U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-HRT-05-068
Date: October 2005
Analysis of LTPP data indicated major differences in roughness progression between doweled and nondoweled pavements. The doweled and nondoweled pavements were divided into three data sets based on the rate of increase of roughness:
The percentage of nondoweled pavements that fell into data sets 1, 2, and 3 was equal (each set consisted of 33 percent of nondoweled pavements). The percentage of doweled sections that fell into data sets 1, 2, and 3 was 56, 25, and 19 percent, respectively. These results clearly show that doweled pavements are providing a superior performance from a roughness point of view.
The roughness progression plots for PCC pavements showed a parallel pattern. This finding indicates that pavements that are built smoother will provide a longer service life before reaching a terminal roughness value, compared to pavements having a lower initial smoothness level.
There was a significant difference in faulting between doweled and nondoweled pavements. Ninety-four percent of the doweled pavements had a total faulting within a 152.4-m (500-ft)-long section that was less than 38 mm (1.5 inches), compared to 39 percent for the nondoweled pavements. A total faulting of 38 mm (1.5 inches) corresponds to an average faulting of less than 1.2 mm (0.05 inch) at a joint. The average total faulting for data sets 1, 2, and 3 for doweled pavements were 15, 13, and 38 mm (0.59, 0.51, and 1.5 inches), respectively, whereas the values for nondoweled pavements were 32, 47, and 119 mm (1.26, 1.85, and 4.69 inches), respectively. A strong relationship existed between the roughness level and total faulting for nondoweled pavements (correlation coefficient of 0.70), whereas a similar relationship was not seen for doweled pavements (correlation coefficient of 0.07). The dowels in the doweled pavements are serving their intended function by providing load transfer and thus preventing faulting.
The majority of nondoweled pavements had an upward curvature; 73 percent of the sections showed an upward curvature at the last profile date. Generally, the pavements that fell into data set 1 have not shown a major change in curvature over the monitored period, which on average was about 10 years. However, pavements that fell into data sets 2 and 3 have shown an increase in curvature over time. Sections in data set 3 showed significant upward curvature, and the 172 amount of curvature was increasing over time. The pavements with a high degree of curvature had high IRI values, showed a high rate of increase of roughness, and had high faulting. The pavements that had very little curvature initially and whose curvature did not change over time showed very good performance from a roughness point of view. Increasing values of the following factors were associated with higher curvature: CTE, freezing index, and elastic modulus of concrete, while increasing values of the following factors were associated with lower curvature: mean annual temperature, annual precipitation, number of wet days per year, and PCC slab thickness.
Overall, the magnitude of slab curvature seen in doweled pavements was much less than that seen for nondoweled pavements. The majority of the sections had downward curvature; 63 percent of the sections showed a downward curvature at the last profile date. Generally, the pavements that fell into data set 1 had very little slab curvature, and they were essentially flat, and showed little change in curvature over time. However, the range of curvature seen for pavements in data set 2 increased over time. The slabs in all pavements that fell into data set 3 showed downward curvature. Sections that showed excessive amounts of downward curvature had high roughness values and high rates of increase of roughness. For doweled pavements, higher values of the following factors resulted in more downward curvature: freezing index, number of days in a year below 0 °C (32 °F), ratio between PCC elastic modulus and split tensile strength, and weight of cement in mix. Higher values of the following parameters resulted in lower downward curvature: mean annual temperature, annual precipitation, days in a year with temperature greater than 32 °C (90 °F), and PCC slab thickness.
The strongest factor affecting roughness progression was faulting. Many sections that had high faulting also had high CI values. It appears that excessive PCC slab curvature increases the potential for faulting. The worst performing nondoweled pavements from a roughness point of view were mostly located either in areas where the mean annual temperature was low or the freezing index was high. (In freezing regions, the freezing index is correlated to mean annual temperature.) In nondoweled pavements, load transfer is provided by aggregate interlock. When the temperatures are low, the amount of load transfer between the slabs will be low. This situation increases the potential for faulting.
