Investigating the Relationship of As-Constructed Asphalt Pavement Air Voids to Pavement Performance (2021)

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.

155 validation data: the first dataset using local pavements that fit within the range of the model’s input dataset and the second dataset using pavements that are beyond the model’s range. The first dataset was key to validating the study’s models. It was more likely that the second dataset for validation will be high truck traffic highway pavements and the models’ predicted performance may (or may not) agree with measured performance. The first validation dataset should include 30 to 50 pavement sections representing the full range of climate, traffic, and pavement structures of the agency’s network within the model’s input dataset boundaries. The depth of input data required and the units used to quantify the data for each model may pose a challenge for some agencies. An agency can limit their validation to the critical performance models. For example, northern agencies may consider thermal cracking as critical but southern agencies may disregard it. The validation pavement performance dataset does not require sufficient field measured values to construct the entire performance curve. The validation performance dataset can focus on one or two key increments in time. For example, rutting at year 4 and year 8 and fatigue cracking at 10 years. Once the agency’s local validation dataset is assembled, apply the input data into the model(s) and obtain the predicted performance at the target years. The field measured performance is then plotted against the predicted performance to assess the degree of correlation. If there is reasonable correlation, then the agency can use the study’s models with confidence to examine the influence of as-constructed AV. The second validation dataset of high truck traffic highway pavements can be 10 to 20 pavement sections. Once the validation data is plotted against the model prediction data, the agency can determine if (1) the correlation is acceptable, (2) there is a bias that can be adjusted, or (3) the scatter in the data is significant, indicating that the model does not work for those pavement conditions. Using the Models An agency can select one or more analysis methods to examine the influence of as- constructed AV on pavement performance using input values that represent the agency’s conditions. Each analysis method requires a specific set of input values and the selection of an analysis method(s) may be restricted by the availability of the required input data. If the agency is using this guide to consider revisions to their standard construction specifications, it is recommended that the agency examine a range of common pavements to obtain a better understanding of the influence of as-constructed AV because the degree of influence varies with the climate, traffic, and pavement structure. Changes to as-constructed air void requirements in standard construction specifications should reflect the risk and cost-benefit, which vary by agency and are not addressed in this guide. It is very likely that some of the agency’s existing climate, traffic, and pavement data do not match the analysis input data units used in this guide. An agency will need to collect associated data to compute the required input values. For example, the required traffic input for all three analysis methods is annual ESALs and some agencies will need to collect pavement structure, truck traffic volume, and axle load distribution to compute ESAL values. Once the agency has the required input data, the instructions given below should be followed for each analysis method to insert the data. In all cases, the output will be presented separately for each performance characteristic. Analysis Method 1 involves searching through

156 created charts to match the agency input data to the chart subset input data ranges. Analysis Methods 2 and 3 will generate a family of performance curves using a range of as-constructed AV for each of the four performance characteristics. The agency can use the analysis output data to determine the predicted influence of revising the as-constructed AV construction specification. Analysis Method 1 Subgroup Scatter Plots The outcome from Analysis Method 1 is presented as a catalog of graphs quantifying the level of pavement performance for measured as-constructed AV on a defined subgroup of LTPP sections with common climate, traffic, and pavement structure. The results for new construction and pavement rehabilitation are separate but presented on the same catalog page. The complete catalog of graphs is provided in the appendix. Table 3 describes the required input parameters and subgroup ranges. Table 3. Subgroup Parameters and Ranges Catalog subgroups are divided into four climate zones based on 25-year average annual temperature and 25-year average annual precipitation. An individual agency’s boundaries may encompass more than one climate zone. Climate Zone 1 Dry-Freeze (less than 37 inches precipitation, less than 56°F) Climate Zone 2 Dry-No freeze (less than 37 inches precipitation, greater than 56°F) Climate Zone 3 Wet-Freeze (greater than 37 inches precipitation, less than 56°F) Climate Zone 4 Wet-No freeze (greater than 37 inches precipitation, greater than 56°F) Catalog subgroups are divided into two traffic levels based on 5-year cumulative ESALs from all LTPP sections in the study and the split may or may not compare to an individual agency’s low and high traffic ranges. LOW Traffic (less than 1,000,000 ESALs cumulative over 5 years) HIGH Traffic (greater than 1,000,000 ESALs cumulative over 5 years) Catalog subgroups are divided into three pavement structures based on the SN of the pavement for all LTPP sections in the study; the splits are a compromise of comparable SN ranges and obtaining a sufficient number of sections in each analysis subgroup. THIN Pavement (less than SN=4.4) MEDIUM Pavement (between SN=4.4 and 6.4) THICK Pavement (greater than SN=6.4) For each catalog subgroup, there is a page displaying the influence of as-constructed AV for each performance characteristic based on the study’s LTPP dataset. Table 4 presents the performance criteria used for Analysis Method 1 and Figure 2 presents an example of a display of graphs for one subgroup. Each cell in the figure is color-coded with a subjective interpretation of the scatter-plot data. A green bar signifies that the data met expectation (lower as-constructed AV improved performance), a yellow bar indicates the data had no influence, a red bar indicates the data contradicted expectation, and a grey cell indicates insufficient data. Table 4. Performance Criteria Used for Analysis Method 1 Rehabilitation • Rutting is quantified as rutting depth measured in inches after 4 years of traffic. • Fatigue cracking, which is assumed to be reflective cracking, is quantified as percent of lane cracked after 10 years based on cracks in the 39-inch wheel path. • Thermal cracking, which is assumed to be reflective cracking, is quantified as the number of years after rehabilitation when transverse cracking begins and the amount of transverse cracks measured in lineal feet/mile 3 years after the cracks initially appear. • Ride is quantified as the IRI expressed in inches/mile after 10 years of traffic. New Construction • Rutting is quantified as rut depth measured in inches after 4 years of traffic. • Fatigue cracking is quantified as percent of lane with visible cracking after 10 years based on cracks in a 39-inch wheel path. • Thermal cracking is quantified as the number of years after construction when transverse cracking begins and the amount of transverse cracks measured in lineal feet/mile 3 years after the cracks initially appear. • Ride is quantified as the IRI expressed in inches/mile after 10 years of traffic.

