-grained soils. The site was opened to traffic in December 1994. The ‘proposed’ traffic volume is 377 KESAL/year. No “noticeable” distress was observed in sections 214, 218, 219, 220, 222 and 223.

304 Transverse cracking, in 6% of slabs, occurred in section 217. Longitudinal cracking, of total length less than 5.0 m, occurred on sections 213, 222 and 224. All longitudinal cracking is of medium severity. In sections 214 through 218, and 221, more than 30% of joints have measurable (1.0 mm or more) wheelpath joint faulting. Two joints each in sections 215 and 216 have faulted in excess of 2.0 mm. The initial roughness of the sections at this site ranged from 1.0 to 2.2 m/km. The latest IRI measurement indicates that after about 8 years of service, the roughness (IRI) ranges from 1.1 to 2.0 m/km. Kansas, KS (20) This test site is located in the Wet Freeze zone and was built on fine-grained soils. The sections at this site, which is the oldest site in the experiment, were opened to traffic in August 1992. The ‘proposed’ traffic volume is 757 KESAL/year. No “noticeable” distress was observed in sections 208 and 209. Transverse cracking occurred in sections 201 and 202. 4 transverse cracks occurred in 201 and 2 cracks occurred in 202. Longitudinal cracking of length less than 7.0 m occurred in sections 201, 206 and 208. Measurable faulting occurred at 20% to 45% of joints in all sections except 202, 204, 207, 209, and 211. Moreover, faulting of 2.0 mm or more occurred at 1 or 2 joints of all sections except 209 and 211. The initial roughness of the sections at this site ranged from 1.1 to 2.1 m/km. After about 12 years of service, the roughness (IRI) ranges from 1.0 to 2.0 m/km. Michigan, MI (26) This test site is located in the Wet Freeze zone and was built on fine-grained soils. The site was opened to traffic in November 1993. The ‘proposed’ traffic volume is 1505 KESAL/ year. Sections 213, 215, 217, and 218 have been de-assigned from the experiment after rehabilitation was done to the sections. Among the sections that are in the experiment, only 214 and 218 have “noticeable” distress. Among the sections that are in the experiment, transverse cracking (5 cracks) was observed in section 214.

305 Measurable faulting occurred at less than 20% of joints in all sections (in the experiment) except 216. Faulting greater than 1.0 mm occurred at less than 12% of joints in sections 223 and 214. The initial roughness of the sections at this site ranged from 0.9 to 1.8 m/km. Roughness (IRI), as per the latest data, ranges from 1.1 to 4.1 m/km. Nevada, NV (32) This test site is located in Dry Freeze zone and was opened to traffic in September 1995. All sections except 201 and 205 were built on coarse-grained soils. The ‘proposed’ traffic volume is 800 KESAL/year. Section 212 had severe cracking following paving and it was replaced with nonconforming materials. Thus the section was removed from the experiment in 1995. In addition, sections 202 and 206 were de-assigned from the experiment following rehabilitation work. All the sections exhibited “noticeable” distress. In sections 203 and 205, 168 and 221 transverse cracks occurred, of which 80 to 95 % of is medium or high severity cracking. In all sections, except 206 and 209, at least 70 % of cracking is of medium or high severity. Also, more than 125 m of longitudinal cracking occurred in the same sections. In sections 203, 205, and 207, 70 to 95% of cracking is of medium to high severity. Except in sections 201 and 210, at least 20% of the joints have measurable faulting. Faulting greater than 1.0 mm was observed at 2 joints each in sections 205 and 207. The initial roughness of the sections at this site ranged from 0.8 to 1.6 m/km. Roughness (IRI), as per the latest data, ranges from 1.1 to 2.5 m/ km. North Carolina, NC (37) This test site is located in Wet No Freeze zone and was built on fine-grained soils. Traffic was opened on the site in July 1994. The ‘proposed’ traffic volume is 715 KESAL/year. “Noticeable” distress was observed in sections 201, 202, 204, 205, and 210. Transverse cracking was observed in sections 201 and 205. A total of 12 transverse cracks (36% of slabs) occurred in section 205 while 1 crack occurred in section 201. Longitudinal cracking occurred in 203, 205 and 210. A total of 6 m of longitudinal cracking occurred in section 205 whereas cracking of less than 1.0 m length occurred in 205 and 210. All cracking (transverse and longitudinal) is of low severity.

306 Measurable faulting occurred at 20% to 35% of joints in sections 201, 202, 206, and 210. The initial roughness of the sections at this site ranged from 1.1 to 1.6 m/km. Roughness (IRI), as per the latest data, ranges from 1.1 m/ km to 1.8 m/km. North Dakota, ND (38) This test site is located in Wet Freeze zone and was opened to traffic in November 1994. The site was built on fine-grained soils. The ‘proposed’ traffic volume is 420 KESAL/year. “Noticeable” distress did not occur in sections 213, 221, 222, and 223. 8 transverse cracks (half of high severity) in 217 and 1 crack each in sections 219, 220, and 224 were observed. No cracking occurred in other sections. Longitudinal cracking was observed in sections 217, 218, and 224. Section 217 exhibited longitudinal cracking of total length equal to 75 m whereas 218 and 224 exhibited cracking less than 10 m. Almost all of the cracking is of medium or high severity in sections 217 and 224. Except for sections 223 and 218, all the other sections had measurable faulting at 30% of joints or more. Among sections with “noticeable” distress, all sections except 218 have faulting greater than 1.0 mm at 1 to 6 joints. The initial roughness of the sections at this site ranged from 1.2 to 2.0 m/km. After about 10 years of service, the roughness (IRI) ranges from 1.1 to 1.9 m/km. Ohio, OH (39) This test site is located in Wet Freeze zone and was opened to traffic in October 1996. The site was built on fine-grained soils. The ‘proposed’ traffic volume is 608 KESAL/year. No “noticeable” distresses were observed in sections 203, 207 and 208. Longitudinal cracking did not occur on any of the test sections. Sections 203, 207, 208, and 211 exhibited no transverse cracking. Section 205 exhibited 25 transverse cracks (in 76% of slabs) while other cracked sections exhibited less than 7 cracks. 10 cracks in 205 and 6 cracks (out of 7) in 210 are of medium or high severity. Sections 203, 204, 207, and 208 had at least 20% of the joints that faulted 1.0 mm or more. Sections 204, 205, and 211 had one joint each with measurable faulting. The initial roughness of the sections at this site ranged from 0.9 to 1.5 m/km. After about 8 years of service, the roughness (IRI) ranges from 0.9 to 1.8 m/km.

307 Washington, WA (53) This test site is located in Dry Freeze zone and was opened to traffic in November 1995. The site was constructed on fine-grained soils. The ‘proposed’ traffic at the site is 462 KESAL/year. “Noticeable” distress occurred only in sections 205, 206 and 212. Transverse cracks (less than 5 cracks i.e. 15% of slabs) occurred in sections 205 and 206. Longitudinal cracking occurred in section 206 with a total length of 4 m (low severity). Measurable faulting was recorded in all sections except 202, 209 and 210, and 211. Among the sections that have faulted joints, less than 6 joints have faulting of 1.0 mm. The initial roughness of the sections at this site ranged from 0.8 to 1.2 m/km. Roughness (IRI), as per the latest data, ranges from 0.8 to 1.8 m/km. Wisconsin, WI (55) This test site is located in Dry Freeze zone and was opened to traffic in November 1997. The site was constructed on coarse-grained soils. The ‘proposed’ traffic volume is 462 KESAL/year. “Noticeable” distress occurred in sections 216 and 222. No cracking (transverse or longitudinal) has occurred on any of the sections. While 12% to 36% joints in 213, 214, 219, 221, and 222 had measurable faulting, 15% of joints in 222 faulted more than 1 mm. The initial roughness of the sections at this site ranged from 0.8 to 1.6 m/ km. Roughness (IRI), as per the latest data, ranges from 0.8 to 1.6 m/ km.

308 6.5 SITE-LEVEL ANALYSES This section of the report is a discussion of the results obtained from site-level analyses of the SPS-2 experiment data. The concepts of performance index (PI) and relative performance were used to perform site-level analyses (details in Chapter 4), as in the case of SPS-1 experiment. These analyses were conducted separately for each performance measure. These performance measures are: • Transverse cracking, • Longitudinal cracking, • Faulting, and • Roughness (IRI). Site-level analyses deal with each SPS-2 project separately. For each site, the climatic conditions, subgrade type (for most of the sites) and traffic are same. Construction conditions, material sources and surveys were also considered to be same for all sections within each SPS-2 site. As described in Chapter 4, the site-level analyses consists of two types of comparisons: (i) Level-A — In this analysis all designs (201 through 212, or 213 through 224) at a given site are compared (among themselves) such that only one factor (design feature) is held common within the sections of each group under comparison; (ii) Level-B — In this analysis, most of the factors (design features) are “controlled” for comparisons. The analysis process is summarized in Figure 6-7. The results from level-A and level-B comparisons, in terms of relative performance ratio, can be found in Appendix B4. Non-parametric tests (Wilcoxon Signed Ranks test and Friedman test) were performed on relative performance ratio to determine the statistical significance of the difference in relative performance ratio of different levels within each design factor. For example, the relative performance ratio corresponding to transverse cracking, for sections with 203 mm (8-inch) slab and sections with 279 mm (11-inch) slab were compared to investigate the statistical significance of the consistency of the effect of PCC slab thickness on transverse cracking across sites. A p- value less than or equal to 0.05 was considered to be indicative of a statistically significant consistency of an effect.

309 In site-level analyses, statistical significance of an effect needs to be interpreted as the significance of the effect’s consistency across sites but not necessarily as significance of its effect on the magnitude of distress. In this chapter, the discussion of results for level-A and level-B analyses is presented separately. Some basic descriptive statistics regarding the performance of the test sections are also presented, to corroborate the results. Though these statistics are not at site-level they are meant to give the reader an insight about the extent/occurrence of distresses. In the SPS-2 experiment, only the test sections built on PATB were provided with in- pavement drainage. As a consequence of this, the impact of drainage alone or base type alone cannot be studied. In other words, the effect of PATB and the effect of drainage cannot be separated. Therefore, an assumption was made that DGAB and PATB are structurally the “same” (as in the case of SPS-1 experiment [12]), and the analysis was performed by comparing performance of sections constructed on DGAB and sections constructed on PATB. It is important to note that the effect of drainage discussed in this report would be a result of comparison between sections on DGAB and sections on PATB. Furthermore, to study the effect of base type, the performance of sections with DGAB, sections with LCB and sections with PATB were compared. Here too the effect of PATB is fused with the effect of drainage.

310 Figure 6 - 7 Methodology for site-level analyses (SPS-2) Site Level Analysis Level-A Comparisons Level-B Comparisons Effect of Drainage Yes vs. No Effect of PCC Thickness 203 mm vs. 279 mm Effect of Base Type DGAB vs. LCB vs. PATB/DGAB Effect of Drainage Yes vs. No Controlling for other factors Effect of Base Type DGAB, LCB, PATB/DGAB Controlling for other factors Effect of PCC Thickness 203 mm vs. 279 mm Controlling for other factors Effects of design features and site factors Effect of Flexural Strength 3.8 MPa vs. 6.2 MPa Effect of Lane width 3.7 m vs. 4.3 m

311 6.5.1 Effect of design features on performance- Comparisons at level-A The discussion of results from level-A analyses is presented here. These results are presented taking one design feature at a time. Drainage To investigate the effects of drainage, sections 201 through 204, and, 213 through 216 were considered as “without drainage” and sections 209 through 212, and, 220 through 224 were considered as “with drainage”. Hence it is important to note that the effects of drainage that are discussed here are from comparisons only between sections built on DGAB and sections built on PATB. From level-A analysis, the effects of drainage on cracking, faulting and roughness are inconclusive, at this point in time. This observation should not be interpreted as drainage not having a significant impact on pavement performance in general. All the observations and conclusions need to be interpreted keeping in view the age of the test sections and the low occurrence of distresses in the SPS-2 test sections. Table 6- 2 is the summary of effects of drainage on cracking (transverse and longitudinal). The effect of drainage on cracking in different climates is also inconclusive. Table 6- 3 is a summary of results obtained from level-A analysis on wheelpath joint-faulting and roughness. Base Type Sections built on each of the three base types, DGAB, LCB and PATB, were compared at each site to study their relative impact on performance. The analysis is a comparison among 56 sections built on DGAB, 56 sections built on LCB and 55 sections built on PATB. Base type was found having a consistent effect on cracking. However, the effect is not consistent (across sites) for faulting and roughness. Approximately 59% of sections built on LCB have exhibited cracking compared to 38% of sections built on DGAB and 25% of sections built on PATB. Though the analysis indicates higher cracking in sections built on LCB, the conclusions need to be considered in light of the construction issues (details in Appendix B1) and, the magnitude and severity of cracking. Table 6- 4 is the summary of the effects of base type on cracking.

312 Table 6-5 is the summary of the effects of base type on faulting and roughness. In general, the trend of faulting suggests higher faulting in the sections built on DGAB. The effect of base type on roughness seems to be inconclusive because at most of the sites the difference in IRI of sections built on the three base types is not considerably high. As of latest distress survey, 80% of the sections in the experiment have IRI less than 1.8 m/km. The extent and magnitude of the distresses are to be considered along with the conclusions.