Generally, pavements having higher split tensile strength values appear to be performing better from a roughness point of view. Higher split tensile values indicate higher flexural strengths. Generally, pavements having high elastic modulus values (greater than 35,000 MPa (5.08 million psi)), or pavements having a high value for the ratio between elastic modulus of concrete and split tensile strength (greater than 8,000) appear to be showing high rates of increase of roughness. Several pavement sections showed a high rate of increase of IRI, where a primary factor causing the increase in IRI was the increase in upward slab curvature over time.
No factor that had a strong relationship to roughness progression could be identified. There were some weak trends in the data that slab length, split tensile strength, and coarse-to-fine aggregate ratio may be affecting roughness progression. The median slab length of the set of pavements that showed a rate of change of IRI of less than 0.02 m/km/yr (1.27 inches/mi/yr) and greater than 0.02 m/km/yr (1.27 inches/mi/year) were 4.8 and 5.3 m (15.7 and 17.4 ft), respectively. Overall, a higher split tensile strength, which is related to higher flexural strength of concrete, appears to be beneficial to providing a good performance from a roughness viewpoint. It appears that higher values of coarse-to-fine aggregate ratio in the concrete provided pavements that maintained their smoothness over a long period. The trends seen for nondoweled pavements for elastic modulus, or the ratio between elastic modulus and tensile strength, were not seen for the doweled pavements. Several doweled sections had high elastic modulus values or high values for the ratio between elastic modulus and split tensile strength that did not exhibit poor performance.
In the two Iowa projects, the joints were formed by performing a sawcut that had a width of 6 mm (0.25 inch) and a depth of 25 mm (1 inch), and then the joints were sealed. Profile data were collected on these projects when the joint was in an unsealed as well as in the sealed condition. For both cases, the lightweight profiler recorded the joint as a feature spread over a distance of 300 mm (11.8 inches). The recorded depths of the joint in the unsealed and sealed condition were 2 mm (0.08 inch) and 1 mm (0.04 inch), respectively.
In the test section located in Pennsylvania, the joints were formed by performing an initial sawcut that was 3 mm (0.12 inch) wide, then a joint reservoir having a width of 9.5 mm (0.37 inch) and a depth of 38 mm (1.50 inches) was sawed, and then the joint was sealed. Profile data were collected when the joints were in these three conditions. For all three cases, the lightweight profiler recorded the joint as a feature spread over a distance of 220 mm (8.7 inches). The recorded depth of the joint when the initial sawcut was present, when the reservoir was sawed with the joint unsealed, and when the joint was sealed were 1.5, 3.5, and 1.5 mm (0.06, 0.14, and 0.06 inch), respectively.
In the test sections located in Michigan, the joint forming procedure was similar to that used in the Pennsylvania project. The joint dimensions were also similar to those in the Pennsylvania project. The joint could not be detected in the data collected when the 3-mm (0.12-inch)-wide initial sawcut was present on the pavement. For the other two cases, the profiler recorded the joint as a feature spread over a distance of 450 mm (17.7 inches). The recorded depth of the joint when the reservoir was sawed with the joint unsealed, and when the joint was sealed, was 4 and 1.4 mm (0.16 and 0.05 inch), respectively.
The lightweight profilers used to measure profiles on the five paving projects had data recording intervals ranging from 30 to 76 mm (1.2 to 3.0 inches). However, these profilers obtain height sensor data at much closer intervals, and average the values when computing the profile at each data recording interval. In addition, the profilers appear to be applying a low-pass anti-alias filter on the data. The averaging procedure used on the height sensor data and the anti-alias filter will flatten out the depth of the joint that appears in the profile and cause the joint to be spread over a much wider distance than the actual width. There are differences in these procedures between profilers manufactured by different manufacturers. Hence, the differences observed in these projects on how a joint was recorded in the profile data are attributed to different procedures used by the different profilers in manipulating the data before recording. Depending on the procedure used to manipulate the data, the condition of the joint during data collection may or may not have an influence on the smoothness index computed from that data.
The profilers used in the Iowa and Pennsylvania projects significantly attenuated the depth that was recorded at a joint when the joint was unsealed. In these projects, the IRI after the joint was sealed was almost identical to the IRI obtained when the joint was unsealed. This is because the depth of the joint that was recorded when the joint was unsealed was not large enough to affect the IRI. For the projects in Michigan, IRI obtained when the joint reservoir was formed, but with the joint unsealed, was 18 percent lower that that obtained when the reservoir was sealed. In the Michigan projects, the depth of the joint recorded by the profilers when the joint was unsealed was large enough to affect the IRI.