157 CLIMATE ZONE 3 Wet-Freeze > 37 in. annual precip. & < 56 F annual mean temp. TRAFFIC HIGH > 1,000,000 ESALs for 5 years PAVEMENT STRUCTURE MEDIUM 4.4 < SN < 6.4 NEW CONSTRUCTION RUTTING FATIGUE CRACKING RIDE THERMAL CRACKING REHABILITATION RUTTING FATIGUE CRACKING RIDE THERMAL CRACKING Figure 2. Example of Analysis Method 1 Subgroup Summary Graphic

158 Analysis Method 2 Regression Models This section covers the performance models for rutting, fatigue cracking, thermal cracking, and ride, for both new and rehabilitated pavements, developed through regression analysis as part of Analysis Method 2. The first part describes the input variables of the performance models and presents the estimated model parameters. The remainder of the section covers the model equations and includes plots with the predicted performance for a range of as- constructed AV values. Performance Models’ Input Variables Table 5 provides a description and unit of each input variable used in the performance models of the study. As noted in Chapter 1, the models should only be applied within the 5th to 95th range of input values used to develop the model as provided in Table 2. Climate and traffic related variables are time-dependent (indicated with the subscript 𝑦𝑟), while pavement structure attributes were treated as constant within the period of analysis. The AC layers’ attributes were characterized either by using the top AC layer properties–referred to in the list as surface asphalt properties–or computed from all AC layers’ properties–referred to in the list as mean or compound asphalt properties–depending on the performance characteristic under analysis. Further details on the computation, processing, and data sources to generate each variable for model development can be found in the technical report of this study. The study manipulated the input and output values into scaled units as part of developing the regression models. This User Guide revised the regression model input variable parameters so the user can simply apply standard input values without scaling. The input variable PreTRF was included in the User Guide rehabilitation models as a place holder for future study. This value is applied as traffic at time zero. The variable was not a part of model development due to limited available data from the LTPP database and must be studied further. The user can input a value of zero or nominal value representing up to two years of cumulative traffic. The User input variables needed for each performance model are summarized in Table 6. The number of time-dependent annual values required for climate and traffic variables are specific to each model. Table 5. Input Variables Used for Analysis Method 2 Regression Models Variable Description Units AV_s surface asphalt lift as-constructed air voids %Gmm AV_m mean asphalt layer as-constructed air voids %Gmm W lane width inches 𝑇ℎ total thickness of asphalt layers inches 𝑇ℎ total thickness of base and/or subbase layers inches 𝑂𝑣𝑇ℎ total thickness of asphalt overlay inches 𝐸∗ surface asphalt lift mixture dynamic modulus at 130°F and 10 Hz ksi 𝐸∗ compound asphalt layer mixture dynamic modulus at 70°F and 10 Hz ksi 𝐸∗ surface asphalt lift mixture dynamic modulus at 14°F and 0.5 Hz ksi 𝐸∗ surface asphalt lift mixture dynamic modulus at 70°F and 10 Hz ksi 𝑀𝑟 compound resilient modulus of base and/or subbase layers ksi 𝑀𝑟 resilient modulus of subgrade ksi 𝐺𝑟# percent passing sieve #4 of aggregates in surface asphalt lift % 𝐺𝑟# percent passing sieve #200 of aggregates in surface asphalt lift % 𝐵𝐶 surface asphalt lift binder content %mix wt 𝐵𝐶 mean asphalt layer binder content %mix wt 𝑃𝑟𝑒𝑐 total annual precipitation at year 𝑦𝑟 inches 𝑇𝑒𝑚𝑝 annual mean temperature at year 𝑦𝑟 °F