313 Table 6- 2 Effects of drainage on cracking, based on Level-A analysis Design Factor Performance Measure Effect Comments Transverse cracking Inconclusive (p=0.299) • In 5 of the 14 sites, no cracking occurred and the performance of sections with and without drainage is thus similar. • In 6 of the 9 sites with distressed sections, sections without drainage exhibited more cracking than ones with drainage. • Overall, 25% of sections without drainage and 12% of sections with drainage have exhibited cracking. • 21% of sections in WF zone, 13% of sections in WNF, 35% of sections in DF zone, and 19% of sections in DNF zone have exhibited cracking. Drainage* Longitudinal cracking Inconclusive (p= 0.411) • In 5 of the 9 sites with distressed sections, sections without drainage exhibited more cracking than ones with drainage. • In 5 of the 14 sites, no cracking occurred and the performance of sections with and without drainage is thus similar. • Overall, 21% of sections without drainage and 19% of sections with drainage have exhibited cracking. • 9% of sections in WF zone, 19% of sections in WNF, 35% of sections in DF zone, and 25% of sections in DNF zone have exhibited cracking. *Effect of drainage is a result of comparison between sections built on DGAB and sections built on PATB. Table 6- 3 Effects of drainage on faulting and roughness, based on Level-A analysis Design Factor Performance Measure Effect Comments Wheelpath joint-faulting (>1.0 mm) Inconclusive (p= 0.699) • In 7 of the 14 sites, sections without drainage exhibited more faulting than ones with drainage. • In 5 of the 14 sites, sections with drainage performed poorer than sections without drainage. • 46% of sections without drainage and 31% of sections with drainage have faulting>1.0 mm (at one joint or more). • 45% of sections in WF zone, 56% of sections in WNF, 8% of sections in DF zone, and 25% of the sections in DNF zone have exhibited faulting>1.0 mm (at one joint or more). Drainage* Roughness (IRI) Inconclusive (p= 0.084) • In 11 of the 14 sites, the performance of sections with and without drainage is comparable. • Average latest roughness of sections without drainage and sections with drainage are 1.6 and 1.3 m/km. *Effect of drainage is a result of comparison between sections built on DGAB and sections built on PATB.

314 Table 6- 4 Effects of base type on cracking based on Level-A analysis Design Factor Performance Measure Effect Comments Transverse cracking Consistent effect (p= 0.000) • In all the 13 sites with distressed sections, higher cracking was observed in sections built on LCB, compared to other sections. • 25%, 46% and 12% of sections on DGAB, LCB, and PATB, respectively, exhibited cracking. Base type Longitudinal cracking Consistent effect (p= 0.002) • In 11 of the 13 sites with distressed sections, higher cracking was observed in sections built on LCB, compared to other sections. • 21%, 42% and 19% of sections on DGAB, LCB, and PATB, respectively, exhibited cracking. Table 6- 5 Effects of base type faulting and roughness based on Level-A analysis Design Factor Performance Measure Effect Comments Wheelpath joint-faulting (>1.0 mm) Inconclusive (p= 0.238) • In 7 of the 14 sites, more faulting was observed in sections built on DGAB, compared to other sections. In 3 sites sections on PATB and in 2 sites sections on LCB had higher faulting. • 46%, 37% and 31% of sections on DGAB, LCB, and PATB, respectively, have faulting>1.0mm, at one joint or more. • 43% of sections in WF zone, 42% of sections in WNF zone, 23% of sections in DF, and 29% of sections in DNF zone exhibited faulting. Base type Roughness (IRI) Inconclusive (p= 0.064) • In 10 of the 14 sites, comparable roughness was observed in all sections. • In 4 of the 14 sites, more roughness was observed in sections built on DGAB, compared to other sections. • Average latest roughness of sections on DGAB, LCB, and PATB are 1.6, 1.6 and 1.3 m/km, respectively.

315 PCC slab thickness A total of 84 sections with 203 mm (8-inch) PCC slab and 83 sections with 279 mm (11- inch) PCC slab were compared (at site-level) for this analysis. This includes all the sections in the experiment. The effect of slab thickness is consistent in the case of transverse and longitudinal cracking. Though a deviation from target thickness was observed in considerable number of sections (details in Chapter 3), the analysis indicates a significant effect of PCC slab thickness on cracking. Table 6- 6 is the summary of effects of PCC slab thickness on cracking. It is to be noted here that 49% of cracking (transverse and/or longitudinal) has occurred in sections that were built on LCB, of which 67% of the sections are sections with 203 mm (8-inch) PCC slabs. These statistics suggest a noticeable effect of both slab thickness and base type. The effect of PCC slab thickness on faulting and roughness is summarized in Table 6- 7. Sections constructed with 203 mm (8-inch) PCC slab had slightly more faulting in 5 sites while a reverse trend was observed in 4 sites. In addition, the roughness of both 203 mm (8-inch) PCC slab and 279 mm (11-inch) PCC slab sections was found to be comparable at all sites, suggesting an insignificant effect of slab thickness on roughness. The effect of PCC slab thickness on faulting and IRI is thus inconclusive.

316 Table 6- 6 Effects of slab thickness on cracking, based on Level-A analysis Design Factor Performance Measure Effect Comments Transverse cracking Consistent effect (p= 0.001) • In all sites that have distressed sections (13 sites), more cracking was observed in sections with 203 mm (8-inch) PCC slab, compared to sections with 279 mm (11”) PCC slab. • 40% of sections with 203 mm (8-inch) PCC slab and 15% of sections with 279 mm (11”) PCC slab exhibited cracking. PCC slab thickness Longitudinal cracking Consistent effect (p= 0.020) • In 11 of the 13 sites that have distressed sections, more cracking was observed in sections with 203 mm (8-inch) PCC slab, compared to sections with 279 mm (11”) PCC slab. • 41% of sections with 203 mm (8-inch) PCC slab and 14% of sections with 279 mm (11”) PCC slab exhibited cracking. Table 6- 7 Effects of slab thickness on faulting and roughness, based on Level-A analysis Design Factor Performance Measure Effect Comments Wheelpath joint-faulting (>1.0 mm) Inconclusive (p= 0.665) • In 5 of the 14 sites, more faulting was observed in sections with 203 mm (8-inch) PCC slab, while, in 4 sites a reverse trend was observed. • 36% of sections with 203 mm (8-inch) and 40% of sections with 279 mm (11”) PCC slab have faulting>1.0 mm (at one joint or more). PCC slab thickness Roughness (IRI) Inconclusive (p= 0.414) • In all the sites, comparable performance was observed in all sections. • Average latest roughness of both the sections with 203 mm (8-inch) PCC slab and sections with 279 mm (11”) PCC slab is 1.5 m/km.

317 PCC flexural strength The performance of test sections with target 14-day PCC flexural strength of 3.8 MPa and test sections with target 14-day PCC strength of 6.2 MPa was compared to study the effect of PCC flexural strength on the performance of SPS-2 sections. A total of 84 sections with 3.8 MPa concrete and 83 sections with 6.2 MPa concrete were compared. The effect of flexural strength on cracking, faulting and roughness appears to be insignificant. Comparable performance was observed in sections with higher strength concrete (6.2 MPa) and lower strength concrete (3.8 MPa). It is important to consider the deviations from target flexural strength in the sections. A detailed discussion of the deviations was presented in Chapter 3. The deviation from target PCC 14-day flexural strength was studied using the data that is available for 52% of the sections in the experiment. The average 14-day flexural strength of PCC of sections with target strength of 3.8 MPa was 3.6 MPa while in sections with target strength of 6.2 MPa was 5.6 MPa. Among sections with target flexural strength of 3.8 MPa, 34% of sections had PCC flexural strength (at 14-days) that exceeded the allowable range of 3.4 MPa to 4.2 MPa, while 16% failed to reach even the lower limit of the range. In the case of sections with target flexural strength of 6.2 MPa, 34% of sections had PCC flexural strength (at 14-days) below the allowable range, and none of the sections exceeded the range. In half of the sections with target strength of 6.2 MPa that failed to meet the lower limit of the range, the PCC strength reached the required limit at 28-days. These deviations from target strength could be a reason for comparable performance of all the pavements. Table 6- 8 is the summary of effects of PCC flexural strength on cracking. At most of the sites, comparable performance (cracking) was observed for both higher strength and lower strength concrete sections. The effect is thus inconclusive. The effect of PCC flexural strength on faulting and roughness is summarized in Table 6- 9. In light of low occurrence of faulting, the effect of PCC flexural strength is inconclusive. Similarly, the effect of PCC flexural strength on roughness appears to be insignificant.

318 Table 6- 8 Effects of flexural strength on cracking, based on Level-A analysis Design Factor Performance Measure Effect Comments Transverse cracking Inconclusive (p=0.400) • In 6 of the 14 sites, lower strength concrete sections exhibited higher cracking than higher strength concrete sections. • In 5 of the 14 sites, all sections have performed at comparable levels. • 26% of lower strength concrete sections and 25% of higher strength concrete sections exhibited cracking. PCC flexural strength Longitudinal cracking Inconclusive (p=0.944) • In 7 of the 14 sites, lower strength concrete sections exhibited higher cracking than higher strength concrete sections. • In 5 of the 14 sites, higher strength concrete sections exhibited higher cracking than lower strength concrete sections. • 26% of lower strength concrete sections and 25% of higher strength concrete sections exhibited cracking. Table 6- 9 Effects of flexural strength on faulting and roughness, based on Level-A analysis Design Factor Performance Measure Effect Comments Wheelpath joint- faulting (>1.0 mm) Inconclusive (p=0.925) • In 12 of the 14 sites, sections with higher strength concrete and lower strength concrete exhibited comparable level of performance. • 31% of lower strength concrete sections and 40% of higher strength concrete sections have faulting>1.0 mm (at one joint or more). PCC flexural strength Roughness (IRI) Inconclusive (p=0.102) • In all sites of the experiment, similar performance was observed in sections with higher strength concrete and sections with lower strength. • Average latest roughness of lower or higher strength concrete sections is 1.5 m/km.

319 Lane width The widened lane [4.3 m (14 ft)] sections were compared to those with standard lane [3.7 m (12 ft)] to study the effect of lane width on performance of the test sections. For this, 84 sections with standard lane width and 83 sections with widened lane were compared. Some effect of lane width was observed only in the case of faulting. The effect of lane width on other distresses seems to be insignificant. In general, the effect of lane width on transverse cracking seems to be insignificant as widened lane sections and standard lane sections have performed similarly in most of the sites. But some effect of lane width seems to exist on longitudinal cracking. In a majority of the sites (9 of 13) slightly higher longitudinal cracking was observed in widened lane sections compared to standard lane sections. This could be due to the geometry of the wider lane (4.3 m) that causes greater transverse bending stresses in widened lanes, as opposed to a standard lane (for same loading). Table 6- 10 is the summary of effects of lane width on cracking. A consistent effect of lane width was found on wheelpath joint-faulting, in that, sections with standard lane experienced more faulting that ones with wider lane (4.3 m) at most of the sites. This could be because of the greater distance of wheelpath from edge in the case of wider lane (4.3 m) that causes less corner stresses in wider lane (4.3 m) sections. Table 6- 11 presents a summary of the lane width effect on faulting and roughness. The effect of lane width on roughness is inconclusive at this point in time.

320 Table 6- 10 Effects of lane width on cracking based on Level-A analysis Design Factor Performance Measure Effect Comments Transverse cracking Inconclusive (p=0.222) • In 6 of the 14 sites, sections with standard lane exhibited higher cracking than ones with wider lane (4.3 m). • In 5 of the 14 sites, both standard lane and wider lane (4.3 m) sections have shown comparable levels of performance. • 28% and 24% of sections with standard lane and wider lane (4.3 m), respectively, have exhibited cracking. Lane Width Longitudinal cracking Inconclusive (p=0.362) • In 9 of the 14 sites, sections with wider lane (4.3 m) exhibited higher cracking than ones with wider lane (4.3 m). • 25% and 26% of sections with standard lane and wider lane (4.3 m), respectively, have exhibited cracking. Table 6- 11 Effects of lane width on faulting and roughness based on Level-A analysis Design Factor Performance Measure Effect Comments Wheelpath joint- faulting (>1.0 mm) Consistent effect (p=0.003) • In 9 of the 14 sites, sections with standard lane exhibited higher faulting than ones with wider lane (4.3 m). • 39% of standard lane sections and 32% of wider lane (4.3 m) sections have faulting>1.0 mm (at one joint or more). Lane Width Roughness (IRI) Inconclusive (p=0.096) • In all sites of the experiment, similar performance was observed in sections with standard lane and sections with wider lane (4.3 m). • Average latest roughness of standard lane sections and wider lane (4.3 m) sections are 1.6 and 1.5 m/km.

