Characterization of Cementitiously Stabilized Layers for Use in Pavement Design and Analysis (2014)

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.

19 C H A P T E R 4 Data for Model Calibration Calibration of the models developed in this study was per- formed using available field data. However, in those situations where the required data was not available, it was estimated using the relationships developed in this research. Durability and Fatigue Data Repeated loads and/or wet–dry and freeze–thaw cycles could damage the CSL located below the pavement surface and result in change in the modulus of the CSL. The modulus value of CSL also increases as a result of continuous hydration and/or pozzolanic reaction. Data were reported in the litera- ture for durability and fatigue tests, including material proper- ties (e.g., 28-day UCS, maximum dry density, and optimum moisture content), field monitoring results [falling weight deflectometer (FWD) modulus values] over time, and traffic information for several locations; these are listed in Table 4-1. Shrinkage Cracking Data Information on material properties—such as water content, density, CSL thickness, and UCS—and climatic conditions— such as the actual local daily average temperature and RH— was collected for the sections used for model calibration. Table 4-2 shows a summary of the field data related to CSL shrinkage cracking. Because very limited field data were avail- able, the results from the restrained shrinkage cracking tests were used for calibrating the shrinkage cracking model. The data used for model development were estimated from field data using the procedures described in the following subsections. Calculation of Crack Spacing and Width Crack spacing and crack width were reported for some sections but only the total crack length was reported for some sections. For these cases, it was assumed that only the trans- verse cracks that extend throughout the entire pavement width existed and there were no longitudinal cracks. The crack spac- ing is calculated as follows: Crack Spacing section length total crack length pavement width (4-1a)= If the crack spacing is smaller than the pavement width, longitudinal cracking is likely to occur and form a block (square) cracking pattern; the crack spacing is calculated as follows: i Crack Spacing section length total crack length 2 pavement width (4-1b)= The crack width is obtained as follows: Crack Width average crack width of th group crack length of th group total crack length (4-2) i i i n∑ = × where n = total number of group For sections with known number of cracks, the average crack spacing is calculated as follows: Crack Spacing section length No. of crack (4-3)= Coefficient of Friction The bond between the CSL and underlying material is typically strong, such that bond failure often occurs in the weak Model Calibration

20 No. Source Section CSL FWD Test Location Traffic at FWD Test Locations 1 Wen et al. (2011) MnRoad Cell 79 (Minnesota) Fly ash (14%) + RAP (8 in.) Mid-Lane No traffic 2 Bloom Consultants, LLC (2007) Parking lot in Milwaukee (Wisconsin) Fly ash (8%) + coal ash (12 in.) Mid-Lane No traffic 3 Parking lot ramp in Milwaukee (Wisconsin) Fly ash (8%) + sandy clay (12 in.) Mid-Lane No traffic 4 Bang et al. (2011) CM1 section, USH 73, Philip (South Dakota) Cement (3%) + soil (8 in.) Mid-Lane No traffic 5 FA1 section, USH 73, Philip (South Dakota) Fly ash (14%) + soil (8 in.) Mid-Lane No traffic 6 Wen et al. (2011) MnRoad Cell 79 (Minnesota) Fly ash (14%) + RAP (8 in.) Mid-Lane Accelerated 7 Romanoschi et al. (2008) Indoor, Kansas State University (Kansas) Cement (7%) + soil (6 in.) Wheel-Path Accelerated 8 Indoor, Kansas State University (Kansas) Fly ash (18%) + soil (6 in.) Wheel-Path Accelerated 9 Indoor, Kansas State University (Kansas) Lime (6%) + soil (6 in.) Wheel-Path Accelerated 10 King et al. (1996) Section 005, Port Allen (Louisiana) Cement (10%) + soil (8.5 in.) Wheel-Path Accelerated 11 Section 006, Port Allen (Louisiana) Cement (4%) + soil (8.5 in.) Wheel-Path Accelerated 12 Section 010, Port Allen (Louisiana) Cement (4%) + soil (12 in.) Wheel-Path Accelerated 13 Bloom Consultants, LLC (2006) County Highway JK, Waukesha County (Wisconsin) Fly ash (8%) + RPM (12 in.) Wheel-Path ADT 5050, 5% truck 14 Si and Herrera (2007) Cement kiln dust section, Amarillo District (Texas) CKD (2%) + local soil (12 in.) Wheel-Path ADT 400 (estimated) 15 Fly ash section, Amarillo District (Texas) Fly ash (8%) + local soil (12 in.) Wheel-Path ADT 400 (estimated) 16 Lime section, Amarillo District (Texas) Lime (3%) + local soil (12 in.) Wheel-Path ADT 400 (estimated) 17 Edil et al. (2003) HW 60, Madison (Wisconsin) Fly ash (10%) + local soil (12 in.) Wheel-Path Design ESAL: 2.6E6 Table 4-1. Sections used for durability and fatigue model calibration. Source Highway Section Soil Binder Underlying Layer George (2001) Highway #302(Mississippi) Section 1A A-2-4 cement (5.5%) lime (4%) treated subgrade Section 3A A-2-4 cement (5.5%) lime (4%) treated subgrade Section 4 A-2-4 cement (3.5%) and fly ash (8%) lime (4%) treated subgrade Section 6 A-2-4 lime (3%) and fly ash (12%) lime (4%) treated subgrade Gaspard (2002) LA 89 (Louisiana) Section 1 A-4 cement (9%) natural silt Section 4 A-4 cement (5%) natural silt Section 9 A-4 cement (9%) natural silt Sebesta and Scullion (2004), Sebesta (2005) TAMU Campus (Texas) 4% dry cure marginal river gravel cement (4%) natural gravel 4% prime cure marginal river gravel cement (4%) natural gravel 4% moisture cure marginal river gravel cement (4%) natural gravel 8% dry cure marginal river gravel cement (8%) natural gravel 8% prime cure marginal river gravel cement (8%) natural gravel 8% moisture cure marginal river gravel cement (8%) natural gravel Monlux and Huotari (2012) Highway #143 (Montana) Section 2001 A-4 cement (8.7%) natural silt Table 4-2. Summary of CSL data.