159 Variable Description Units 𝐹𝐼 annual freezing index at year 𝑦𝑟 °F-days 𝑇𝑅𝐹 annual total traffic at year 𝑦𝑟 kESALs 𝐶𝑢𝑚𝐸𝑆𝐴𝐿 cumulative traffic at year 𝑦𝑟 (since opening to traffic date for new pavements, or since assignment date for rehabilitated pavements) computed as 𝐶𝑢𝑚𝐸𝑆𝐴𝐿 = 𝑃𝑟𝑒𝑇𝑅𝐹 + ∑ (𝑇𝑅𝐹 ) kESALs 𝑃𝑟𝑒𝑇𝑅𝐹 cumulative traffic before assignment date kESALs 𝑙𝐶𝑢𝑚𝐸𝑆𝐴𝐿 log-transformed (natural logarithm) cumulative traffic at year 𝑦𝑟 computed as 𝑙𝐶𝑢𝑚𝐸𝑆𝐴𝐿 = 𝑙𝑜𝑔(𝐶𝑢𝑚𝐸𝑆𝐴𝐿 + 10) Table 6. Required User Input Dataset for All Models Input Variable Rutting Fatigue Cracking Thermal Cracking Ride New Rehab New Rehab New Rehab New Rehab AV_s X X X X X X AV_m X X W X X Th_AC X X Xi Xi X X Th_BSB X X X X Xi X OvTh_AC X X Xi E*_IRI and E* FtC X Xi Xi X E*_ThC X E*_Rut X Mr_BSB X X X X X Xi Mr_SG Xi X X X X Gr_(#4) X X Xi X X X X Xi Gr_(#200) X X X X X X BC_s X X Xi X X BC_m Xi Xi Temp_yr X X Xi Xi X Xi X Prec_yr X X Xi X X Xi FI_yr X Xi X X PreTRF X Xi X Xi TRF_yr X X X Xi X X Xi Legend: X = independent model input variables; Xi = model input variables only used for interaction terms. Performance Models’ Parameters Table 7 shows the parameters of each performance model for every combination of performance characteristic (rutting, fatigue cracking, thermal cracking, and ride) and pavement type (new and rehabilitated pavement). Variables %𝐺𝑚𝑚, 𝐸∗, and 𝐵𝐶, in the table refer to the AC as-constructed air voids, dynamic modulus, and binder content, respectively. The specific variables to be used for the corresponding performance model are %𝐺𝑚𝑚 for fatigue cracking and %𝐺𝑚𝑚 for rutting, thermal cracking, and ride; 𝐸∗ for rutting, 𝐸∗ for fatigue cracking, 𝐸∗ for thermal cracking, and 𝐸∗ for ride; and 𝐵𝐶 for fatigue cracking and 𝐵𝐶 for rutting, thermal cracking, and ride. Table 7. Model Parameters for Each Performance Characteristic and Pavement Type Variable Rutting Fatigue Cracking Thermal Cracking Ride New Rehab. New Rehab. New Rehab. New Rehab. (Intercept) -6.65E-01 -5.68E-04 2.71E+00 3.42E+00 1.66E+01 3.79E+00 5.64E+00 4.27E+00 Th_AC -5.26E-03 8.59E-03 - - - 1.59E-01 - -2.37E-02 Th_BSB - - 1.43E-02 -4.61E-02 -7.57E-02 6.68E-02 - -3.45E-03 OvTh_AC - -6.12E-03 - -1.22E-01 - - - - E* -2.48E-04 - -1.35E-04 - -7.66E-04 - - 2.21E-04 Mr_BSB -1.82E-04 -2.52E-04 3.22E-03 - - 1.05E-02 -5.79E-03 - Mr_SG - 6.09E-03 1.33E-01 - - - -4.25E-02 2.49E-02 Gr_4 -4.70E-03 -2.82E-03 - 2.29E-02 3.39E-02 4.11E-02 1.77E-02 - Gr_200 2.21E-02 - -4.95E-01 -1.86E-01 -4.33E-01 - -9.67E-02 -5.78E-02 BC 6.91E-02 6.95E-02 - - - - -1.54E-01 7.60E-02 Temp 9.32E-03 1.97E-03 - - -1.29E-01 - - -8.00E-03 Prec -2.29E-03 -1.87E-03 - - - -9.99E-03 -6.49E-03 - FI 7.09E-05 - - 3.73E-04 - - - -5.27E-05

160 Variable Rutting Fatigue Cracking Thermal Cracking Ride New Rehab. New Rehab. New Rehab. New Rehab. lCumESAL 3.51E-02 -2.30E-02 2.57E-01 - - -4.94E-01 -7.15E-02 - %Gmm 1.34E-02 5.24E-02 -3.63E-03 -3.30E-01 -1.46E+00 8.83E-01 -3.11E-01 -1.42E-02 %Gmm * Th_AC 9.84E-04 -9.38E-04 - - -1.37E-02 -3.43E-02 - - %Gmm * Th_BSB - - -2.93E-03 1.25E-02 1.04E-02 -1.08E-02 - - %Gmm * OvTh_AC - - - 2.09E-02 - - - - %Gmm * E* - - - - - - -9.71E-06 -2.96E-05 %Gmm * Mr_BSB - - - - - -1.93E-03 3.02E-04 - %Gmm * Mr_SG - - -1.74E-02 - - - 4.85E-03 -1.88E-03 %Gmm * Gr_4 5.23E-04 4.73E-04 - -4.63E-03 - -6.31E-03 -2.39E-03 - %Gmm * Gr_200 -2.31E-03 - 6.84E-02 4.10E-02 5.56E-02 - 1.71E-02 1.19E-02 %Gmm * BC -6.60E-03 -9.28E-03 - -6.83E-02 - - 3.29E-02 - %Gmm * Temp - -6.37E-04 - 7.32E-03 1.79E-02 - 1.23E-03 - %Gmm * Prec 2.03E-04 2.20E-04 - - - - 8.89E-04 - %Gmm * FI - -7.03E-06 - - - - - - %Gmm * lCumESAL -1.81E-03 - -3.40E-02 - - - 4.46E-03 - Age -8.33E-03 -1.61E-03 1.84E-01 8.16E-01 -1.91E-01 2.19E-01 8.49E-02 4.72E-03 Age * Th_AC -1.53E-03 - - 1.59E-02 - - - - Age * Th_BSB - - 4.75E-03 -2.17E-03 - - -5.92E-04 - Age * OvTh_AC - - - 8.28E-03 - - - -1.46E-03 Age * E* - - 1.40E-04 -8.66E-05 2.52E-04 - - - Age * Mr_BSB -7.83E-05 1.27E-04 - - - - 1.99E-04 2.04E-04 Age * Mr_SG -2.22E-03 - - - - - - - Age * Gr_4 - -3.52E-04 2.77E-03 -2.90E-03 - - - 4.01E-04 Age * Gr_200 2.61E-03 - 1.81E-02 - - - -2.42E-03 - Age * BC - - -3.57E-02 - -6.13E-02 - - - Age * Temp - - -6.44E-03 -7.58E-03 - - -1.07E-03 -5.64E-04 Age * Prec 6.15E-04 - -1.48E-03 - - - - -1.42E-04 Age * FI 1.15E-05 1.23E-05 - -1.68E-04 - - - - Age * lCumESAL 1.91E-03 5.18E-03 -3.21E-02 -3.29E-02 - - 1.89E-03 2.84E-03 Age * %Gmm - - -1.41E-02 -1.66E-02 - - - -9.05E-04 Use %Gmm_m for fatigue cracking models and %Gmm_s for rutting, thermal cracking and ride models E*_Rut for rutting models, E*_FtC for fatigue cracking models, E*_ThC for thermal cracking models, and Use E*_IRI for ride models Use BC_m for fatigue cracking models and BC_s for rutting, thermal cracking and ride models. Rutting Performance Model The rutting performance model is shown as Equation 1. The parameters for new and rehabilitated pavement models are reported in the second and third columns of Table 7, respectively. The intercept parameter is 𝛽 ; the parameters on each variable 𝑗 are 𝛽 ; the interaction terms between each variable 𝑗 and as-constructed surface AV are 𝛽% , ; the interaction terms between each variable 𝑗 and surface age are 𝛽 , ; and the interaction term between as-constructed surface AV and surface age is 𝛽% , . 𝑅𝑢𝑡 = 𝛽 + ∑ (𝛽 𝑋 , ) + %𝐺𝑚𝑚 (𝛽% + ∑ (𝛽% , 𝑋 , ))+ log (𝐴𝑔𝑒 ) (𝛽 + ∑ (𝛽 , 𝑋 , )) + 𝛽% , log (𝐴𝑔𝑒 ) %𝐺𝑚𝑚 (1) Where: 𝑅𝑢𝑡 = estimated rutting at year 𝑦𝑟; 𝐴𝑔𝑒 = age of pavement surface at year 𝑦𝑟; %𝐺𝑚𝑚 = asphalt surface lift as-constructed air-voids; 𝑋 , = model variable 𝑗 at year 𝑦𝑟; and 𝛽 = model parameter on variable 𝑗.

