Researcher's Guide to The LTPP Layer Thickness Data

Previous | Table of Contents | Next

1. Data Sources
2. Typical Deviations Between the Mean Measured and the Design Thicknesses
3. Analysis Results From Designed Versus Constructed Thickness Comparisons

This chapter summarizes analysis results concerning characterization of the differences in the as-designed and as-constructed (measured) thickness data for the newly constructed SPS layers. Only these new SPS layers have design thickness values accurately documented in the LTPP program. Typical summary statistics for as-designed versus as-constructed mean thicknesses are presented by layer types and target thickness values. Statistical test results are provided regarding whether the mean layer thickness deviations by layer type and design thickness levels are normally distributed. Finally, analysis results are presented from analyzing the percentage of the thickness measurements, as well as for the t-tests results comparing the mean layer thickness and the designed thickness values.

Data Sources

Measured Thickness Data

Two thickness data sources with multiple measurements on a given layer exist in the LTPP database:

• Elevation measurements in SPS*_LAYER_THICKNESS tables for experiments SPS-1, SPS-2, SPS-5, SPS6, SPS-7, and SPS-8.
• Pavement core measurements in testing tables TST_AC01_LAYER and TST_PC06.

According to SPS construction guidelines [30-35], rod and level survey measurements are to be taken at a minimum of five offset locations (edge, outer wheel path, mid-lane, inner wheel path, and inside edge of lane) at longitudinal intervals no greater than 15.2 m (50 ft). Typically, 55 elevation measurements are available for a section.

The number of cores taken for each section depends on experiment and layer type and is defined in the corresponding sampling and testing guides [5-10]. The number of cores per section ranges between 1 and 9.

All sections and layers with available thickness data in either one of these tables were studied to quantify design versus constructed variations in thickness.

For these section/layer combinations, an analysis cell is defined to represent a specific layer for which the target thickness was documented. The following fields in LTPP tables define a unique analysis cell:

• Experiment number.
• Layer type.
• Target thickness.

Design Thickness

For newly constructed SPS layers, the design thicknesses are defined in the corresponding SPS Experimental Designs [11-16]. The designed thicknesses are available for the following layer types:

• SB - AC surface and binder thickness (SPS-1, SPS-5, SPS-6, SPS-8).
• DGATB - Dense-graded asphalt-treated base (SPS-1).
• PATB - Permeable asphalt-treated base (SPS1, SPS-2).
• PCC - Portland cement concrete (SPS-2, SPS-7, SPS-8).
• LC - Lean concrete, (SPS-2).
• DGAB - Dense-graded aggregate base (SPS-1, SPS-2, SPS-8).

The design thicknesses for all these SPS experiments and layer types are presented in tables 15 through 20.

Table 15. Design Layer thickness for the SPS-1 experiment.

Notes:
¹Dense-graded aggregate base.
² Permeable asphalt-treated base.
³ Dense-graded asphalt-treated base.
⁴ Surface and binder (asphalt concrete surface courses).

Table 16. Design layer thicknesses for the SPS-2 experiment.

Notes:
¹Dense-graded aggregate base.
² Permeable asphalt-treated base.
³ Lead concrete base.
⁴ Portland cement concrete slab.

Table 17. Design layer thickness for the SPS-5 experiment.

Notes: ¹Surface and binder (asphalt concrete surface coures).

Table 18. Design Layer thickness for the SPS-6 experiment.

Notes: ¹Surface and binder (asphalt concrete surface courses).

Table 19. Design Layer thickness for the SPS-7 experiment.

Notes: ¹Portland cement concrete slab

Table 20. Design layer thickness for the SPS-8 experiment.

Notes:
¹Dense-graded aggregate base.
²Portland cement concrete slab.
³Surface and binder (asphalt concrete surface cources).

Typical Deviations Between the Mean Measured and the Design Thicknesses

Typical mean layer thickness deviations are established by the following:

• Descriptive summary statistics of the average thicknesses deviations between as-designed and as-constructed values for the layers with the same layer material type and the same design thickness.
• Kurtosis and skewness tests of the distribution of the mean thicknesses for the layers with the same layer material type and the same design thickness.

Descriptive Summary Statistics

Mean layer thickness data for SPS experimental sections with newly constructed layers were obtained from TST_AC01_LAYER and TST_PC06 tables (core thickness), and from SPS*_LAYER_THICKNESS tables (elevation thickness), to compute measured thickness deviation from the design value. The analysis was done for the sets of data grouped by target design thickness, material, and layer type. The following statistical indicators were computed:

• Total number of sections or layers.
• Mean thickness deviation.
• Minimum thickness deviation.
• Maximum thickness deviation.
• Standard deviation of thickness deviation.
• COV of thickness deviation.