321 6.5.2 Effect of design features- Paired Comparisons at Level-B Level-B comparisons are those in which all possible factors other than the one of interest are controlled. The individual sections that are compared under this analysis were identified in chapter 4. The effects of drainage, base type, and PCC slab thickness on the performance measures are presented below. Drainage Sections with drainage (i.e. sections with PATB) were compared with sections without drainage (sections with DGAB) controlling the effects of all other factors, namely, PCC slab thickness, lane width, and flexural strength. The effect of drainage is consistent (across sites) on transverse cracking. Slight effect was observed on roughness, whereas no effect was apparent in the case of faulting and longitudinal cracking. For sections with 203 mm (8”) PCC slab, the effect of drainage seems to be consistent on transverse cracking. Table 6- 12 is the summary of effects of drainage on cracking. The effect of drainage on longitudinal cracking is inconclusive. Sections with drainage and without drainage performed similarly in varying conditions. Table 6- 13 is a summary of effects of drainage on faulting and roughness. No consistent effect seems to exist on the occurrence of faulting. In general, more sections without drainage have faulted than ones with drainage. Among sections with 203 mm (8”) slab, a slight effect of drainage (p=0.076) appears to exist on roughness in those with standard lane. Sections without drainage have slightly higher roughness than ones with drainage.

322 Table 6- 12 Effect of drainage on cracking, based on Level-B analysis Design Factor Performance Measure Effect Comments Transverse cracking Consistent effect (p=0.034) • At 5 of the 7 sites with distresses sections, among sections 203 mm (8”) PCC slab, sections without drainage cracked more than sections with drainage. • Among 203 mm (8”) PCC slab, 18% and 43% of sections with drainage and without drainage have exhibited transverse cracking. • Cracking was observed only at two sites in the thicker (279 mm) slab sections. Drainage* Longitudinal cracking Inconclusive • No discernable trends were observed for longitudinal cracking. • Among 203 mm (8”) PCC slab, 25% of sections with drainage and 25% of sections without drainage have exhibited longitudinal cracking. *Effect of drainage is a result of comparison between sections built on DGAB and sections built on PATB. Table 6- 13 Effect of drainage on faulting and roughness, based on Level B analysis Design Factor Performance Measure Effect Comments Wheelpath joint- faulting (>1.0 mm) Inconclusive • In most of the sites, among 203 mm (8”) PCC slab, the sections without drainage faulted more than sections with drainage. • Among the 203 mm (8”) PCC slab, 25% of sections with drainage and 36% of sections without drainage exhibited faulting. Drainage* Roughness (IRI) Slight effect (p=0.076) • Effect of drainage seems to be negligible as the sections with drainage and without drainage have performed similarly in most of the sites. • Among the 203 mm (8”) PCC slab, the average roughness of sections with drainage and without drainage are 1.3 and 1.6 m/km. *Effect of drainage is a result of comparison between sections built on DGAB and sections built on PATB.

323 Base Type Sections built on each of the three base types, DGAB, LCB and PATB, were compared at each site by controlling the effects of PCC slab thickness and lane width. A consistent effect of base type on transverse and longitudinal cracking was observed. The effect of base type on faulting and roughness is not clear. On average, among sections built with 203 mm (8-inch) slabs, those built on LCB have exhibited higher transverse cracking than other sections (see Table 6-14). This trend was observed in a majority of sites. Among sections with 203 mm (8-inch) slab and standard lane width, the trend is consistent (p=0.001) across the sites. This effect may be an “interaction effect”, as the effect of base type was discernable among sections with 203 mm (8-inch) slab and standard lane, and not in sections with 279 mm (11-inch) slab. The effect of base type on longitudinal cracking (see Table 6-14) is consistent among sections with 203 mm (8-inch) slab, in that sections built on LCB have higher cracking than those on other bases. The effect of base type on faulting is inconclusive. In general, a slight effect of base type was observed on faulting, in that sections built on DGAB had higher faulting than other sections at a majority of the sites, irrespective of other design features. Table 6-15 is a summary of effects of base type on faulting and roughness. The effect of base type on roughness is inconclusive. PCC slab thickness The performance of sections with target PCC slab thickness of 203 mm was compared with that of sections with target PCC slab thickness of 279 mm by controlling the effects of base type and PCC flexural strength. The effect of PCC slab thickness was consistent on cracking (transverse and longitudinal), whereas no noticeable effect was found on faulting and roughness. Among sections built with DGAB or LCB, sections with 8-inch (203 mm) slab had higher cracking than sections with 11-inch (279 mm) slab. Table 6-6 is the summary of effects of PCC slab thickness on cracking. In the case of longitudinal cracking, the effect was found consistent among sections built with LCB with higher cracking in sections with 8-inch (203 mm) slab. Table 6- 17 is a summary of the effect of PCC slab thickness on faulting and roughness. The effect of PCC slab thickness on faulting and roughness is inconclusive.

324 Table 6- 14 Effect of base type on cracking, based on Level B analysis Design Factor Performance Measure Effect Comments Transverse cracking Consistent effect (p<0.05) • In 9 sites, among thinner slab (203 mm) sections, sections built on LCB cracked more than sections on other base types. • 43%, 64% and 18% of sections on DGAB, LCB and PATB have exhibited cracking among thinner (203 mm) slab sections. Base type Longitudinal cracking Consistent effect (p<0.05) • In 8 sites, of the 12 sites at which cracking was observed, sections built on LCB exhibited more cracking sections built on other base types. • Among the thinner slab (203 mm) sections, 54% of sections built on LCB and 25% of other sections exhibited cracking. Table 6- 15 Effect of base type on faulting and roughness, based on Level B analysis Design Factor Performance Measure Effect Comments Wheelpath joint- faulting (>1.0 mm) Inconclusive • Those built on DGAB seem to be experiencing higher faulting than other sections, especially among sections built on fine-grained soils. • 39% of sections built on DGAB, 34% of sections built on LCB and 26% of sections built on PATB have faulting >1.0 mm, at one or more joints. Base type Roughness (IRI) Inconclusive Effect of base type seems to be negligible as all the sections performed similarly, in general. Table 6- 16 Effect of slab thickness on cracking, based on Level B analysis Design Factor Performance Measure Effect Comments Transverse cracking Consistent effect (p<0.05) Thinner slab (203 mm) sections exhibit more transverse cracking than thicker slab (279 mm) sections, among sections with DGAB or LCB. PCC slab thickness Longitudinal cracking Consistent effect (p<0.05) Thinner slab (203 mm) sections exhibit more longitudinal cracking than thicker (279 mm) slab, among sections with LCB. Table 6- 17 Effect of slab thickness on faulting and roughness, based on Level B analysis Design Factor Performance Measure Effect Comments Wheelpath joint- faulting (>1.0 mm) Inconclusive Effect of base type seems to be negligible as all the sections perform similarly PCC slab thickness Roughness (IRI) Inconclusive Effect of base type seems to be negligible as all the sections perform similarly

325 6.6 OVERALL ANALYSIS The results obtained from statistical analyses performed on the SPS-2 data are presented in this section. Both the performance and response variables were analyzed to study the effects of various design and site-factors on the pavement sections. Analyses were performed combining all data and is referred to as ‘Overall’ analyses. Analyses were also conducted in each climatic zone combining data from all sections within a zone as per the recommendation of the project panel. Linear Discriminant Analysis (LDA), Binary Logistic Regression (BLR), and Analysis of Variance (ANOVA) are the statistical methods that were employed for analyses. Analysis of Performance Measures The performance measures that were analyzed to investigate the impact of design and site factors on rigid pavement performance are as follows: • Transverse cracking, • Longitudinal cracking, • Wheel path joint-faulting, and • Roughness (IRI)- initial roughness and change in roughness. The significance of factors affecting occurrence of transverse cracking, longitudinal cracking, faulting and roughness were determined using LDA and BLR, which are frequency- based methods (details in Chapter 4). For analyses on cracking, the test sections were grouped into 2 categories−sections with cracking, and sections without cracking. Faulting greater than 1.0 mm was considered to be “noticeable” (distress) given the low levels of faulting in test sections. In LDA and BLR analysis on faulting, sections with at least one occurrence of noticeable faulting were categorized as one group, and the other sections were categorized as another. Analyses were then performed to identify factors that significantly discriminate the groups. In the case of IRI, a threshold of 1.5 m/km was used to separate the test sections as groups. The value corresponds to the threshold between “normal” and “poor” pavements [13] for an age of 7 years, which is the average age of SPS-2 test sections. As mentioned in Chapter 4, ANOVA was used to determine the significance of factors impacting initial roughness (IRI), change in roughness (first survey to latest survey), and joint- faulting. The number of joints that faulted greater than 1.0 mm was taken as the performance

326 measure for faulting. ANOVA method could not be applied to transverse and longitudinal cracking as the assumption of constant variance of residuals was violated owing to occurrence of cracking in not more than 30% of test sections. Wherever required, a natural logarithmic transformation of the variable of interest was done to satisfy the assumptions of ANOVA. Traffic volume and age of test sections were considered as covariates in all analyses to adjust for the difference in traffic loading and age among the sites in the experiment. Analysis of Response (FWD) Measures The response measures that were analyzed to determine the effects of design and site factors are as follows: • Deflection under FWD load (d0): This deflection corresponds to the structural strength of the entire pavement structure. • Deflection at the farthest sensor from FWD load (d6): This deflection gives an idea about the strength of the subgrade soil. • Effective Stiffness (ES) of the PCC slab. • Area Factor (AF) of the PCC slab of the pavement. Each of the above measures was derived or calculated from the midslab FWD testing (J1). FWD testing is conducted at 10 slabs on each section during a typical survey. For a section, an average of the deflections from all the tests corresponding to a survey was used for obtaining the above response measures. Effective Stiffness (ES) of PCC slab and Area Factor (AF) were calculated based on the study by Stubstad [14]. Higher the ES or AF, stiffer the upper layer of the pavement. Analysis was conducted on the above measures corresponding to first survey and to last survey, separately. It was assumed that the measures corresponding to first survey of the sections, gives an idea about the contribution of the design and site factors to the as-built structural condition of the test sections. Also, it was assumed that analysis of the response measures corresponding to the latest (or final) survey provides information about the contribution of the design and site factors to the “long-term” performance of the test sections. It is known that the temperature of the PCC slab at the time of testing has considerable bearing on the deflections of the PCC slab. For this, in all the analyses on response parameters, surface and bottom temperatures at the time of testing were taken as covariates.

327 RESULTS FROM ANALYSES The following is a summary of the main findings from each method of analysis, categorized by performance measure and response indicator. Basic statistics pertaining to the extent of occurrence of distresses have been presented along with results to corroborate the results with data. It is suggested that the results be interpreted keeping in view the extent of distresses (see Chapter 3) that occurred in the test sections. In the discussion of results, the word ‘significance’ needs to be interpreted as statistical significance, unless specified as practical significance. An asterisk in the results indicates both practical and statistical significance of an effect. As mentioned before, all analyses were conducted without including data from the site in Nevada (32), as extensive distresses at the site are related to wide range of construction issues that occurred at the site but not to pavement performance. Inclusion of data from this site will affect results from analyses, significantly. 6.6.1 Extent of Distresses by Experimental Factor This section contains a discussion on the effect of key experimental factors as in the case of the SPS-1 experiment. As stated earlier in Chapter 3, the occurrence of cracking and faulting in the SPS-2 sections was “low”. Hence, only roughness (latest or final roughness) of the test sections is presented to illustrate the effects of experimental factors on roughness (see Figure 6- 8). Figure 6-8 indicates that about 60% of all test sections have shown IRI value higher than 1 m/km, with about 20% of all test sections showing IRI value higher than 1.4 m/km. The effect of specific design and site factors is discussed below. The following is a summary of inferences from analysis of roughness: a) Drainage: The effect of drainage, in terms of higher percentage of test sections showing roughness, is more pronounced at the higher levels of roughness. This could mean that drainage is effective in reducing growth of roughness [see Figure 6-8 (a)]. b) Base Type: The difference in the percentage of test sections that have roughness, between those built on DGAB and those built on PATB are highest among all experimental factors (about 20%). Sections built on DGAB bases showed the highest percentages, while those built on LCB and those built on PATB showed comparable values [see Figure 6-8 (a)].

328 c) PCC Slab Thickness: The percentage of test sections with 203 mm (8-inch) PCC slab that have IRI of at least 1.5 m/km is about 30% as compared to about 50% for test sections with 279 mm (11-inch) PCC slab. The difference in percentage of test sections with higher roughness levels for sections with different thickness is negligible [see Figure 6-8 (b)]. d) Flexural Strength: The percentage of test sections with higher strength concrete (6.2 MPa) that have IRI of at least 1.5 m/km is about 50% as compared to about 30% for test sections with lower strength concrete (3.8 MPa). This difference is lesser at higher levels of roughness [see Figure 6-8 (c)]. e) Lane Width: Consistently, more sections with standard lane width 3.7 m (12-feet) have exceeded (slightly) various IRI levels than sections with widened lane 4.3 m (14-feet) [see Figure 6-8 (d)]. f) Climatic Zone: The effect of climate on roughness appears to be significant; with about 5% to 10% more sections in WF zone exceeding 1.5 m/km than sections in DF zone [see Figure 6-8 (e)]. g) Subgrade Soil Type: Consistently sections built on fine-grained soils have exceeded (slightly) various IRI levels than sections built on coarse-grained soils [see Figure 6-8 (f)].