161 Rutting Performance Model for New Pavements Figure 3 shows the predicted rutting for new pavements between 1 and 15 years of surface age for four as-constructed surface AV values: 3.0%, 4.5%, 6.0%, and 7.5%. The initial traffic value was set to 0 and the annual traffic value was set to 250 kESALs. The remaining variables were set to their median value from the analysis dataset: Th_AC = 5.20 in; E*_Rut = 64.22 ksi; Mr_BSB = 21.49 ksi; Mr_SG = 8.83 ksi; Gr_4 = 55.00 %; Gr_200 = 6.30 %; BC = 4.80%; Prec = 32.30 in; Temp = 58.28°F and FI = 165.20°F-days. Figure 3. Example Influence of As-Constructed Air Voids on Predicted Rutting for New Pavements Rutting Performance Model for Rehabilitated Pavements Figure 4 shows the predicted rutting for rehabilitated pavements between 1 and 15 years of surface age for four as-constructed surface AV values: 3.0%, 4.5%, 6.0%, and 7.5%. The initial traffic value, PreTRF, was set to 1000 kESALs and the annual traffic value was set to 250 kESALs. The remaining variables were set to their median value from the analysis dataset: Th_AC = 8.30 in; OvTh_AC = 3.00 in; Mr_BSB = 22.03 ksi; Mr_SG = 8.83 ksi; Gr_4 = 56.00%; BC = 4.95%; Prec = 39.00 in; Temp = 54.68°F and FI = 221.00°F-days.

162 Figure 4. Example Influence of as-constructed air voids on predicted rutting for rehabilitated pavements Fatigue Cracking Performance Model The definition for fatigue cracking is given in Table 1 and the fatigue performance regression model is shown as Equation 2. The parameters for new and rehabilitated pavement models are reported in the fourth and fifth columns of Table 7, respectively. The intercept parameter is 𝛽 ; the parameters on each variable 𝑗 are 𝛽 ; the interaction terms between each variable 𝑗 and as-constructed mean AV are 𝛽% , ; the interaction terms between each variable 𝑗 and surface age are 𝛽 , ; and the interaction term between as-constructed mean AV and surface age is 𝛽% , . 𝐹𝑡𝐶 = 𝑒𝑥𝑝(−𝑒𝑥𝑝(𝛽 + ∑ (𝛽 𝑋 , ) + %𝐺𝑚𝑚 (𝛽% + ∑ (𝛽% , 𝑋 , )) + 𝐴𝑔𝑒 (𝛽 + ∑ (𝛽 , 𝑋 , )) + 𝛽% , 𝐴𝑔𝑒 %𝐺𝑚𝑚 )) (2) Where: 𝐹𝑡𝐶 = estimated fatigue cracking at year 𝑦𝑟 (wheelpath cracking, 𝑊𝑝𝐶 , for new pavements and reflected fatigue cracking, 𝑅𝑓𝐶 , for rehabilitated pavements); 𝑤𝑝 = wheelpath width = 39 inches (1.0 m) as defined in the 2016 HPMS Field Manual; 𝑊 = lane width, inches; 𝐴𝑔𝑒 = age of pavement surface at year 𝑦𝑟; %𝐺𝑚𝑚 = mean asphalt layer as-constructed air-voids; 𝑋 , = model variable 𝑗 at year 𝑦𝑟; and 𝛽 = model parameter on variable 𝑗.