The analyses were done separately for the thickness data obtained from core measurements and for the data from elevation measurements. Table 21 summarizes layer thickness deviations by different layer and material types based on an analysis of elevation measurements. Table 22 summarizes mean core examination layer thickness deviations from their designed values by different layer and material types. The following observations are made based on these summary statistics:

• The computed descriptive statistics using elevation measurement data are different from those using core examination data. However, based on statistical analyses, the differences in the mean layer thicknesses and standard deviations are not significant for a majority of the layers. [29]
  
• The mean constructed layer thicknesses for PCC layers and lean concrete base layers are generally above the designed values.
  
• For the same layer and material type, the mean constructed layer thicknesses tend to be above the designed value for the thinner layers, and below the design value for the thicker layers.

These summary statistics characterizing the differences between as-designed and mean as-constructed layer thicknesses can be used as benchmarks for use in pavement design reliability and other research studies

Table 21. Summary of differences between mean elevation thickness measurements and target thicknesses.

Table 22. Summary of differences between mean core thickness measurements and target thickness.

Distribution Type of the Deviations between Designed and Constructed Layer Thicknesses

Table 23. Distribution of the mean thickness deviations from the design thickness based on kurtosis and skewness tests.

As shown in table 23, for the elevation data, eight of the distributions appear to be normal while three are skewed to one side, five to the other side, and one sample tested wide spread but not skewed. The only reasonable distribution (of the differences between mean thicknesses from elevation measurements and target thicknesses) to assume for all material types and thickness is then the normal distribution. Again since the core data are of the same type as the elevation data, it is reasonable to assume normality for these data as well. Supporting the assumption for core data samples is that for the actual tests of skewness and kurtosis nine out of thirteen material types and thickness groups were determined to be normal, while the remaining five groups were skewed and wide spread.

For the cases where distributions were skewed, the reasons were further investigated, and in some cases explanations were discovered by simply examining the data points. For example, figure 4 shows an example of unreasonably large differences from the target thickness for seven PCC layers. Five layers with the largest differences belong to the same SPS project (10-0200). It appears that all core values for these layers are about 8 inch, while elevation data for the same layers show that they are 11 in or thicker. One explanation might be that the incomplete layer thicknesses were obtained during coring.

The following conclusions could be obtained from the analysis of the distribution of the differences between target and measured mean thickness values:

• It is reasonable to assume the mean thickness deviations follow the same kind of distribution;
• There is no trend being skewed to only one side for the elevation data; and
• Majority of the data tested to be normal;

As figure 4 indicates the skewness could be due to unreasonable outlier sections where it is possible that (1) incomplete layer thicknesses were obtained from the core, (2) contractor consistently deviated from the design thickness (under design or over design), or (3) erroneous data exist in the database.

The conclusions drawn from both the descriptive statistics and the kurtosis and skewness tests of their distribution types will be useful for pavement designers and researchers. They will be especially useful in reliability based mechanistic-empirical pavement performance analysis and design.

Figure 3: Chart. Example of normally distributed thickness deviations (elevation data, LC, target thickness 152 mm [6 in]).

Figure 4: Chart. Example of a skewed distribution for layer thickness deviation (core data, PCC, target thickness 279 mm [11 in]).

Analysis Results From Designed Versus Constructed Thicknesses Comparisons

This section presents a summary and conclusions derived from two types of statistical analyses that compared as-constructed versus as-designed thicknesses or target values. Detailed discussion of these comparisons can be found in a report titled, Evaluation and Analysis of LTPP Pavement Layer Thickness Data. [29]

First, both elevation data and core examination data were evaluated to establish the percentage of the individual measurements that is either within or outside the specified values from the target thickness. Three tolerance levels of 6.35 mm (0.25 in), 12.7 mm (0.5 in), and 25.4 mm (1 in) were used for this comparison.

Second, a statistical analysis of the measured mean thickness values versus the as-designed values was performed. Two types of thickness comparisons were performed for both data sources. The two-sided t-test with 95 percent reliability level was used for each section and layer to estimate whether the difference between as-designed and as-constructed thicknesses was significant. The one-sided t-test with 95 percent reliability level was used for each layer for the difference between as-designed thickness and the mean as-constructed thickness and for tolerance levels of 6.35 mm (0.25 in), 12.7 mm (0.5 in), and 25.4 mm (1 in). Analysis results from these comparisons and statistical analyses are presented in the following sections.

Analysis of the Percentage Distribution Results

Based on percentage distributions of the elevation measurements:

• Overall, about 35 percent of the measurements are within ± 6.35 mm (0.25 in) of the target value, with about 30 percent lower than the target and about 35 percent higher than the target value by more than 6.35 mm (0.25 in).
  
  
• Thickness measurements for AC surface and binder layers and thin bonded PCC layers consistently show the highest deviations from the target values.
  
  
• The percentage of the thickness measurements greater than the target value for PCC slab and lean concrete base layers is significantly higher than the percentage of measurements that are lower than the target value. Only 2 percent of thickness measurements are lower, and almost 80 percent are higher than the target value by more than 6.35 mm (0.25 in) for thin PCC bonded layers (76 mm [3 in] and 127 mm [5 in] thick).
  