329 0 10 20 30 40 50 60 70 80 90 100 1 1.25 1.5 1.75 2 IRI, m/km P e r c e n t o f s e c t i o n s a b o v e DGAB LCB PATB (a) Base type 0 10 20 30 40 50 60 70 80 90 100 1 1.25 1.5 1.75 2 IRI m/km P e r c e n t o f s e c t i o n s a b o v e . 203 mm 279 mm (b) PCC slab thickness 0 10 20 30 40 50 60 70 80 90 100 1 1.25 1.5 1.75 2 IRI m/km P e r c e n t o f s e c t i o n s a b o v e . 3.8 MPa 6.2 MPa (c) PCC flexural strength 0 10 20 30 40 50 60 70 80 90 100 1 1.25 1.5 1.75 2 IRI m/km P e r c e n t o f s e c t i o n s a b o v e . 3.7 m 4.3 m (d) Lane Width 0 10 20 30 40 50 60 70 80 90 100 1 1.25 1.5 1.75 2 IRI m/km P e r c e n t o f s e c t i o n s a b o v e . WF WNF DF DNF (e) Climatic zone 0 10 20 30 40 50 60 70 80 90 100 1 1.25 1.5 1.75 2 IRI m/km P e r c e n t o f s e c t i o n s a b o v e . Coarse-grained Fine-grained (f) Subgrade Figure 6 - 8 Effect of experimental factors on roughness

330 6.6.2 Frequency-based Methods Two frequency-based methods were used- Linear Discriminant Analysis and Binary Logistic Regression (details in Chapter 4). The results from these analyses are as follows: Linear Discriminant Analysis Based on this method all the distresses were analyzed using the following thresholds for categorization of the sections. • Transverse or longitudinal cracking: Cracked versus non-cracked • Wheel path joint faulting: Faulting <1.0 mm versus Faulting> 1.0 mm • Roughness: IRI (final)< 1.5 m/km versus IRI (final)>1.5 m/km IRI (initial)<1.25 m/km versus IRI (initial)> 1.25 m/km. This analysis is intended to identify the experimental factors that best discriminate the distresses versus non-distressed pavement sections. As the pavements in the SPS-2 experiment have not shown a “high” level of distress, this analysis will help in finding the significant design and site factors contributing to the occurrence of distresses (rather than magnitude), at this point in time. Traffic and pavement age, are considered as covariates in this analysis. Transverse cracking The design factors drainage, base type, and target PCC thickness were significant in discriminating between cracked or un-cracked sections. Table 6- 18 summarizes the effect of the design and site factors on the occurrence of transverse cracking, in general. In the WF zone, the effects of PCC thickness (p=0.041) and subgrade soil type (p= 0.007) were statistically significant in discriminating between sections with cracking and without cracking. Table 6- 19 summarizes effects of experimental factors based on the results of LDA on transverse cracking for sections in WF zone. It was observed that 33% of sections with 203 mm slab have exhibited transverse cracking while 14% of sections with 279 mm slab have exhibited cracking. Moreover, 14% of sections with thinner (203 mm) slab have exhibited high severity cracking where as none of the thicker (279 mm) slab sections exhibited high severity cracking. While 33% of sections built on fine-grained soils manifested transverse cracking, 4% of sections built on coarse-grained exhibited cracking. Also, about 22% (13 of the 60 sections) of sections built on fine-grained subgrade soils have exhibited high severity cracking while 4% (1 of the 24 sections) of the sections built on coarse-grained soils exhibited cracking.

331 Table 6- 18 Summary of results from LDA on transverse cracking- Overall Factor Category Factor Effects on transverse cracking p-value Drainage Presence of drainage significantly reduces the chances of occurrence of cracking Yes (0.001) Target PCC thickness Thicker (279 mm) PCC thickness reduces the chances of occurrence of cracking Yes (0.001) Base type The type of base significantly impacts the chances of the occurrence of cracking Yes (0.044) Flexural Strength No significant effect. In general, the 6.2 MPa mixes tend to mitigate cracking. No (0.716) Design Lane Width No significant effect. 4.3 m wide lane sections tend to inhibit cracking. No (0.467) Climatic Zone No significant effect. Designs constructed in Dry zones tend to crack more. No (0.147) Site Subgrade soil type No significant effect, however the model indicates that sections on fine subgrade soils tend to crack more than sections on coarse subgrade soils No (0.538) Table 6- 19 Results from LDA on transverse cracking, WF zone Factor Effects on transverse cracking p-value Drainage No significant effect. Presence of drainage reduces the chances of occurrence of cracking No (0.151) Target PCC thickness Sections with thicker (279 mm) PCC slabs crack significantly less Yes (0.041) Base type No significant effect. Sections on LCB tend to crack more. No (0.214) Flexural Strength No significant effect. In general, the 6.2 MPa mixes tend to crack more No (1.000) Lane Width No significant effect. 4.3 m wide lane sections tend to inhibit cracking No (0.614) Subgrade soil type Sections on fine subgrade soils tend to crack significantly more than sections on coarse subgrade soils Yes (0.007)

332 Longitudinal cracking The effect of target PCC thickness, base type and the climatic zone are significant in discriminating between cracked and un-cracked sections. Table 6-20 summarizes the effect of the design and site factors on the occurrence of longitudinal cracking. The effects of PCC thickness and base type, in WF zone, were statistically significant in discriminating between sections with cracking and sections with no cracking. Table 6- 21 summarizes the effects of experimental factors based on the results from LDA on longitudinal cracking, for pavements in WF zone. While 19% of sections with 203 mm (8-inch) PCC slab have exhibited longitudinal cracking, 10% of sections with 279 mm (11-inch) PCC slab have exhibited cracking. Also, 7%, 25%, and 11% of sections built on DGAB, LCB and PATB, respectively, have exhibited longitudinal cracking.

333 Table 6- 20 Summary of results from LDA on longitudinal cracking Factor category Factor Effects on longitudinal cracking p-value Drainage No significant effect. Presence of drainage increases the chances of occurrence of cracking. No (0.180) Target PCC thickness Thicker (279 mm) PCC thickness reduces the chances of occurrence of cracking Yes (0.000) Base type The type of base significantly impacts the chances of the occurrence of cracking Yes (0.004) Flexural Strength No significant effect. In general, the 6.2 MPa mixes tend to mitigate cracking. No (0.834) Design Lane Width No significant effect. 4.3 m wide lane sections tend to have more cracking. No (0.834) Climatic Zone Designs constructed in Dry zones tend to crack more. Yes (0.009) Site Subgrade soil type No significant effect, however the model indicates that sections on fine subgrade soils tend to crack more than sections on coarse subgrade soils No (0.456) Table 6- 21 Results from LDA on Longitudinal cracking, in WF zone Factor Effects on longitudinal cracking p-value Drainage No significant effect. Sections with drainage have cracked more than the ones without drainage No (0.193) Target PCC thickness Thicker (279 mm) PCC thickness significantly reduces the chances of occurrence of cracking Yes (0.026) Base type Sections on LCB crack significantly more than other sections Yes (0.023) Flexural Strength No significant effect. In general, the 6.2 MPa 900-psi mixes tend to crack more. No (1.000) Lane Width No significant effect. 4.3 m wide lane sections tend to inhibit cracking. No (0.463) Subgrade soil type No significant effect, the model indicates that sections on fine subgrade soils tend to crack more than sections on coarse subgrade soils No (0.296)

334 Faulting None of the design or site factors are discriminating between sections with faulting and without faulting, at this point in time. Climate appears to have some effect (p-value= 0.098), in that; the pavements located in Wet zones have higher faulting than those located in Dry zones (see Table 6-22). Analysis was also conducted combining data from sections located in the WF zone and none of the factors were found to be significantly affecting the occurrence of faulting. Roughness The initial and current roughness of the test sections were analyzed by categorizing the variables using the thresholds mentioned before. Table 6-23 and Table 6-24 are results from these analyses. PCC thickness was the only factor that was found to be discriminating between “smooth” and “rough” pavement sections based on the initial IRI categories. Based on LDA on current roughness, drainage, PCC thickness, and base type were found to be the significant factors.

335 Table 6- 22 Summary of LDA on Faulting Factor category Factor Effects on faulting p-value Drainage No significant effect. Presence of drainage decreases the chances of faulting. No (0.202) Target PCC thickness No significant effect. Thicker (279 mm) PCC thickness increases the chances of occurrence of pumping. No (0.623) Base type No significant effect. Lesser faulting occurs in LCB. No (0.315) Flexural Strength No significant effect. In general, the 6.2 MPa mixes tend to have more faulting. No (0.251) Design Lane Width No significant effect. 3.7 m wide lanes tend to have more faulting. No (0.412) Climatic Zone No significant effect. Designs constructed in Wet zones tend to have more faulting. No (0.098) Site Subgrade soil type No significant effect, however the model indicates that sections on fine subgrade soils tend to fault lesser than sections on coarse subgrade soils No (0.846) Table 6- 23 Summary of results from LDA on initial roughness Factor category Factor Effects on initial roughness p-value Drainage* Presence of drainage decreases the chances of higher initial roughness. No (0.090) Target PCC thickness Sections with 279 mm PCC slab have higher chances of being built rougher than the 203 mm ones. Yes (0.054) Base type No significant effect. Lesser roughness was observed on sections built with PATB. No (0.329) Flexural Strength No significant effect. In general, the 6.2 MPa mixes tend to have more roughness. No (0.201) Design Lane Width No significant effect. 4.3 m wide lanes tend to have more roughness. No (0.750) Climatic Zone No significant effect. Designs constructed in Wet zones tend to have more roughness. No (0.232) Site Subgrade soil type No significant effect Sections on fine subgrade soils tend to have more initial roughness. No (0.342) * The effect of drainage is based on comparison between sections built on PATB and sections built on DGAB.

336 Table 6- 24 Summary of results from LDA on initial roughness, in WF zone Factor Effects on initial roughness p-value Drainage* No significant effect. Presence of drainage decreases the chances of higher initial roughness. No (0.356) Target PCC thickness No significant effect. Sections with 279 mm PCC slab have higher chances of being built rougher than 203 mm ones. No (0.835) Base type No significant effect. Sections with PATB tend to be built smoother. No (0.0.594) Flexural Strength No significant effect. In general, the 6.2 MPa mixes tend to have more roughness. No (0.190) Lane Width No significant effect. 3.7 m wide lanes tend to have more roughness. No (0.190) Subgrade soil type No significant effect. Sections on fine subgrade soils tend to have more initial roughness. No (0.071) *The effect of drainage is based on comparison between sections built on PATB and sections built on DGAB. Table 6- 25 Summary of results from LDA on roughness Factor category Factor Effects on final (latest) roughness p-value Drainage Presence of drainage inhibits increase in roughness. Yes (0.000) Target PCC thickness Thicker (279 mm) PCC thickness increases the chances of higher roughness. Yes (0.036) Base type Sections on DGAB and LCB have higher increase in roughness. Yes (0.017) Flexural Strength No significant effect. In general, the 6.2 MPa mixes tend to cause more roughness. No (0.076) Design Lane Width No significant effect. 12-foot wide lanes tend to have lesser roughness. No (0.873) Climatic Zone No significant effect. Designs constructed in Wet zones tend to have more roughness. No (0.588) Site Subgrade soil type No significant effect. Sections on fine subgrade soils have more roughness. No (0.317) Table 6- 26 Summary of results from LDA on roughness, in WF zone Factor Effects on final (latest) roughness p-value Drainage Presence of drainage inhibits increase in roughness. Yes (0.008) Target PCC thickness No significant effect. Thicker (279 mm) PCC thickness decreases the chances of higher roughness. No (1.000) Base type Sections on DGAB have higher increase in roughness. Yes (0.032) Flexural Strength No significant effect. In general, 6.2 MPa mixes tend to cause more roughness. No (0.081) Lane Width No significant effect. 3.7 m wide lanes tend to have more roughness. No (0.193) Subgrade soil type Sections on fine subgrade soils have more roughness. Yes (0.004)

337 Binary Logistic Regression (BLR) The BLR model was used to model the probability of occurrence for the various performance measures. Thresholds similar to the ones used for LDA were used to categorize the test sections for this analysis. The results are summarized in Table 6-27 and Table 6-28. Transverse Cracking The BLR model for transverse cracking was significant with a p-value of 0.000. Moreover, 88.5% of the times, the model correctly differentiate cracked sections from non- cracked sections. Based on this analysis, the effects of significant factors are as follows: PCC slab thickness—Sections built with 203 mm (8-inch) PCC slab have significantly higher probability of cracking than the ones built with 279 mm (11-inch) PCC slab. Base Type—Sections built on PATB have significantly higher likelihood of cracking than those built on LCB. Subgrade—Sections built on fine subgrade soils have significantly higher probability of cracking than the ones built on coarse subgrade soils. Based on the BLR on data from sections in WF zone, the effect subgrade soil type (0.029 in BLR) was statistically significant in discriminating between sections with cracking and sections without cracking. Longitudinal Cracking The BLR model for longitudinal cracking was significant with a p-value of 0.000. Moreover, 88.5% of the times, the model correctly differentiates cracked sections from un- cracked sections. Based on this analysis, the following conclusions can be made. Base Type—Sections on LCB have significantly higher chances of cracking compared to the DGAB sections. PCC slab thickness—Sections with 203 mm thick slab have significantly higher chances of cracking than the Sections with 279 mm thick slab. Climatic Zone—Sections in Dry No Freeze have significantly higher chances of cracking than Wet Freeze.