163 Fatigue Cracking Performance Model for New Pavements Figure 5 shows the predicted wheelpath cracking expressed as percent of lane width for new pavements between 1 and 15 years of surface age for four as-constructed surface AV values: 3.0%, 4.5%, 6.0%, and 7.5%. The initial traffic value was set to 0 and the annual traffic value was set to 250 kESALs. The remaining variables were set to their median value from the analysis dataset: W = 144 in; Th_BSB = 15.90 in; E*_FtC = 1,065.42 ksi; Mr_BSB = 21.56 ksi; Mr_SG = 9.48 ksi; Gr_4 = 56.00%; Gr_200 = 6.10%; BC = 4.62%; Prec = 33.93 in and Temp = 60.08°F. Figure 5. Example Influence of As-Constructed Air Voids on Predicted Fatigue Cracking for New Pavements Fatigue Cracking Performance Model for Rehabilitated Pavements Figure 6 shows the predicted reflected fatigue cracking expressed as percent of lane width for rehabilitated pavements between 1 and 15 years of surface age for four as-constructed surface AV values: 3.0%, 4.5%, 6.0%, and 7.5%. The initial traffic value was set to 1000 and the annual traffic value was set to 250 kESALs. The remaining variables were set to their median value from the analysis dataset: W = 144 in; Th_AC = 8.30 in; Th_BSB = 15.00 in; OvTh_AC = 7.12 in; E*_FtC = 1,047.58 ksi; Gr_4 = 53.00%; Gr_200 = 6.40%; BC = 4.67%; Temp = 53.78°F and FI = 279.50°F-days.

164 Figure 6. Example Influence of As-Constructed Air Voids on Predicted Reflected Fatigue Cracking for Rehabilitated Pavements Thermal Cracking Performance Model The average probable time to crack initiation is 10-13 years combining all LTPP sections used in the study. In contrast, Analysis Method 1 showed that reflective transverse cracking in rehabilitated pavements occurred in 2 to 6 years and thermal cracking in new construction occurred in 10 to 12 years. The performance model of thermal cracking length after crack initiation is shown as Equation 3. The parameters for new and rehabilitated pavement models are reported in the sixth and seventh columns of Table 7, respectively. The intercept parameter is 𝛽 ; the parameters on each variable 𝑗 are 𝛽 ; the interaction terms between each variable 𝑗 and as- constructed surface AV are 𝛽% , ; the interaction terms between each variable 𝑗 and surface age (after transverse crack initiation) are 𝛽 , ; and the interaction term between as-constructed surface AV and surface age (since transverse crack initiation) is 𝛽% , . 𝑇ℎ𝐶 = 𝑒𝑥𝑝(𝛽 + ∑ (𝛽 𝑋 , ) + %𝐺𝑚𝑚 (𝛽% + ∑ (𝛽% , 𝑋 , )) + 𝐴𝑔𝑒 (𝛽 + ∑ (𝛽 , 𝑋 , )) + 𝛽% , 𝐴𝑔𝑒 %𝐺𝑚𝑚 ) (3) Where: 𝑇ℎ𝐶 = estimated thermal cracking length at year 𝑦𝑟 (transverse cracking, 𝑇ℎ𝐶, for new pavements and reflected transverse cracking, 𝑅𝑡𝐶 , for rehabilitated pavements); 𝐴𝑔𝑒 = age of pavement surface at year 𝑦𝑟, in years after transverse crack initiation; %𝐺𝑚𝑚 = asphalt surface lift as-constructed air-voids; 𝑋 , = model variable 𝑗 at year 𝑦𝑟; and 𝛽 = model parameter on variable 𝑗.

165 Thermal Cracking Performance Model for New Pavements Figure 7 shows the predicted thermal cracking for new pavements between 1 and 10 years after transverse crack initiation for four as-constructed surface AV values: 3.0%, 4.5%, 6.0%, and 7.5%. The initial traffic value was set to 0 and the annual traffic value was set to 250 kESALs. The remaining variables were set to their median value from the analysis dataset: Th_AC = 6.76 in; Th_BSB = 14.88 in; E*_ThC = 3,050.11 ksi; Gr_4 = 55.16%; Gr_200 = 6.52%; BC = 4.65% and Temp = 57.58°F. Figure 7. Example Influence of As-Constructed Air Voids on Predicted Thermal Cracking After Crack Initiation for New Pavements Thermal Cracking Performance Model for Rehabilitated Pavements Figure 8 shows the predicted reflected thermal cracking for rehabilitated pavements between 1 and 10 years after transverse crack initiation for four as-constructed surface AV values: 3.0%, 4.5%, 6.0% and 7.5%. Note that the y-axis scale is only one-third of the range compared to new pavements in Figure 6 above. The initial traffic value was set to 1000 and the annual traffic value was set to 250 kESALs. The remaining variables were set to their median value from the analysis dataset: Th_AC = 7.16 in; Th_BSB = 16.96 in; Mr_BSB = 18.84 ksi; Gr_4 = 56.11% and Prec = 33.80 in.