• Thickness measurements for PATB are more evenly distributed around the target value.

Based on percentage distributions of the individual core thickness measurements:

• Overall, less than 35 percent of core measurements are within ± 6.35 mm of the design thickness value. For some material types and target thickness values, such as thin PCC layers (76 mm [3 in] or 123 mm [5 in] thick) and 203 mm (8 in) thick SB layer, this percentage is below 20.
  
  
• For LC and PCC layers, a much larger percentage of cores have thickness higher than designed. For PATB, the situation is just the opposite.
  
  
• For DGATB, SB, and PCC layers, the percentage of sections with as-constructed thickness below the target value increases with the target thickness. For PCC layers, the percentage of sections with as-constructed thickness above the target value decreases with increasing target thickness.

Statistical Analysis Results

Analysis Results from Two-sided t-tests

Based on the elevation measurements:

• Overall, only about 20 percent of the layers had mean constructed thicknesses not significantly different from their target thicknesses.
  
  
• All 24 sections with 76 mm (3 in) or 123 mm (5 in) target thicknesses for bonded PCC overlays are constructed significantly thicker.
  
  
• For only 4 to 15 percent of the sections with SB layers and target thicknesses between 51 mm (2 in) and 178 mm (7 in), the as-constructed mean thickness is not significantly different from the as-designed thickness.
  
  
• The lowest deviations from as-designed thickness are observed for DGAB layer for which more than 30 percent of the sections have as-constructed mean thickness not significantly different from the target value.

Based on the core thickness measurements:

• Overall, the mean constructed thickness for more than 45 percent of sections/layers is not significantly different from the target thickness. The percentage is highest for DGATB, and lowest for PCC and LC.
  
  
• DGATB has the highest number (overall 61 percent) of sections with mean constructed thickness not different from the target value. For almost 80 percent of the sections with DGATB layer and 102 mm (4 in) target thickness, constructed thickness is not significantly different from the designed thickness.
  
  
• PCC and LC layers have the lowest number (between 34 and 37 percent) of layers with mean constructed thickness not significantly different from the target value. For thin PCC slabs this percentage is 20 or below.

A comparison between analysis results from the elevation and core thickness measurements shows that the percentage of measurements within tolerance limits for all three tolerance levels is approximately the same. However, the percentage of measurements lower than target value is consistently higher for core measurements than for elevation measurements.

Analysis Results from One-sided t-tests

Based on the elevation measurements:

• The AC surface and binder layers have the highest number of sections with the mean constructed thicknesses tested to deviate more than their target values plus or minus all three tolerance levels (6.35 mm [0.25 in], 12.7 mm [0.5 in], and 25.4 mm [1 in]).
  
  
• For most sections (about 70 percent), the mean constructed thicknesses for the dense-graded aggregate base layers are within ±6.35 mm (0.25 in) of their target thicknesses.
  
  
• For portland cement concrete slabs and lean concrete bases, a much higher percent of sections had mean thicknesses greater than the target values plus tolerance levels than the ones below the target values. For thin bonded PCC overlays (76 mm [3 in] and 127 mm [5 in] thick) there are no sections with an as-constructed thickness significantly lower than the target value for all three tolerance levels.
  
  
• For all layer material types except AC surface and binder layers and thin bonded PCC slabs, more than 90 percent of sections have mean layer thicknesses tested within ±25.4 mm (1 in) of their target values.
  

Based on the core thickness measurements:

• The PCC layers have the highest percentage of sections with mean measured thicknesses above their target thicknesses for all three tolerance levels. This percentage decreases with the increased PCC target thickness. For thin bonded PCC layers (76 mm [3 in] or 123 mm [5 in] thick), there are no sections with a layer thickness significantly lower than the target value. For very thin bonded PCC overlays (76 mm [3 in] thick), 80 percent of sections have mean thicknesses that are significantly higher than the target value for more than 6.35 mm (0.25 in). This percentage decreases with increasing target thickness.
  
  
• For all material types except PATB and 178-mm- (7-in-) and 203-mm- (8-in-) thick SB layers, a much larger percentage of layers have mean thicknesses that are significantly higher than designed. For PATB, the situation is just the opposite, with more than 40 percent of layers having values that are significantly lower than the target value for more than 6.35 mm (0.25 in). For 203-mm- (8-in-) thick SB layers, there are no sections with mean measured thicknesses significantly higher than designed.
  
  
• For DGATB and SB layers, the number of sections with the mean thickness below the target thickness increases with the design thickness.
  
  
• All sections with DGATB and LC layers, except one, have thicknesses within ± 25.4 mm (1 in) of the target thickness.

Different conclusions were drawn between the statistical tests performed on the elevation measurements and the statistical tests conducted on the core examination measurements. For example, more core measurement data than elevation measurement data suggest that the measured mean layer thicknesses are not significantly different from their designed thicknesses.

Previous | Table of Contents | Next