338 From BLR on data from sections in WF zone, it was found that PCC slab thickness has a slight effect (p = 0.084) on cracking. Faulting The BLR model for faulting was significant with a p-value of 0.010. Moreover, 69.9% of the times, the model correctly differentiates sections with faulting from sections without faulting. From the analysis the following effects were found to be significant: Subgrade type—Sections on coarse-grained soils have significantly higher chances of faulting than the sections that are built on fine-grained soils. Climatic Zone— The chances of occurrence of faulting are slightly higher for the non-drained sections. Also, the chances of occurrence of faulting in Wet freeze climatic zone are higher than those in Dry No Freeze zone.

339 Table 6- 27 Summary of p-values from BLR for determining the effect of experimental factors on pavement performance measures- Overall Roughness Experimental Factors Transverse cracking Longitudinal cracking Faulting Initial Current Drainage 0.083 (3.4) 0.616 (0.68) 0.06 (2.3) 0.28 (1.6) 0.003 (7.5) Base type 0.073 (4.9)* 0.099 (2.7) 0.17 (1.4) 0.12 (2.4) 0.007 PCC thickness 0.019 (3.6) 0.001 (8.7) 0.955 (0.98) 0.018 (0.49) 0.381 (0.67) Flexural Strength 0.550 (1.4) 0.825 (1.13) 0.24 (0.65) 0.17 (0.63) 0.077 (0.44) Lane Width 0.389 (1.6) 0.800 (1.15) 0.381 (1.37) 0.687 (0.87) 0.876 (0.931) Subgrade type 0.003 (9.5) 0.361 (1.98) 0.046 (0.38) 0.186 (1.73) 0.283 (1.97) Climatic Zone 0.571 0.017 (18) 0.104 0.400 0.262 Note: Values in parenthesis are odds ratios. *LCB vs. PATB. Table 6- 28 Summary of p-values from BLR for determining the effect of experimental factors on pavement performance measures- WF Zone Roughness Experimental Factors Transverse cracking Longitudinal cracking Faulting Initial Current Drainage 0.736 (0.74) 0.55 (0.35) 0.140 (2.34) 0.510 (1.5) 0.004 (47.4) Base type 0.767 0.237 0.32 0.383 0.014 PCC thickness 0.650 (1.43) 0.084 (11.3) 0.763 (1.15) 0.058 (.039) 0.066 (5.05) Flexural Strength 0.643 (0.70) 0.868 (0.84) 0.723 (0.85) 0.141 (0.48) 0.222 (0.37) Lane Width 0.593 (1.49) 0.234 (4.65) 0.337 (1.57) 0.147 (2.1) 0.655 (1.39) Subgrade type 0.029 (19.5) 0.488 (0.23) 0.636 (0.70) 0.066 (2.97) 0.368 (3.009) Note: Values in parenthesis are odds ratios.

340 Roughness The BLR model for initial roughness was significant with a p-value of 0.023. Moreover, 61.5% of the times, the model correctly differentiates sections with “poor” roughness from other sections. Based on this analysis, base type and PCC slab thickness are significant factors that discriminate between the categories. The effects of the factors are: Base Type—Sections built on LCB have significantly (p-value=0.040) higher probability of roughness than the ones built on PATB. PCC slab thickness—Sections built with thinner slab (203 mm) have significantly (p- value=0.038) lesser probability of being built rougher than the ones built with thicker slab (279 mm). The BLR model for latest roughness was significant with a p-value of 0.000. Moreover, 78.8% of the times, the model correctly differentiate sections with “poor” roughness from other sections. Based on this analysis, only base type has a significant effect on the categories and the effect is described below. Base Type—Sections built on DGAB or LCB have significantly (p-value=0.007) higher probability of roughness than the ones built on PATB. This may also be interpreted that sections with drainage have significantly lesser chances of becoming rougher compared to sections without drainage (sections on DGAB).

341 6.6.3 Analysis of Variance ANOVA was performed on roughness and faulting of sections in the SPS-2 experiment. The procedure adopted for this analysis is the same as that for analysis of SPS-1 data. As mentioned before, the analyses were performed combining data from all sections (overall) in the experiment. Also ANOVA was conducted on data for sections within each zone. It should be noted that analysis within zone was presented only for the WF zone. Each of the other zones has two sites each and this sample size did not yield meaningful results because of less statistical “power”. The effects of design factors and site factors on pavement performance are discussed next leading to a discussion on results from analysis of pavement response. Effects of design factors on pavement performance The results from this analysis are summarized in Table 6-29 and these results indicate that the most significant design factor is the base type, which has a significant effect on ∆IRI and IRIo. In addition to the effect of base type, ∆IRI is affected by drainage and PCC thickness (slight effect), IRIo is affected by PCC slab thickness, and faulting is affected by lane width. For investigating the practical (operational) significance of the mean difference between the levels of design factors, the marginal means (predicted cell means from the model) were transformed back to the original scale of the distress, as discussed in Chapter 5. These conversions were necessary in order to find out the practical/operational mean difference. Table 6- 30 shows the back-transformed marginal means for all levels of design factors. The following discussion summarizes significant effects of design factors on pavement performance: • Effect of drainage: Pavement sections with drainage have shown significantly lower change in roughness than those without drainage. This effect was found to be practically significant (>0.10 m/km per year). • Effect of base type: Pavement sections with DGAB have shown the highest change in roughness while those with PATB have shown the least change in roughness. This difference in change in roughness between sections with DGAB and sections with PATB is practically significant (>0.10 m/km per year). Sections built with LCB had the highest initial roughness while other sections had comparable initial roughness.

342 • Effect of PCC thickness: Significantly (practically and statistically) higher initial roughness was observed on sections with 279 mm (11”) PCC slab compared to sections with 203 mm (8”) PCC slab. However, the change in roughness was slightly higher in sections with 203 mm PCC slab compared to those with 279 mm PCC slab. • Effect of flexural strength: No significant effect of PCC flexural strength was found on roughness and faulting, at this point. • Effect of lane width: Pavement sections with wider lane (4.3 m) have shown significantly lower faulting than those with standard lane (3.7 m). No significant effect of lane width was found on roughness.

343 Table 6- 29 Summary of p-values from ANOVA for determining the effects of design factors on pavement performance−Overall Roughness (IRI) Factor Faulting ∆IRI IRIo Drainage 0.477 0.002* 0.373 PCC thickness 0.243 0.094* 0.003 Base Type 0.585 0.009* 0.037 PCC Flexural Strength 0.903 0.544 0.246 Lane Width 0.050 0.860 0.313 Site (blocked) 0.000 0.006 0.000 N=110 R2=0.411 N=114 R2=0.305 N=156 R2=0.344 * Also shows operational/practical significance Table 6- 30 Summary of marginal means from ANOVA for determining the effect of main design factors on pavement performance measures—Overall IRI Design Factor ∆IRI (m/km) IRIo (m/km) No 0.34 1.33 Drainage Yes 0.14 1.29 203 mm 0.32 1.25 PCC thickness 279 mm 0.22 1.36 DGAB 0.39 1.29 LCB 0.30 1.37 Base Type PATB 0.16 1.25 3.8 MPa 0.29 1.28 PCC Flexural Strength 6.2 MPa 0.24 1.32 3.7 m 0.26 1.32 Lane Width 4.3 m 0.27 1.28 MSE 1.343 0.032

344 Similar ANOVA was performed on data from sections in the WF zone and Table 6- 31 is the summary of results from the analysis. Table 6- 32 shows the back transformed marginal means for all levels of design factors. As mentioned before, the results from analysis on data from other climatic zones are not presented, as limited amount of data is available for those zones. The following discussion summarizes significant effects of design factors on performance of sections in the WF zone: • Effect of drainage: Pavement sections with drainage have shown significantly lower change in roughness than those without drainage. This effect was found to be practically significant (>0.10 m/km per year). • Effect of base type: Pavement sections with LCB have shown significantly higher change in roughness compared to those with PATB and with DGAB. This difference in change in roughness between sections with LCB and sections with PATB is practically significant (>0.10 m/km per year). Sections with PATB have shown the least change in roughness. Sections built with LCB were found to have the higher initial roughness compared to other sections, which have comparable initial roughness. • Effect of PCC thickness: Significantly (practically and statistically) higher initial roughness was observed on sections with 279 mm PCC slab compared to sections with 203 mm PCC slab. However, the change in roughness was significantly higher in sections with 203 mm PCC slab compared to those with 279 mm PCC slab. • Effect of flexural strength: Significant effect of PCC flexural strength was found on initial roughness. Sections with higher strength concrete (6.2 MPa) had higher initial roughness compared to those with lower strength concrete (3.8 MPa). • Effect of lane width: Slight effect of lane width was observed for initial roughness in that sections with standard lane (3.7 m) width were constructed with higher initial roughness.

345 Table 6- 31 Summary of p-values from ANOVA for determining the effects of design factors on pavement performance−WF Zone Roughness (IRI) Factor Faulting ∆IRI IRIi Drainage 0.420 0.012* 0.688 PCC thickness 0.668 0.033* 0.057 Base Type 0.701 0.004* 0.557 PCC Flexural Strength 0.907 0.650 0.009 Lane Width 0.127 0.987 0.071 Site (blocked) 0.102 0.000 0.000 N=52 R2=0.303 N=58 R2=0.505 N=84 R2=0.426 * Also shows operational/practical significance Table 6- 32 Summary of marginal means from ANOVA for determining the effect of main design factors on pavement performance measures—WF Zone IRI Design Factor ∆IRI (m/km) IRIo (m/km) No 0.30 1.32 Drainage Yes 0.13 1.29 203 mm 0.29 1.26 PCC thickness 279 mm 0.16 1.36 DGAB 0.26 1.30 LCB 0.34 1.34 Base Type PATB 0.11 1.27 3.8 MPa 0.20 1.24 PCC Flexural Strength 6.2 MPa 0.23 1.38 3.7 m 0.22 1.35 Lane Width 4.3 m 0.22 1.26 MSE 0.947 0.033

346 Effect of Site Factors on Pavement Performance Given the unbalanced nature of the experimental design with respect to climatic zone, a one-way ANOVA was performed study the main effects of the subgrade soil type (fine-grained versus coarse-grained soils) and climatic zone (wet versus dry, freeze versus no-freeze), one at a time. The p-values and mean performances from these analyses are summarized in Table 6- 33 and Table 6- 34, respectively. To indicate the direction of effects, “+” and “-“ signs are reported along with p-values. A “+” indicates that the first level has exhibited more distress than second level within a factor, while a “-“ indicates otherwise. For example, in the case of the effect of subgrade on faulting, “+” indicates higher faulting in pavements constructed on fine-grained soils compared to those constructed on coarse-grained soils, as fine-grained soil is the first level and coarse-grained soil is the second. The p-values indicate that subgrade type appears to be significant in affecting the initial roughness. Pavements built on fine-grained subgrade showed higher initial roughness than those constructed on coarse-grained subgrade. Climate has a significant effect on faulting and initial roughness. Pavements located in “wet” climate have exhibited significantly higher faulting than those located in “dry” climate. A slight effect of climate was found within “dry” zones. Pavements located in DF zone exhibited higher faulting than those located in DNF zone. Pavements constructed in “wet” climate were found to have slightly higher initial roughness compared to those in “dry” climate.