166 Figure 8. Example Influence of As-Constructed Air Voids on Predicted Reflected Thermal Cracking After Crack Initiation for Rehabilitated Pavements Ride Performance Model The ride performance model is shown as Equation 4. The parameters for new and rehabilitated pavement models are reported in the eighth and ninth columns of Table 7, respectively. The intercept parameter is 𝛽 ; the parameters on each variable 𝑗 are 𝛽 ; the interaction terms between each variable 𝑗 and as-constructed surface AV are 𝛽% , ; the interaction terms between each variable 𝑗 and surface age are 𝛽 , ; and the interaction term between as-constructed surface AV and surface age is 𝛽% , . 𝐼𝑅𝐼 = 𝑒𝑥𝑝(𝛽 + ∑ (𝛽 𝑋 , ) + %𝐺𝑚𝑚 (𝛽% + ∑ (𝛽% , 𝑋 , )) + 𝐴𝑔𝑒 (𝛽 + ∑ (𝛽 , 𝑋 , )) + 𝛽% , 𝐴𝑔𝑒 %𝐺𝑚𝑚 ) (4) Where: 𝐼𝑅𝐼 = estimated ride at year 𝑦𝑟; 𝐴𝑔𝑒 = age of pavement surface at year 𝑦𝑟; %𝐺𝑚𝑚 = surface asphalt lift as-constructed air-voids; 𝑋 , = model variable 𝑗 at year 𝑦𝑟; and 𝛽 = model parameter on variable 𝑗. Ride Performance Model for New Pavements Figure 9 shows the predicted ride for new pavements between 1 and 15 years of surface age for four as-constructed surface AV values: 3.0%, 4.5%, 6.0%, and 7.5%. The initial traffic value was set to 0 and the annual traffic value was set to 250 kESALs. The remaining variables were set to their median value from the analysis dataset: Th_BSB = 15.80 in; E*_IRI = 895.47

167 ksi; Mr_BSB = 21.90 ksi; Mr_SG = 9.55 ksi; Gr_4 = 55.00%; Gr_200 = 6.20%; BC = 4.99%; Prec = 35.46 in and Temp = 58.10°F. Figure 9. Example Influence of As-Constructed Air Voids on Predicted Ride for New Pavements Ride Performance Model for Rehabilitated Pavements Figure 10 shows the predicted ride for rehabilitated pavements between 1 and 15 years of surface age for four as-constructed surface AV values: 3.0%, 4.5%, 6.0%, and 7.5%. The initial traffic value was set to 1000 kESALs and the annual traffic value was set to 250 kESALs. The remaining variables were set to their median value from the analysis dataset: Th_AC = 8.30 in; Th_BSB = 15.00 in; OvTh_AC = 3.00 in; E*_IRI = 1,045.63 ksi; Mr_BSB = 20.00 ksi; Mr_SG = 8.42 ksi; Gr_4 = 55.00%; Gr_200 = 6.40%; BC = 4.97%; Prec = 39.74 in; Temp = 53.06°F and FI = 334.40°F-days.

168 Figure 10. Example of Influence of As-Constructed Air Voids on Predicted Ride for Rehabilitated Pavements Using Analysis Method 3 ANN Models Analysis Method 3 covers the performance models for rutting, fatigue cracking, thermal cracking, and ride for both new and rehabilitated pavements developed through the artificial neural network (ANN) analysis. The section covers the input variables required for each performance model and an example using the performance models to predict the influence of as- constructed asphalt pavement AV. Performance Models’ Input Variables Table 8 provides a description and units of each input variable used in the different ANN performance models. As noted in Chapter 1, the models should only be applied within the 5th to 95th range of input values used to develop the model as provided in Table 2. Table 9 identifies what input variables are needed for each performance model. Climate and traffic related variables are time-dependent, indicated with the subscript 𝑦𝑟, while pavement section attributes are treated as constants within the period of analysis. The AC layers’ attributes are characterized by using the top AC layer properties (referred to in the list as surface asphalt properties) or computed from all AC layers’ properties (referred to in the list as average or compound asphalt properties) depending on the performance characteristic under analysis. Further details on the computation, processing, and data sources to generate each variable for model development can be found in the technical report of this study. The study manipulated the input and output values into scaled units as part of developing the ANN models. This User Guide revised the ANN model variable parameters so the user can simply input standard values without scaling. The ANN models create the required scale value for each user input value.

169 Table 8. Input Variables for ANN Pavement Performance Models Abbreviation Description Units t pavement age year T average annual temperature °F P average annual precipitation inch FI average annual freezing index °F-days TR average annual traffic volume kESAL hac total asphalt layer thickness inch hOL asphalt overlay thickness inch Pbw asphalt content of asphalt layer, weighted average %mix wt Pbt asphalt content of top asphalt lift %mix wt AVw as-constructed air voids of asphalt layer, weighted average %Gmm AVt as-constructed air voids of top asphalt lift %Gmm PNo. 4 aggregate passing sieve No. 4 % PNo.200 aggregate passing sieve No. 200 % E130°F, 10Hz dynamic modulus of asphalt mixture at 130°F and 10Hz ksi E70°F, 10Hz dynamic modulus of asphalt mixture at 70°F and 10Hz ksi E14°F, 0.5Hz dynamic modulus of asphalt mixture at 14°F and 0.5Hz ksi hbase base layer thickness inch Ebase composite base modulus ksi Esg subgrade modulus ksi Table 9. Input Variables of ANN Performance Models Input Variables ANN Models Rutting Fatigue Cracking Thermal Cracking Ride new rehab new rehab new rehab new rehab t × × × × × × × × T × × × × × × × × P × × × × × × × × FI × × × × × × × × TR × × × × × × × × hac × × × × × × × × hOL × × × × Pbw × × Pbt × × × × × × AVw × × AVt × × × × × × PNo.4 × × × × × × × × PNo.200 × × × × × × × × E130°F, 10Hz × × E70°F, 10Hz × × × × E14°F, 0.5Hz × × hbase × × × × × × × × Ebase × × × × × × × × Esg × × × × × × × × Rutting Performance Model for New Construction The LTPP Section 01-0105 was selected as an example of a new pavement that illustrates how to use the ANN model to determine the influence of as-constructed AV on rutting performance of an asphalt pavement. Table 10 lists the values of the ANN model inputs for the LTPP Section 01-0105. By inputting these values into the ANN RN (rutting for new construction) model, the corresponding rut depth results are predicted, as shown in Figure 11.