For the Iowa project, the RN values obtained after the joints were sealed were about 2 to 3 percent higher than those obtained before sealing the joints. For the Michigan projects and the Pennsylvania project, RN obtained from data collected when the joint reservoir was formed but the joint was unsealed was on average 14 and 28 percent lower, respectively, than that obtained after the joint was sealed.
These results indicate that profiling should not be performed when the joint reservoir is made on the pavement with the joint unsealed if RN is to be computed from the data. It is best to follow a similar approach when collecting data to compute IRI. For some profilers, this may not be an issue, but it could be for some profilers. There was no evidence to suggest IRI or RN computed from data collected under the following two conditions would be different: (1) 3-mm (0.12-inch)-wide initial sawcut was present on the pavement and (2) joint reservoir was sawed and sealed.
For the project in Pennsylvania, the two projects in Michigan, and the I-80 project in Iowa, negligible PCC slab curvature was observed for all data collection sequences. The slabs were essentially flat. In all of these projects, virtually no changes in slab curvature were observed between the different profiling sequences.
The U.S. 20 project had the highest CI at the first profile date of all projects. Some changes in CI between the profiling sequences were noted at this section. An evaluation of the profile data indicated that CI was being influenced by variations in the profile within the slab, and not by changes that occurred because of movements at the joints. The right wheel path in this section had a repetitive wave having a wavelength of approximately 1.6 m (5.2 ft). The amplitude of this wavelength in the profile appeared to be different during the different profiling sequences, and this phenomenon was having an influence on the CI.
For the Pennsylvania project, little change in IRI was noted for the different test sequences. IRI can be considered to have remained at the same value over the 3.5-month monitored period.
At the U.S. 20 project in Iowa, little change in IRI was noted for the different test sequences along the left wheel path over the 3-month monitoring period. Along this wheel path, changes in IRI compared to IRI obtained immediately after paving ranged from -1 to 5 percent for the different profiling sequences. Along the right wheel path, little change in IRI was noted for data obtained up to 9 days (change of -1 to -3 percent). However, the data collected at 3 months showed a reduction in IRI of 10 percent compared to IRI obtained 1 day after paving. The cause for this reduction could not be determined from the profile data.
In the I-80 project in Iowa, two 152.4-m (500-ft)-long sections were considered. The lowest changes in IRI were noted along the right wheel path of the first 152.4-m (500-ft)-long section, whereas the changes in IRI with respect to IRI immediately after paving for the different profiling sequences ranged from -2 to 6 percent. The highest changes in IRI were noted along the left wheel path of the second 152.4-m (500-ft)-long section, where changes in IRI with respect to IRI immediately after paving ranged from -1 to 14 percent.
For the U.S. 23 project in Michigan, profile data collected 1 day after paving showed humps appearing at joint locations. These humps appear to have been caused by the joint saw residue and caused IRI to increase. A reduction in IRI from that obtained immediately after paving was seen for data collected at 5 days, 10 days, and 1 year after paving. For the 10-day data, the reduction in IRI compared to IRI obtained immediately after paving varied from 15 to 21 percent for the different wheel paths. This reduction is attributed to the residue from the joint sawing operation being washed away by rain. IRI obtained 1 year after paving was lower than that obtained at 10 days after paving for all wheel paths, except for the right wheel path of the outside lane. The reduction in IRI between 1-year and 10-day data averaged 15 percent for the three wheel paths where a reduction in IRI occurred, while for the other wheel path IRI increased by 1 percent. It is unclear whether this reduction in IRI was due to changes in pavement shape or due to differences in profile data collection capabilities between the lightweight profiler used to obtain the 10-day data and the high-speed profiler used to obtain the 1-year data.