347 Table 6- 33 Summary of p-values from one-way ANOVA for determining the effect of site factors on pavement performance measures Roughness (IRI) Site Factor Faulting ∆IRI IRIi Subgrade Fine-grained vs. Coarse-grained 0.184 (-) 0.920 (+) 0.050 (+) Climate Wet vs. Dry Freeze vs. No Freeze Wet Freeze vs. Wet No Freeze Dry Freeze vs. Dry No Freeze 0.006 (+) 0.924 (-) 0.163 (-) 0.092 (+) 0.406 (-) 0.632 (-) 1.000 (+) 1.000 (-) 0.052 (+) 0.317 (-) 1.000 (-) 1.000 (-) Table 6- 34 Summary of marginal means from one-way ANOVA for determining the effect of site factors on pavement performance measures Roughness (IRI) Site Factor Faulting ∆IRI IRIi Fine-grained 1.24 0.37 1.33 Subgrade Coarse-grained 1.64 0.38 1.25 Wet 1.68 0.35 1.33 Dry 0.94 0.42 1.25 Freeze 1.39 0.36 1.29 No Freeze 1.41 0.41 1.33 Wet Freeze 1.34 0.36 1.31 Wet No Freeze 2.42 0.33 1.38 Dry Freeze 1.18 0.38 1.21 Climate Dry No Freeze 0.52 0.50 1.29

348 Effect of Design Factors on Pavement Performance based on standard deviates As explained before in Chapter 4, the experiment design and the performance of the test sections have rendered the SPS-2 experiment unbalanced (statistical). Most of the sites, 9 of the 14, in the experiment were constructed in “wet” climate of which 7 are in the WF zone. Also, all 24 unique designs were not built in every soil-climate combination. Furthermore, non-occurrence of distresses in a considerable number of sections has also contributed to the unbalance. In light of these concerns, a simplified analysis considering one design factor at a time (univariate) was performed, as in the case of analysis of the effects of site factors. The performance of similar designs was not found to be consistent across sites indicating a considerable influence of the site conditions at each site. The site conditions that could have contributed to this variation in performance are traffic, age, construction quality, measurement variability, and material properties, apart from site factors (i.e. subgrade and environment). In order to separate the “true” effects of the experimental factors, this “noise” should be nullified. The concept of standard deviate, as in the case of the SPS-1 experiment, was thus employed for the analyses on roughness and faulting (see Chapter 5). This measure (standard deviate) indicates the relative performance of a design with respect to the other designs at the same site. As this measure was calculated for each section with respect to other sections in a site, it indicates the relative standing of the section compared to other sections in that site. It thus helps nullify the variation in performance (due to site conditions) across sites, as the sections are evaluated with respect to companion sections in each site. This approach of using the standard deviate is similar to blocking of the site factors performed in the multivariate analysis. One-way analysis of variance (univariate) was performed on the standard deviates of the sections to study the effects of the design factors, by considering one design factor at a time. As the SPS-2 experiment design calls for study of the effects of design factors at different site conditions, the univariate analysis was performed accordingly. The analyses were performed on data from all sections and also on subsets of data corresponding to different subgrade types, climates and combination of these. This helped in identification of the effects of design factors under different site conditions. To study the “pure” effect of each design factor, comparisons of standard deviates were also calculated by controlling for the other factors, as in the case of level-B analyses (site-level).

349 As mentioned earlier, these comparisons were performed only on roughness, and faulting. The method is not appropriate for cracking because of “low” occurrence of the distresses at this point in time. The effects of design factors, based on the above-mentioned analyses, on roughness and faulting are discussed next. Roughness The effects of the design and site factors, in terms of standard deviate, are shown in Figure 6-9. The change in roughness was considered as the performance measure for analyses. The summary of p-values corresponding to these analyses is Table 6- 35. The mean change in roughness corresponding to each comparison presented in Table 6- 35 is shown in Table 6- 36. Though the univariate analyses were performed on standard deviates, these means were used to identify operational significance of the effects. The effects of the design factors on change in roughness, based on this analysis, are presented below: Drainage: On the whole, the effect of drainage is statistically and operationally significant. Un- drained sections have exhibited higher change in roughness than drained sections. This effect is consistent in all sections, irrespective of subgrade soil type. In addition, the effect is more prominent in “wet” climate. Among the sections located in WF zone, those built on fine-grained soils have shown greater change in roughness compared to those built on coarse-grained soils. This effect is marginally significant (statistical) but is of practical significance. Base Type: As the effect of base type is confounded with the effect of drainage because of the SPS-2 experiment design, the results presented here are from comparisons between the performance of sections built with DGAB and sections built with LCB, which are both un- drained. On average, sections with LCB have shown lower change in roughness than sections with DGAB. This effect is not statistically significant. A significant (statistical and operational) effect of base type was observed among sections located in WNF zone, in that sections with DGAB have shown higher change in roughness than those with LCB. However, this effect should be interpreted with caution, as only two sites are located in WNF zone.

350 PCC slab thickness: The effect of PCC slab thickness is significant (statistical and operational) among sections built on fine-grained soils, especially when constructed in WF zone. Sections with 203 mm slab have shown higher change in roughness than those with 279 mm slab. PCC flexural strength: The effect of PCC flexural strength was not found to be significantly affecting roughness of the SPS-2 sections, at this point in time. Lane width: The effect of lane width was not found to be significantly affecting roughness of the SPS-2 sections. On average, among sections built on fine-grained soils and located in WF zone, those with standard lane (3.7 m) have shown higher change in roughness than those with wider lane (4.3 m).

351 -0.60 -0.40 -0.20 0.00 0.20 0.40 0.60 N D D D G A B LC B PA TB 20 3 m m 27 9 m m 3. 8 M Pa 6. 2 M Pa 3. 7 m 4. 3 m Design Fators M ea n st d. D ev ia t (a) Overall -0.60 -0.40 -0.20 0.00 0.20 0.40 0.60 N D D D G A B LC B PA TB 20 3 m m 27 9 m m 3. 8 M Pa 6. 2 M Pa 3. 7 m 4. 3 m Design Factors M ea n St d. D ev ia t Coarse Fine (b) By subgrade type -0.60 -0.40 -0.20 0.00 0.20 0.40 0.60 N D D D G A B LC B PA TB 20 3 m m 27 9 m m 3. 8 M Pa 6. 2 M Pa 3. 7 m 4. 3 m Design Factors M ea n St d. D ev ia t (c) Wet Freeze (WF) zone Figure 6 - 9 Effect of design factors on change in IRI

352 Table 6- 35 Summary of p-values for comparisons of standard deviates— Change in roughness By subgrade By climatic zone By subgrade and zone WF WNF DF DNF Design Factor Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C Drainage Drainage vs. No-Drainage 0.000 0.009 0.002 0.052 0.024 0.004 0.084 0.071 0.497 0.297 0.048 0.051 0.084 Base type DGAB vs. LCB vs. PATB/DGAB 0.001 0.024 0.016 0.102 0.003 0.003 0.123 0.094 0.852 0.082 0.017 0.055 0.123 PCC thickness 203 mm vs. 279 mm 0.240 0.029 0.469 0.067 0.626 0.773 0.230 0.015 0.718 0.565 0.949 0.923 0.230 PCC flexural Strength 3.8 MPa vs. 6.2 MPa 0.962 0.530 0.506 0.442 0.226 0.832 0.603 0.797 0.065 0.255 0.639 0.941 0.603 Lane Width 3.7 m vs. 4.3 m 0.500 0.770 0.639 0.530 0.776 0.992 0.667 0.491 0.934 0.790 0.975 0.772 0.667 Note: Shaded cells indicate statistical significance at 90% or higher level of confidence (p<0.1). Table 6- 36 Summary of mean change in roughness SG Zone WF WNF DF DNF Design Factor Comparison Overall F C WF WNF DF DNF F C F C F C C ND 0.3 0.3 0.3 0.3 0.3 0.1 0.4 0.5 0.1 0.1 0.5 0.1 0.1 0.4 Drainage D 0.1 0 0.1 0 0.1 -0.1 0.2 0 0 0.1 0.1 0 -0.2 0.2 DGAB 0.3 0.3 0.3 0.3 0.3 0.1 0.4 0.5 0.1 0.1 0.5 0.1 0.1 0.4 LCB 0.2 0.3 0.1 0.3 0.1 0.3 0.2 0.4 0.1 -0.1 0.2 0.3 0.3 0.2 Base type PATB 0.1 0 0.1 0 0.1 -0.1 0.2 0 0 0.1 0.1 0 -0.2 0.2 203 mm 0.2 0.3 0.2 0.3 0.2 0.1 0.2 0.4 0.1 0.1 0.3 0.1 0.1 0.2 PCC slab thickness 279 mm 0.2 0.2 0.2 0.2 0.1 0.1 0.3 0.2 0.1 0 0.3 0.1 0 0.3 3.8 MPa 0.2 0.2 0.2 0.2 0.1 0.1 0.2 0.3 0.1 0 0.3 0.1 0.1 0.2 PCC flexural strength 6.2 MPa 0.2 0.2 0.2 0.2 0.2 0.1 0.2 0.3 0 0.1 0.3 0.1 0 0.2 3.7 m 0.2 0.3 0.2 0.3 0.1 0.1 0.3 0.4 0.1 0.1 0.3 0.1 0.1 0.3 Lane width 4.3 m 0.2 0.2 0.2 0.2 0.2 0.1 0.2 0.2 0.1 0.1 0.3 0.1 0 0.2

353 Faulting The effects of the design and site factors on faulting, in terms of standard deviate, are shown in Figure 6-10. The number of joints that faulted more than 1.0 mm was considered as the performance measure for PI, and the standard deviate is based on PI. The summary of p-values corresponding to analyses on faulting is Table 6- 37. The mean PI corresponding to each comparison presented in Table 6- 37 is shown in Table 6- 38. Though the univariate analyses were performed on standard deviates, these mean PIs were used to identify operationally significant effects. The effects of the design factors on faulting, based on this analysis, are presented below. It is to be noted that the magnitude of faulting in most of the sections is “low” at this point in time (details in Chapter 3). Drainage: The effect of drainage is not conclusive. However, on average, un-drained sections have exhibited more faulting than drained sections. Base Type: As the effect of base type is confounded with the effect of drainage because of the SPS-2 experiment design, the results presented here are from comparisons between the performance of sections built with DGAB and sections built with LCB, which are both un- drained. On average, sections with LCB have shown lesser faulting than sections with DGAB. This effect is inconclusive. Also, among sections located in WF zone and built on fine-grained soils, this effect is more prominent. PCC slab thickness: On average, among pavements located in WF zone sections, those with 203 mm (8”) PCC slab have sown higher faulting than those with 279 mm (11”) slab. However, the effect of PCC slab thickness is inconclusive at this point. PCC flexural strength: The effect of PCC flexural strength is not significantly affecting faulting. However, on average, sections with 3.8 MPa (550-psi) concrete have shown slightly higher faulting than those with 6.2 MPa (900-psi) concrete. Lane width: The effect of PCC flexural strength was found to be significantly affecting faulting in the SPS-2 sections. On the whole, sections built with standard lane (3.7 m) have shown lesser faulting than those with wider lane (4.3 m). This effect is more prominent among sections located in WF zone and built on fine-grained soils. In general, higher faulting was observed in sections located in WNF zone compared to those located in other zones. However, it is to be noted that only two sites are located in WNF zone.

354 -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 N D D D G A B LC B PA TB 20 3 m m 27 9 m m 3. 8 M Pa 6. 2 M Pa 3. 7 m 4. 3 m Design Factors M ea n St d. D ev ia t (a) Overall -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 N D D D G A B LC B PA TB 20 3 m m 27 9 m m 3. 8 M Pa 6. 2 M Pa 3. 7 m 4. 3 m Design Factors M ea n St d. D ev ia t Coarse Fine (b) By subgrade type -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 N D D D G A B LC B PA TB 20 3 m m 27 9 m m 3. 8 M Pa 6. 2 M Pa 3. 7 m 4. 3 m Design Factors M ea n St d. D ev ia t (c) Wet Freeze (WF) zone Figure 6 - 10 Effect of design factors on wheelpath joint faulting

355 Table 6- 37 Summary of p-values for comparisons of standard deviates— Faulting By subgrade By climatic zone By subgrade and zone WF WNF DF DNF Design Factor Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C Drainage Drainage vs. No-Drainage 0.459 0.228 0.796 0.297 0.780 0.296 0.49 0.057 0.315 0.561 0.407 0.989 0.469 Base type DGAB vs. LCB vs. PATB/DGAB 0.307 0.277 0.677 0.451 0.169 0.528 0.719 0.134 0.588 0.516 0.159 0.979 0.719 PCC thickness 203 mm vs. 279 mm 0.974 0.646 0.545 0.197 0.282 0.394 0.750 0.942 0.009 0.429 0.525 0.889 0.750 PCC flexural Strength 3.8 MPa vs. 6.2 MPa 0.899 0.593 0.663 0.724 0.994 0.360 0.965 0.570 0.816 0.423 0.414 0.137 0.965 Lane Width 3.7 m vs. 4.3 m 0.009 0.023 0.085 0.045 0.448 0.130 0.276 0.133 0.190 0.589 0.597 0.075 0.276 Note: Shaded cells indicate statistical significance at 90% or higher level of confidence (p<0.1). Table 6- 38 Summary of mean PIs for faulting SG Zone WF WNF DF DNF Design Factor Comparison Overall F C WF WNF DF DNF F C F C F C C ND 1.5 1.3 1.8 1.3 4.4 0.3 0.2 1.6 0.7 0.8 8.1 0.4 0.1 0.2 Drainage D 1 0.8 1.2 0.7 2.7 0.9 0.2 0.4 1.4 2.3 3.7 0.5 2.3 0.2 DGAB 1.5 1.3 1.8 1.3 4.4 0.3 0.2 1.6 0.7 0.8 8.1 0.4 0.1 0.2 LCB 0.7 0.6 0.8 0.7 1 0.9 0.2 0.6 0.9 0.4 1.6 0.8 1.5 0.2 Base type PATB 1 0.8 1.2 0.7 2.7 0.9 0.2 0.4 1.4 2.3 3.7 0.5 2.3 0.2 203 mm 1 0.8 1.4 1.1 2.2 0.6 0.2 0.8 1.7 1.2 3.6 0.6 0 0.2 PCC slab thickness 279 mm 1.1 0.9 1.2 0.7 3.2 0.8 0.3 0.9 0.2 1.5 5.7 0.5 1.6 0.3 3.8 MPa 1.2 0.9 1.5 1.1 3 0.6 0.2 1.1 0.9 0.8 6 0.4 1.8 0.2 PCC flexural strength 6.2 MPa 0.9 0.8 1.1 0.7 2.4 0.8 0.2 0.6 1 1.8 3.3 0.8 0.9 0.2 3.7 m 1.3 1.2 1.5 1.3 3.1 0.9 0.3 1.2 1.4 1.8 5.7 0.7 1.2 0.3 Lane width 4.3 m 0.8 0.5 1.1 0.6 2.3 0.6 0.2 0.6 0.6 0.7 3.9 0.4 1.3 0.2