170 Table 10. ANN Model Inputs for LTPP Section 01-0105 Input Variable Value Input Variable Value t 2, 4, 6, 8, 10 AVt 3, 4, 5, 6, 7, 8, 9 T 63.7 PNo.4 56 P 52.4 PNo.200 6.5 FI 59.4 E130°F, 10Hz 52 TR 330 hbase 8.1 hac 4.1 Ebase 32 Pbt 5.2 Esg 12 Figure 11. Influence of As-Constructed Air Voids on Rut Depth for LTPP Section 01-0105 Rutting Performance Model for Rehabilitation The LTPP Section 09-0960 was selected as an example of a rehabilitated pavement that illustrates how to use the ANN model to determine the influence of as-constructed AV on rutting performance of an asphalt pavement. Table 11 lists the values of the ANN model inputs for the LTPP Section 09-0960. By inputting these values into the ANN RR (rutting for rehabilitation) model, the corresponding rut depth results are predicted, as shown in Figure 12. Table 11. ANN Model Inputs for LTPP Section 09-0960. Input Variable Value Input Variable Value t 2, 4, 6, 8, 10 AVt 3, 4, 5, 6, 7, 8, 9 T 49.8 PNo.4 56 P 49.4 PNo.200 3.5 FI 553.1 E130°F, 10Hz 98 TR 658 hbase 24.7 hac 6.5 Ebase 25 hOL 3.2 Esg 6.25 Pbt 4.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0 2 4 6 8 10 12 Pr ed ic te d R ut D ep th (i nc h) Pavement Age (Year) Section 01-0105 New Construction AV=3 AV=4 AV=5 AV=6 AV=7 AV=8 AV=9

171 Figure 12. Influence of As-Constructed Air Voids on Rut Depth for LTPP Section 09-0960 Fatigue Performance Model for New Construction The LTPP Section 05-0115 was selected as an example of a new pavement that illustrates how to use the ANN model to determine the influence of as-constructed AV on fatigue performance of an asphalt pavement. The definition for fatigue cracking is given in Table 1. Table 12 lists the values of the ANN model inputs for the LTPP Section 05-0115. By inputting these values into the ANN FCN (fatigue cracking for new construction) model, the corresponding fatigue cracking results are predicted, as shown in Figure 13. Table 12. ANN Model Inputs for LTPP Section 05-0115 Input Variable Value Input Variable Value t 2, 4, 6, 8, 10 AVw 3, 4, 5, 6, 7, 8, 9 T 60.3 PNo.4 61 P 49.2 PNo.200 6.7 FI 157.5 E70°F, 10Hz 1257 TR 320 hbase 7.4 hac 7.0 Ebase 100 Pbw 4.4 Esg 8 0 0.05 0.1 0.15 0.2 0.25 0.3 0 2 4 6 8 10 12 Pr ed ic te d R ut D ep th (i nc h) Pavement Age (Year) Section 09-0960 Rehabilitation AV=3 AV=4 AV=5 AV=6 AV=7 AV=8 AV=9

172 Figure 13. Influence of As-Constructed Air Voids on Fatigue Cracking for LTPP Section 05-0115 Fatigue Performance Model for Rehabilitation The LTPP Section 09-0902 was selected as an example of a rehabilitated pavement that illustrates how to use the ANN model to determine the influence of as-constructed AV on reflected fatigue performance of an asphalt pavement. The definition for fatigue cracking is given in Table 1. Table 13 lists the values of the ANN model inputs for the LTPP Section 09- 0902. By inputting these values into the ANN FCR (fatigue cracking for rehabilitation) model, the corresponding reflective fatigue cracking results are predicted, as shown in Figure 14. Table 13. ANN Model Inputs for LTPP Section 09-0902 Input Variable Value Input Variable Value t 2, 4, 6, 8, 10 AVw 3, 4, 5, 6, 7, 8, 9 T 50.4 PNo.4 56 P 49.3 PNo.200 3.5 FI 456.3 E70°F, 10Hz 675 TR 361 hbase 24.9 hac 7.2 Ebase 25 hOL 3.8 Esg 6 Pbw 5.0 0 10 20 30 40 50 0 2 4 6 8 10 12 Pr ed ic te d Fa tig ue C ra ck in g (% ) Pavement Age (Year) Section 05-0115 New Construction AV=3 AV=4 AV=5 AV=6 AV=7 AV=8 AV=9

173 Figure 14. Influence of As-Constructed Air Voids on Reflected Fatigue Cracking for LTPP Section 09-0902 Thermal Cracking Performance Model for New Construction The LTPP Section 31-0122 was selected as an example of a new pavement that illustrates how to use the ANN model to determine the influence of as-constructed AV on thermal cracking performance of an asphalt pavement. Table 14 lists the values of the ANN model inputs for the LTPP Section 31-0122. By inputting these values into the ANN TCN (thermal cracking for new construction) model, the corresponding transverse cracking results are predicted, as shown in Figure 15. Table 14. ANN Model Inputs for LTPP Section 31-0122 Input Variable Value Input Variable Value t 2, 4, 6, 8, 10 AVt 3, 4, 5, 6, 7, 8, 9 T 52.7 PNo.4 74 P 26.2 PNo.200 6.9 FI 541.9 E14°F, 0.5Hz 2957 TR 117 hbase 32.4 hac 3.7 Ebase 34 Pbt 5.3 Esg 9.5 0 10 20 30 40 50 0 2 4 6 8 10 12Pr ed ic te d Fa tig ue C ra ck in g (% ) Pavement Age (Year) Section 09-0902 Rehabilitation AV=3 AV=4 AV=5 AV=6 AV=7 AV=8 AV=9