For the I-69 project in Michigan, a reduction in IRI from that obtained at 1 day after paving was seen for data collected at 6 days and 10 days after paving. The reduction in IRI ranged from 4 to 11 percent for the different wheel paths and test dates. The joints in the pavement had been sawn just before profile data collection on the first day the pavement was profiled, and residue from the joint sawing operation was present adjacent to the joints at the time of profiling. The higher IRI obtained for day 1 may have been caused by this residue. IRI obtained 4.5 months after paving was higher than that obtained immediately after paving by 9 and 22 percent for the left and the right wheel paths, respectively. The 1-day profiling was performed with a lightweight profiler, while the 4.5-month profiling was performed with a high-speed profiler. It is unclear whether the difference in IRI was caused by a change in the pavement profile or is related to differences in the two profilers' data collection capabilities. Also, a median barrier was present when the 4.5-month data were collected, and the path followed by the high-speed profiler during profiling may have been different from that followed by the lightweight profiler. The fact that different paths were followed might also have been a contributing factor to differences in IRI.
The CTE values of the concrete used in these projects ranged from 8.25 x 10-6 to 13.2 x 10-6 per °C (4.58 x 10-6 to 7.33 x 10-6 per °F). No effect of CTE on short-term changes in IRI could be detected.
The effect of joint condition (sawed, not sealed versus sealed) was discussed previously. When addressing changes in RN for the different projects, RN obtained when the joint reservoir is sawed but unsealed will not be considered.
In the Pennsylvania project, little change in RN was noted for the different data sets. The changes in RN of the different wheel paths over the 3.5-month monitoring period, compared to RN obtained 1 day after paving, ranged from -6 to 3 percent.
For the U.S. 20 project in Iowa, a slight increase in RN was noted over the 3-month monitoring period when RN obtained at different test sequences was compared with the 1-day values. However, the increase in RN was less than 5 percent.
For the I-80 project in Iowa, the RN values for both wheel paths of the two 152.4-m (500-ft)-long sections showed little change for the different profiling sequences over a 1-month period. The change in RN compared to RN obtained immediately after paving was within ±3 percent.
For the U.S. 23 project in Michigan, the RN obtained at profiling times that varied from 5 days after paving to 1 year after paving showed little change when compared to RN obtained immediately after paving. The variations in RN that occurred, compared to the 1-day RN, ranged from -4 to 7 percent with no consistent trend occurring over time.
For the I-69 project in Michigan, little change in RN occurred for the different profiling sequences that varied from 1 day after paving to 4.5-months after paving. When all profiling paths and test sequences were considered, the variations in RN that occurred, compared to the 1-day RN, ranged from -6 to 7 percent.
The surface of the pavement must be clean when performing profile measurements. Residue from the sawcutting operation present adjacent to the transverse joint can appear as small humps in the measured profile and can affect the smoothness indices computed from the profile data. On one paving project where such residue was observed, an increase in IRI ranging from 17 to 25 percent for the four evaluated wheel paths could be attributed to the humps created by the residue.
For each data set, the lightweight profilers obtained three repeat runs. When IRI obtained from the three repeat runs for the entire section were evaluated for all data sets, the average difference between the maximum and minimum IRI for the S.R. 6220 project in Pennsylvania, U.S. 23 project in Michigan, and I-69 project in Michigan were 0.03, 0.06, and 0.04 m/km (1.9, 3.8, and 2.5 inches/mi), respectively. These three projects had transverse tining. When a similar analysis was performed for the two projects in Iowa that had longitudinal tining, the average difference between the maximum and minimum IRI was 0.06 and 0.09 m/km (3.8 and 5.7 inches/mi) for the U.S. 20 and I-80 project, respectively. The lateral wander during profiling would cause the laser dot of the height sensor to obtain measurements on top of the tine as well as the bottom of the tine. This is attributed as the cause for the lower repeatability of IRI observed on the longitudinally tined pavements.