This page is intentionally left blank. 356

357 6.6.4 Effect of Experimental Factors on Pavement Response As mentioned earlier, the dependent variables that were considered for the ANOVA on pavement response include deflections do, d6, Area Factor (AF) and Effective Stiffness (ES), based on deflections from the J1 location (midslab). ANOVA was performed on initial values and final (or latest) values of these deflection parameters, separately. The computational details regarding the AF and the ES can be found in the report by Stubstad [9]. A natural logarithmic transformation was applied to the deflection parameters, for the data to be used for ANOVA without any violation of statistical assumptions. The surface and bottom temperatures of the PCC slab at the time of testing were included as covariates in ANOVA along with age of the sections at the time of testing and variability in the PCC slab thickness. The results (see Table 6- 39 and Table 6- 40) from ANOVA are discussed next. Peak Deflection under FWD load (do) When the effects of design features on do-initial were studied by blocking the site factors, the main effects of base type (p=0.000) and PCC thickness (p=0.000) were found to be significant. The pavements constructed on DGAB have shown significantly higher deflections than those constructed on PATB. The sections constructed on LCB have the least amount of deflections. Also the sections with 203 mm (8”) PCC slab deflected significantly more than 279 mm (11”) PCC slab. When effects of both the design and site factors do-initial were studied, it was found that interaction between subgrade soil type and climatic zone is significant (p =0.000). Among the sections located in WF zone, those built on fine-grained subgrade soils have significantly higher deflections than those built on coarse-grained subgrade soils. Based on analyses on final survey d0, the interaction effects between base type and lane width (p=0.014), base type and subgrade soil type (p=0.014), and climate and PCC thickness (p=0.020) were found significant. Among sections with LCB, those built with standard lane (3.7 m) [3.7m (12’)] have shown significantly higher d0 than those with widened lane [4.3 m (14’)]. Also among the sections with LCB, those built on fine-grained soils have shown significantly higher deflections than those built on coarse-grained soils. It is to be noted that, in general, sections with DGAB have significantly higher d0 values than sections with LCB. For sections

358 located in WF zone, those with 203 mm PCC slab have shown significantly higher deflections than those with 279 mm PCC slab thickness. These results imply that the design factors significantly interact over time to affect d0, unlike in the initial conditions where only the main effects of design factors were important. Far Sensor Deflection (d6) The effects of design factors on d6-initial were studied blocking the site factors. The main effects of base type (p=0.000) and PCC thickness (p=0.000) were found to be significant. The pavements constructed on DGAB have shown significantly higher deflections than those constructed on PATB or LCB. The sections constructed on LCB experienced the least amount of deflections. Also the sections with 203 mm PCC slab deflected more than those with 279 mm PCC slab. When the effects of both design and site factors were studied, the effects of site factors were not found to be important. The effect of drainage (p=0.002), PCC slab thickness (p=0.002), and base type (p=0.006) are significantly affecting d6 of final year for the test sections. The sections with PATB (with drainage) have significantly lesser d6 than sections with DGAB or LCB. The effect of PCC slab thickness remained the same as for initial year d6. It was also found that the main effect of subgrade is marginally significant (p =0.069) when effects of both the design and site factors were studied for final survey d6. Sections built with fine-grained soils have slightly higher d6 than those built on coarse-grained soils. Area Factor (AF) When the effects of design factors on AFinitial were studied by blocking the site factors, the main effect of PCC thickness (p=0.000) was found to be significant. The sections with 279 mm thick slab have higher AF than those with 203 mm thick slab. The main effect of climate (p=0.000) was significant when the effects of both the design and site factors were studied AFinitial. Sections located in “wet” climate have higher AF than those located in “dry” climate. An interaction between base type and subgrade (p=0.041) was also found significant, in that among sections with LCB, those constructed on coarse-grained subgrade have higher AF than those constructed on fine-grained subgrade soils. These effects were not found on final survey AF.

359 When the effects of design features on final survey AF were studied by blocking the site factors, the main effects of PCC slab thickness (p=0.001), and PCC flexural strength (p=0.024) were found to be significant. The effect of PCC slab thickness is similar to that for initial AF. Sections with higher strength concrete (target 14-day strength of 6.2 MPa (900-psi)) have significantly higher AF than those with lower strength concrete (target 14-day strength of 3.8 MPa (550-psi)). Effective Stiffness (ES) The effects of design features were studied by blocking the site factors. The interaction effects between base type and PCC thickness (p-value=0.053), and base type and flexural strength (p-value=0.012) were significant. The effect of PCC thickness on ES is more, among sections with DGAB or PATB than among sections with LCB. Also, the effect of PCC flexural strength on ES is more, in the case of sections built on DGAB or PATB than in the case of sections built on LCB. When effects of both the design and site factors were studied, it was found that the main effects of climatic zone (p=0.000) and subgrade (p=0.018) were significant. Pavements located in “wet” climate have significantly higher ES compared to those located in “dry” climate. Also, the upper layers of sections built on coarse-grained subgrade soil were significantly stiffer than those built on fine-grained soil. Drainage (p=0.05), PCC slab thickness (p=0.000), base type (p=0.016) and flexural strength (p=0.020) were found to have significant main effects when the effects of design features on final survey ES were studied by blocking the site factors (see Table 6- 39). The effects of PCC thickness and base type were similar as in the case of initial ES. However, sections built with drainage have higher ES than those without drainage. Also, sections with higher strength concrete have higher ES than those with lower strength concrete.

360 Table 6- 39 Summary of p-values for the effects of design factors on rigid pavement response do d6 Area Factor Effective stiffness of PCC slab Design Factor Initial Final Initial Final Initial Final Initial Final Drainage 0.000 0.000 0.002 0.002 0.352 0.922 0.234 0.050 PCC thickness 0.000 0.000 0.000 0.002 0.000 0.001 0.000 0.000 Base Type 0.000 0.000 0.000 0.006 0.387 0.863 0.000 0.016 PCC Flexural Strength 0.948 0.550 0.746 0. 602 0.300 0.024 0.586 0.021 Lane Width 0.425 0.327 0.186 0.941 0.673 0.609 0.570 0.366 Site (blocked) 0.000 0.012 0.000 0.000 0.000 0.003 0.000 0.006 N=156 R2=0.816 N=156 R2=0.701 N=156 R2=0.772 N=156 R2=0.716 N=156 R2=0.604 N=156 R2=0.427 N=156 R2=0.630 N=156 R2=0.450 Table 6- 40 Summary of p-values for the effects of design factors on rigid pavement response do d6 Area Factor Effective stiffness of PCC slab Design Factor Initial Final Initial Final Initial Final Initial Final Subgrade 0.001 0.014 0.707 0.069 0.687 0.207 0.018 0.792 Zone 0.210 0.570 0.000 0.025 0.000 0.569 0.000 0.173 Subgrade*Zone 0.000 0.360 0.003 0.211 0.087 0.187 0.306 0.207 N=156 R2=0.783 N=156 R2=0.665 N=156 R2=0.55 N=156 R2=0.55 N=156 R2=0.254 N=156 R2=0.314 N=156 R2=0.465 N=156 R2=0.357

361 6.7 APPARENT RELATIONSHIP BETWEEN RESPONSE AND PERFORMANCE In this section of the report the observations regarding apparent relationships between rigid pavement response (FWD testing) and performance are presented. The usefulness of such relationships can be of two kinds— explanatory and predictive (see Chapter 5). Relationships could not be established based on data for all the sections in SPS-2 experiment because of “low” occurrence of distresses (see Chapter 3). Predictive relationships were established based on bivariate correlation analyses and scatter plots, at the site level, for sections in those sites that exhibited some distress at this point in time. The DLR data were used for predictive relationships regarding the instrumented sections in Ohio and North Carolina. 6.7.1 Site Level Analyses— Predictive Relationships This section summarizes the findings regarding predictive relationships between initial response (FWD deflection or deflection-based indices) and future pavement performance (cracking), at the site level. Various deflection-based indices were calculated based on the individual deflection basins for each section. These indices include: • do (peak deflection under the load), • d6 [farthest deflection at 1.5 m (60 in.) away from the load], • AF (Area Factor), and • ES (effective stiffness of upper (bound) layer. Bivariate correlation analyses between response parameters (deflections or deflection basin parameters (DBPs)) and transverse cracking were conducted for AZ (4), OH (39) and MI (26) of the SPS-2 experiment. These sites were selected based on the extent of occurrence of the distress. Relationships between roughness and deflection parameters were explored within each site. The latest performance for each section within the SPS-2 experiment was used in these analyses. It is to be noted that for a site, age, traffic, construction, material properties and environment are the same and thus this provides a good opportunity for seeking apparent relationships.

362 Figure 6-11 through Figure 6-16 are examples of bivariate relationships between Effective Stiffness (ES), Area Factor (AF), and d0 with transverse cracking and roughness. Among sections in AZ (4), a negative correlation was observed between ES and IRI. Sections with higher stiffness (i.e. higher ES) showed higher roughness (see Figure 6-11). Sections with higher stiffness also showed potential for transverse cracking (see Figure 6-12). Sections with lower stiffness (i.e. lower AF or lower ES) have shown higher transverse cracking in MI (26) (see Figures 6-13 and 6-14). Among the sections in OH (39), higher roughness was observed for those with higher peak deflection and sections with higher stiffness had higher transverse cracking (see Figures 6-15 and 6-16). Table 6-41 is a summary of correlation coefficients from the bivariate analyses for roughness and various deflection parameters within each site. The deflection basin parameters do not have a consistent association with roughness. This inconsistency may be explained in light of “low” magnitude of roughness for SPS-2 sections at this point in time.

363 y = -3E-06x + 1.9165 R2 = 0.3448 0 0.5 1 1.5 2 0 100000 200000 300000 400000 Effective Stiffness (psi) IR I ( m /k m ) Figure 6 - 11 Roughness and ES relationship ─ State (4) Arizona y = 9E-05x - 12.604 R2 = 0.6044 0 5 10 15 20 0 100000 200000 300000 400000 Effective Stiffness (psi) Tr an sv er se C ra ck in g (N o. ) Figure 6 - 12 Transverse cracking and ES relationship─ State (4) Arizona y = -3E-05x + 8.9627 R2 = 0.4252 0 2 4 6 8 10 12 14 0 100000 200000 300000 400000 Effective Stiffness (psi) Tr an sv er se C ra ck in g (N o. ) Figure 6 - 13 Transverse cracking and ES relationship ─ State (26) Michigan y = -0.1036x + 10.452 R2 = 0.5374 0 2 4 6 8 10 12 14 0 50 100 150 Area Factor Tr an sv er se C ra ck in g (N o. ) Figure 6 - 14 Roughness and ES relationship ─ State (4) Arizona y = 0.005x + 1.0002 R2 = 0.1768 0 0.5 1 1.5 2 0 50 100 150 Peak Deflection (microns) IR I ( m /k m ) Figure 6 - 15 Roughness and Do relationship ─ State (39) Ohio y = -0.0001x + 26.158 R2 = 0.2072 0 5 10 15 20 25 30 0 100000 200000 300000 Effective Stiffness (psi) Tr an sv er se C ra ck in g (N o. ) Figure 6 - 16 Transverse cracking and ES relationship ─ State (39) Ohio

364 Table 6- 41 Summary of correlations for deflections and DBPs with IRI State ID State Do AF ES Zone SG 10 DE -0.29 0.53 0.36 WF C 19 IA 0.26 0.18 -0.09 WF F 20 KS -0.38 -0.36 0.24 WF F 26 MI -0.05 -0.38 -0.27 WF F 38 ND -0.44 0.30 0.23 WF F 39 OH 0.41 -0.03 -0.32 WF F 55 WI 0.31 -0.25 -0.42 WF C 5 AR 0.37 0.31 0.00 WNF C 37 NC -0.19 0.53 0.66 WNF F 8 CO -0.37 -0.27 0.02 DF F 53 WA -0.42 -0.55 -0.44 DF F 4 AZ 0.22 -0.43 -0.58 DNF C 6 CA -0.56 0.04 0.04 DNF C (-) ρ 8 7 6 (+) ρ 5 6 7 Mean -0.09 -0.03 -0.04 Std 0.36 0.37 0.36 CoV 4.1 13.0 8.3

365 6.7.2 Relationship between strain and performance This section is regarding the Dynamic Load Response (DLR) of the instrumented rigid pavement sections in the states of Ohio and North Carolina. Four sections were instrumented with strain gauges and LVDTs to measure the pavement “response”, at each site (details in Chapter 2). An attempt was made to relate the observed performance of the instrumented sections with measured responses (strains and deflections of PCC slab). The attempt was limited by low distresses for the instrumented pavement sections, especially among sections in NC. Consequently, no significant findings could be made, at this point in time. However, some observations regarding the dynamic load response of the instrumented test sections are summarized below: • Higher deflections and strains occurred in the section with DGAB compared to the sections with LCB. • Strains and deflections were higher in the sections with LCB compared to the section with PATB. • Higher strains and deflections were observed in the section with 8”-thick [203 mm] PCC slab compared to the section with 11”-thick [279 mm] PCC slab