174 Figure 15. Influence of As-Constructed Air Voids on Thermal Cracking for LTPP Section 31-0122 Reflected Thermal Cracking Performance Model for Rehabilitation The LTPP Section 53-6056 was selected as an example of a rehabilitated pavement that illustrates how to use the ANN model to determine the influence of as-constructed AV on reflected thermal cracking performance of an asphalt pavement. Table 15 lists the values of the ANN model inputs for the LTPP Section 53-6056. By inputting these values into the ANN TCR (thermal cracking for rehabilitation) model, the corresponding reflected transverse cracking results are predicted, as shown in Figure 16. Table 15. ANN Model Inputs for LTPP Section 53-6056 Input Variable Value Input Variable Value t 2, 4, 6, 8, 10 AVt 3, 4, 5, 6, 7, 8, 9 T 51.8 PNo.4 60 P 17.1 PNo.200 6.7 FI 207.5 E14°F, 0.5Hz 2739 TR 58 hbase 11.3 hac 8.8 Ebase 20 hOL 11.3 Esg 5.8 Pbt 4.7 0 40 80 120 160 200 0 2 4 6 8 10 12T ra ns ve rs e C ra ck in g (fe et /m ile ) Pavement Age (Year) Section 31-0122 New Construction AV=3 AV=4 AV=5 AV=6 AV=7 AV=8 AV=9

175 Figure 16. Influence of As-Constructed Air Voids on Reflected Thermal Cracking for LTPP Section 53-6056 Ride Performance Model for New Construction The LTPP Section 31-0122 was selected as an example of a new pavement that illustrates how to use the ANN model to determine the influence of as-constructed AV on ride performance of an asphalt pavement. Table 16 lists the values of the ANN model inputs for the LTPP Section 31-0122. By inputting these values into the ANN IRIN (international roughness index for new construction) model, the corresponding IRI results are predicted, as shown in Figure 17. Table 16. ANN Model Inputs for LTPP Section 31-0122 Input Variable Value Input Variable Value t 2, 4, 6, 8, 10 AVt 3, 4, 5, 6, 7, 8, 9 T 52.3 PNo.4 74 P 27.4 PNo.200 6.9 FI 703.4 E70°F, 10Hz 897 TR 117 hbase 32.4 hac 3.7 Ebase 34 Pbt 5.3 Esg 9.5 0 100 200 300 400 500 600 0 2 4 6 8 10 12 Tr an sv er se C ra ck in g (fe et /m ile ) Pavement Age (Year) Section 53-6056 Rehabilitation AV=3 AV=4 AV=5 AV=6 AV=7 AV=8 AV=9

176 Figure 17. Influence of As-Constructed Air Voids on Ride for LTPP Section 31-0122 Ride Performance Model for Rehabilitation The LTPP Section 53-6020 was selected as an example of a rehabilitated pavement that illustrates how to use the ANN model to determine the influence of as-constructed AV on ride performance of an asphalt pavement. Table 17 lists the values of the ANN model inputs for the LTPP Section 53-6020. By inputting these values into the ANN IRIR (international roughness index for rehabilitation) model, the corresponding IRI results are predicted, as shown in Figure 18. Table 17. ANN Model Inputs for LTPP Section 53-6020 Input Variable Value Input Variable Value t 2, 4, 6, 8, 10 AVt 3, 4, 5, 6, 7, 8, 9 T 51.1 PNo.4 61 P 14.5 PNo.200 6.5 FI 230.5 E70°F, 10Hz 558 TR 79 hbase 18.3 hac 8.0 Ebase 20 hOL 2.7 Esg 9.3 Pbt 5.1 0 40 80 120 160 200 0 2 4 6 8 10 12 Pr ed ic te d R id e (I R I i nc h/ m ile ) Pavement Age (Year) Section 31-0122 New Construction AV=3 AV=4 AV=5 AV=6 AV=7 AV=8 AV=9

177 Figure 18. Influence of As-Constructed Air Voids on Ride for LTPP Section 53-6020 Methods Summary Chart The research team used three methods to analyze over 400 LTPP sections to determine the influence of as-constructed AV on pavement performance. A separate analysis was performed on each of four pavement performance characteristics: rutting, fatigue cracking, thermal cracking, and ride. The results of the three analyses indicated the influence of as- constructed AV on pavement performance was mixed. Lower as-constructed AV improved performance (met expectation) under some combinations of climate, traffic, and pavement structure but showed no influence or contradicted expectation for other combinations. Further, the three analysis methods do not always agree. This conclusion supports the understanding that there are numerous major factors that influence pavement performance and as-constructed AV was only a contributing factor. Table 18 is a summary of the findings. An agency must keep a proper perspective about this chart, as it represents pavement performance across the entire country. Performance for Analysis Method 1 was a subjective conclusion based on the trend of each pavement subgroup scatter plot. Performance for Analysis Method 2 was based on the series of performance curves created by regression models. Performance for Analysis Method 3 was based on the series of performance curves created by ANN models. Details regarding the R2 values listed in the table are discussed in the study report. 0 40 80 120 160 200 0 2 4 6 8 10 12 Pr ed ic te d R id e (I R I, in ch /m ile ) Pavement Age (Year) Section 53-6020 Rehabilitation AV=3 AV=4 AV=5 AV=6 AV=7 AV=8 AV=9