An evaluation of short-interval IRI repeatability using 15-m (49-ft) segment lengths showed the average difference between the maximum and minimum IRI obtained from the repeat runs for the 15-m (49-ft)-long segments in the S.R. 6220 project in Pennsylvania, U.S. 23 project in Michigan, and I-69 project in Michigan were 0.09, 0.10, and 0.11 m/km (5.7, 6.3, and 7.0 inches/mi), respectively. A similar analysis for the two projects in Iowa indicated values of 0.25 and 0.12 m/km (15.9 and 7.6 inches/mi) for the U.S. 20 and I-80 projects, respectively. The short-interval IRI repeatability for the U.S. 20 project was very poor. These results indicate that implementing an IRI-based specification that relies on short-interval IRI (e.g., 15 m (49 ft)) is not suitable for pavements that have longitudinal tining. There were several cases where significant differences in IRI between the runs were noted for individual 15-m (49-ft)-long segments in all projects. Although there can be differences in IRI between 15-m (49-ft)-long segments, when IRI is computed over longer distances such as 152.4 m (500 ft), these differences can cancel out, and IRI from repeat runs can give excellent agreement.
If a profiler is used to measure smoothness of concrete pavements, it is advisable to certify the profiler on PCC sections. There could be differences in the way profilers treat joints, which can affect smoothness indices obtained from the profile data. In addition, treatment of tining may be different between devices. Hence, certifying profilers on asphalt surfaces and using the profiler to measure smoothness on concrete surfaces may not necessarily mean comparable smoothness indices will be obtained between devices.
Based on the five projects used in this study, it appears that smoothness measurements can be performed at any time within the first few months after paving.
Usually smoothness indices like IRI are computed for each lane over a 161-m (528-ft) length for construction acceptance. The IRI for the overall section does not provide any information about how IRI varies within the section, or where rough spots within the section are located. A roughness profile of the section can be used to investigate how roughness varies within the section and identify where rough spots within the section are located. If rough spots are detected within the section, the location of these events could be correlated to the construction process or a pavement feature to obtain information about a specific construction event or a pavement feature at that location that resulted in a high roughness value. Overlaid roughness profiles of the left and right wheel paths can be used to see how roughness varies between the wheel paths. In addition, overlaid roughness profiles of both wheel paths of the inside and outside lane can be used to see how roughness varies across the entire pavement width. These procedures help detect problems in the construction process that results in rough spots.
PSD plots can be used to identify whether a roughness associated with a specific wavelength is predominant in the pavement. A wavelength that has a significant contribution to the roughness will appear as a spike in the PSD plot. If spikes are detected, the cause for the prominent wavelength to occur in the profile can be investigated, whether it stems from the equipment used for paving or the finishing process. For example, data from the U.S. 20 project showed a repetitive wave with a wavelength of 1.6 m (5.2 ft) appearing along the right wheel path. Any repetitive feature occurring in a profile can be easily detected by this technique.
Analyzing roughness profiles, as well as using PSD plots to evaluate data collected at the start of a paving project, will indicate whether any features in the profile are contributing to a high roughness. If such features are detected, these methods will provide an opportunity to troubleshoot, identify the problem, and then correct it at the start of the project.
Overall, when all tested wheel paths and lanes were considered, the average IRI values for the S.R. 6220 project in Pennsylvania, U.S. 20 project in Iowa, I-80 project in Iowa, U.S. 23 project in Michigan, and I-69 project on Michigan were 1.11, 1.44, 0.95, 0.80, and 1.07 m/km (70, 91, 60, 51, and 68 inches/mi), respectively.
The best smoothness was obtained at the U.S. 23 project. This is the only project where tie bars were not inserted by the paver, but fixed to the base on chairs. Fixing the tie bars to the base will be more costly than inserting them during paving. It appears that the contractor in this project believed that better smoothness could be achieved by using this procedure. The contractor did indeed construct a very smooth pavement in this project.
The only project where a spreader was not used, and also where dowel bars were inserted, was the I-69 project. Out of the five projects studied, this project had the third best smoothness.
A survey of State DOT personnel and concrete industry personnel was performed to get their opinion on whether contractors have been adjusting their mix designs to achieve higher smoothness. The general consensus was that no modifications have been required in the concrete mix design to achieve higher smoothness. It was indicated that when the mix design for a project is approved by the DOT personnel, it usually cannot be changed without the approval from the DOT engineers. It was also indicated that no pavement performance problems have been encountered because of the implementation of a smoothness specification. The general consensus was that the smoothness is mostly affected by the paving equipment and the construction process.