366 6.8 SYNTHESIS OF ANALYSES RESULTS FOR RIGID PAVEMENTS This section of the report summarizes all the findings from various analyses performed on SPS-2 data. The methods employed in this study were explained in Chapter 4 and the results obtained from these analyses were presented in this chapter. The synthesis of results from analyses is similar to that of flexible pavements of the SPS-1 experiment (see Chapter 5). As mentioned before, broadly two types of analyses (overall) were performed−magnitude-based and frequency-based. ANOVA, which is a method for comparing means, is the magnitude-based analysis. Binary Logistic Regression (BLR) and Linear Discriminant Analysis (LDA) are frequency-based analyses, which give the likelihood of occurrence and non-occurrence of distresses. The site-level analyses were used to compare the performance of pavements within each site. The results from site-level analysis were used to ascertain the consistency of the effects (of experimental factors) across all sites. The magnitude-based methods, though powerful, are more appropriate for analyses of distresses, which have “fairly high” occurrence (for example, roughness, and faulting). On the other hand, the frequency-based methods are more suitable when magnitude of a distress is low but the occurrence of the distress is considerable (for example, transverse cracking, and longitudinal cracking). An attempt has been made to ‘summarize’ the above said effects of design and site features on the performance and response measures. The results were interpreted in light of the type of analysis, and occurrence and extent of distress. ANOVA being the most “powerful” among the methods was given higher importance for distresses with “good” occurrence and/or extent (roughness and faulting). However, the results from this analysis suffer seriously in case of limited (low occurrence of distress) and unbalanced data. Therefore, in these cases, the effects of experimental features, mainly on occurrence of distresses, were investigated using BLR and LDA. The results from site-level analyses (paired comparisons and comparison of designs) and data exploration (extent of distresses) were then considered to confirm the findings. Based on the site-level analyses the consistency of effects was ascertained. All results need to be interpreted in light of the experiment design, occurrence and extent of distresses, and analyses methods used. A “weak” effect at this point in time may become a “medium” or “strong” effect in the long term. Hence, all the conclusions are based on “mid- term” performance of the ongoing SPS-2 experiment.

367 The synthesis of results is presented next, separately for each performance measure. A ‘simple’ summary of results from all analyses is

This page is intentionally left blank. 368

369 Table 6- 42. The summary is only meant to give the reader an idea about the effects. The reader is strongly recommended to read the following write-up for a better understanding of all the effects. It is important to note that a “strong”, “medium” or “weak” effect of a factor should only be interpreted as “strong”, “medium”, or “weak” difference in effects of different levels of a factor, respectively. In other words, a “strong” effect of PCC slab thickness and a “strong” effect of subgrade soil type should not be interpreted as PCC slab thickness and subgrade type having the same strength of effect. In this summary of effects of design and site factors, a black circle was used to indicate “strong” effect (significant), a grey circle was used to indicate a “medium” effect, and a white circle was used to indicate a “weak” effect. Operational significance was determined only for “strong” or “medium” effects. It should be noted that an effect can be statistically significant (meaning that it is not a happenstance) but may not be operationally/ practically significant, at this point in time. The SPS-2 experiment, entitled Strategic Study of Structural Factors for Rigid Pavements, is one of nine special pavement studies in the LTPP program. The main objectives of this experiment are to determine the relative influence and long-term effectiveness of the structural factors affecting performance of jointed plain concrete pavements (JPCP). These factors include PCC slab thickness, base type, in-pavement drainage, PCC flexural strength and lane width. The key conclusions from this study are summarized below. 6.8.1 Effects of experimental factors for SPS-2 experiment This section is subdivided into three parts: (i) pavement performance, (ii) pavement response, and (iii) relationship between response and performance. The structural/design factors include PCC slab thickness, base type, drainage, PCC flexural strength and lane width. The experiment also includes studying the secondary effects of site factors, namely subgrade type and climatic zones. 6.8.1.1 Effect of Design and Site Factors on Pavement Performance The effects of the experimental factors on each performance measure are discussed below, one performance measure at a time.

370 PCC slab thickness and base type seem to be the most important factors affecting the occurrence of transverse cracking, whereas, drainage seems to have a marginal effect. The occurrence of transverse cracking among pavements with 203 mm (8-inch) PCC slab thickness is higher than that among those with 279 mm (11-inch) PCC slab thickness. Also, the occurrence of transverse cracking among pavements constructed with LCB is higher than that among those with PATB/DGAB or with DGAB. Pavements with PATB/DGAB have shown the “best” performance (least occurrence of cracking). These effects of PCC thickness and base type are statistically significant, as suggested by the frequency-based analyses. The analyses indicate a marginal effect of drainage on the occurrence of transverse cracking. Sections without drainage have slightly higher likelihood of cracking than sections with drainage. On average, among sections built with LCB, those with 203 mm PCC slab have higher occurrence of cracking than those with 279 mm PCC slab. It is important to interpret these results in light of the construction issues, i.e. shrinkage cracking in LCB. Pavements built on fine-grained soils have slightly higher chances for the occurrence of transverse cracking than those built on coarse-grained soils. This effect was found to be marginally significant. Longitudinal Cracking PCC slab thickness and base type seem to be the most important factors affecting the occurrence of longitudinal cracking. The occurrence of longitudinal cracking among pavements with 203 mm PCC slab is higher than among those with 279 mm. Also, the occurrence of longitudinal cracking among pavements constructed with LCB is higher than among those with PATB/DGAB or with DGAB. Pavements with PATB/DGAB have shown the “best” performance (least occurrence of cracking). These effects of PCC thickness and base type are statistically significant, as suggested by the frequency-based analyses. On average, the above effect of PCC slab thickness is more prominent among sections built with LCB. However, it is important to interpret these results in light of the construction issues i.e. shrinkage cracking in LCB.

371 The extent of faulting among the test sections is “low”, with 62% of the sections having no joints with faulting greater than 1 mm. Only 33% of the sections have 0 to 20% of the joints that faulted more than 1.0 mm, and just 5% of the sections have more than 20% of the joints that faulted more than 1.0 mm. A majority of SPS-2 sections seem to be exhibiting “good” performance with respect to joint faulting, at this point in time. This performance seems to be reasonable as the test sections are “young” and have doweled joints at 4.6 m (15 ft) spacing. Therefore, the results at this point may only indicate the initial trends/observations that may not be of much practical significance. Among all the design factors, lane width seems to be most important for faulting of PCC joints. In general, pavements with standard lane [3.7 m (12 ft) wide lane] have shown higher faulting than those with widened lane [4.3 m (14 ft) wide lane]. This effect of lane width is statistically significant, as suggested by magnitude-based methods. However, the effect may not be of practical significance because of the low occurrence of faulting. The effect of lane width is more prominent among sections built on fine-grained soils than among those built on coarse-grained soils. Also, the effect is more pronounced among sections located in WF zone. Among sections located in WF zone and built on fine-grained soils, those with drainage have slightly lower (with marginal statistical significance) faulting than those without drainage. These effects are of statistical significance, and may not be practically significant, at this point in time. Among sections located in WF zone and built on coarse-grained soils, sections with 8” (203 mm) PCC slab have slightly higher faulting than those with 279 mm PCC slab. This effect is statistically significant but not of practical significance. It is important to note that according to the experiment design of SPS-2, sections with 203 mm PCC slabs are built with dowels of 32 mm (1.25-inch) diameter, whereas sections with 279 mm PCC slabs are built with dowels of 38 mm (1.5-inch) diameter. Hence, the effect of dowel diameter and the effect of PCC slab thickness on faulting are not separable. Faulting

372 The initial roughness (smoothness) of the pavement sections in the experiment seems to be affected by the PCC slab thickness. Pavements with thicker slab (279 mm) were found to have more initial roughness compared to those with thinner slab (203 mm). Drainage and base type seem to be the most important factors affecting the growth in roughness, whereas, slab thickness seems to have a marginal effect. Pavements without drainage have shown higher change in roughness than those with drainage. Also, pavements constructed with PATB have shown lower change in IRI (∆IRI) compared to those with DGAB or LCB, while pavements with DGAB have the highest change in roughness. These effects of drainage and base type are of statistical and practical significance, as suggested by magnitude-based methods. Among pavements constructed with standard lane (3.7 m), sections with DGAB have shown higher ∆IRI than those with LCB or PATB. This effect is of practical significance but is only marginally significant, statistically. Among pavements built on fine-grained soils, those with 203 mm PCC slab have higher ∆IRI than those with 279 mm PCC slab. This effect is more prominent among sections located in WF zone. Also, the effect of drainage (i.e. sections with PATB) is more prominent among sections located in WF zone and built on fine-grained soils. Among sections located in WF zone and built on fine-grained soils, those with drainage (i.e. sections with PATB) have shown lower ∆IRI compared to those without drainage. These marginally significant effects of drainage and PCC slab thickness are of practical significance. The above results suggest that the change in roughness can be inhibited by constructing pavements with PATB and drainage as compared to sections with DGAB or LCB, especially in the case of pavements built on fine-grained soils. Also, among pavements built on fine-grained soils, an increase in PCC slab thickness from 203 mm to 279 mm seems to help prevent an increase in pavement roughness. 6.8.1.2 Effect of Design and Site Factors on Pavement Response An ANOVA was conducted with the peak deflection under the FWD load plate (do), the far sensor deflection at 60 inches (1524 mm) from the FWD load (d6), the “Area Factor” (AF), and Effective Stiffness (ES) of the PCC slab as the dependent variables. All the response parameters have been calculated using the midslab deflections. Roughness

373 The sections constructed on DGAB have shown significantly higher deflections than the ones constructed on PATB. The sections constructed on LCB experienced the least amount of deflections. Also the 203 mm (8-inch) thick slab sections deflected more than the 279 mm (11- inch) thick slabs. In the Wet Freeze zone, the sections with fine subgrade soils were found to have significantly higher deflections than the sections that were built on coarse subgrade soils. Similar results were obtained from the analysis of the latest (or final) do values. Far Sensor Deflection (d6) The sections constructed on DGAB have shown significantly higher deflections than the ones constructed on PATB. The sections constructed on LCB exhibited the lowest deflections. Also the sections with 203 mm (8-inch) PCC slabs deflected more than those with 11-inch (279 mm) PCC slabs. Similar results were obtained from the analysis of the latest (or final) do values. In the Wet Freeze zones, the sections with fine subgrade soils were found to have significantly higher deflections than the sections that were built on coarse subgrade soils. Sections located in “freeze” climate have shown significantly higher deflections than the ones located in “no freeze” climate. Area Factor (AF) The sections with 11-inch (279 mm) slab have higher AF than those with 203 mm (8- inch) slab. Sections located in “wet” climate have higher AF than those in “dry” climate. An interaction between base type and subgrade was found to be significant, in that, among sections with LCB, those constructed on coarse-grained subgrade have higher AF than those constructed on fine-grained subgrade soils. These effects were not statistically significant on final survey AF values. Sections with higher strength concrete have significantly higher AF than those with lower strength concrete. Effective Stiffness (ES) The effect of PCC thickness on ES is more significant among sections with DGAB than among those with LCB. The effect of PCC flexural strength on ES is more apparent for sections with DGAB or PATB than for sections with LCB. Pavements located in “wet” climate have Peak Deflection under FWD load (do)

374 significantly higher ES values compared to those located in “dry” climate. The sections built on coarse-grained subgrade soil were significantly stiffer than those built on fine-grained soil. The effects of PCC thickness and base type on ES from final survey were similar as in the case of initial ES. However, sections built with drainage have higher ES than those without drainage. Also, sections with higher strength concrete have higher ES than those with lower strength concrete. A simplified summary of results from all analyses is given in Table 6-42. The summary is only meant to give an overall assessment of the effects. The reader is strongly recommended to read the following write-up for a better understanding of all the effects. It is important to note that a “strong”, “medium” or “weak” effect should only be interpreted in terms of the difference in effects at the various levels of a factor. As an example, a “strong” effect of PCC slab thickness and a “strong” effect of subgrade soil type should not be interpreted as PCC slab thickness and subgrade type having the same strength of effect. A black circle indicates a “strong” effect (significant); a grey circle indicates a “medium” effect, and a white circle indicates a “weak” effect. Operational significance was determined only for “strong” or “medium” effects. It should be noted that an effect can be statistically significant (meaning that it is not a coincidence) but may not be operationally/ practically significant, at this point in time.

375 Table 6- 42 Summary of effects of design and site factors for rigid pavements Performance Measures Response Measures (initial) Design Factor Transverse cracking Longitudinal cracking Faulting Roughness do d6 AF ES Drainage PCC thickness Base type Flexural Strength Lane Width Climatic Zone Subgrade type Note: This table is solely for the purpose of summarizing some of the effects in a ‘simple’ format. The reader is urged to read relevant text in the report for a better understanding. Symbol Description Strong Effect (Main effect exists) Medium Effect (Interaction effect) Weak Effect