21 material that lies between the CSL and underlying materials (Romanoschi and Metcalf 2001). Thus, the coefficient of fric- tion for the weak materials between the CSM and underlying material, determined from the direct shear strength test, can be used as Level 1 input. For the Level 2 input, the coefficient of friction can be estimated from the IDT strength, as shown in Figure 3-22. Typical coefficient of friction values for dif- ferent materials in an underlying layer are listed in Table 4-3 (Zhang and Li 2001); these values are used in this study. Age of CSL The age of the CSL is taken as the number of days after construction until shrinkage cracking is observed in the field. Properties of CSL Materials Data on density, water content, binder content, and thick- ness of CSL were obtained from the information reported in the literature. The calcium contents by mass for cement, lime, C fly ash, and F fly ash were assumed to be 63%, 90%, 27%, and 2.7%, respectively (Ramme and Tharaniyil 2004). The field 28-day UCS values were obtained from data reported in the literature. However, when only the 7-day UCS lab values were provided, the 28-day UCS values were estimated using the growth model given by Equation 3-4. Because no values for the COTE of CSM were available in the literature, COTE values were assumed based on the values used in this study. Daily Climate Information Daily climate information, including daily maximum tem- perature, minimum temperature, and average RH, was obtained from www.wunderground.com/history/. IDT Strength and Modulus Growth The cumulative IDT strength/modulus growth was deter- mined using the growth model presented in Equation 3-11. Ultimate Shrinkage Strain and Shrinkage Strain on Top of CSL The ultimate shrinkage was determined from Equation 3-17; the cumulative shrinkage strain of the top CSL at any day was estimated from Equation 3-18. Calibration Procedures Durability and Fatigue Calibration of the durability and fatigue model was per- formed using the following six-step procedure. Step 1, Determination of Number of Freeze–Thaw and Wet–Dry Cycles in the CSL The number of freeze–thaw and wet–dry cycles in the CSL was determined based on MEPDG simulations. Input data for the MEPDG were the pavement structure, material properties, and local climate database. The daily frost depth determined by the MEPDG was then used to calculate the freeze–thaw cycles in the CSL. A freeze–thaw cycle consists of freezing of the bottom of the CSL followed by thawing of top of the CSL. The changes in moisture content of the CSL are not considered in the MEPDG although field measurements have shown significant change in CSL moisture (see Figure 4-1). To determine the number of wet–dry cycles, the CSL is substituted by a granular layer in the MEPDG modeling, and the changes in modulus values resulting from the fluctuation of moisture in the unbound materials are determined. The number of modulus cycles in the unbound layers are taken as the number of wet–dry cycles (see Figure 4-2). Step 2, Prediction of Material Properties for Each Month Other pavement materials, such as HMA and unbound materials (e.g., subgrade), are also affected by climate. The mod- ulus of HMA is affected by temperature; it varies throughout Material in Underlying Layer Coefficient of Friction (psi/in.) Cement-Stabilized 13415 Granular 169 Lime-Treated Clay 146 Clay 22 Table 4-3. Typical coefficient of friction values (Zhang and Li 2001). 0 10 20 30 40 50 2008/1/3 2008/6/1 2008/10/29 2009/3/28 2009/8/25 2010/1/22 M oi st ur e C on te nt (% ) Date Figure 4-1. Moisture variation in CSL of MnRoad.

22 the depth of the HMA layer because of the varying temper- ature throughout the depth. The modulus of HMA also is affected by oxidation that occurs during pavement life. The modulus of the unbound layer is affected by moisture content and/or temperature (e.g., freezing). The MEPDG analysis accounts for the effects of cli- mate on HMA and unbound materials. In this analysis, the asphalt layer and unbound material layer are divided into sublayers; the modulus of each sublayer is varied on a monthly basis. The predicted modulus values for the HMA and unbound materials in each sublayer over the course of the pavement life are used in this analysis. The flexural modulus values of the CSL are estimated from the UCS using Equation 3-7. In addition, the modulus values of the CSL are subject to growth and degradation due to the combined effect of wet–dry and freeze–thaw cycles. The growth models and durability models developed in this study were used to determine the combined modulus values of the CSL on a monthly basis. Step 3, Determination of Pavement Response After determining the modulus values for the different pavement layers/sublayers, the pavement responses were calculated based on the linear elastic layered theory (Maina and Matsui 2004). The traffic load spectrum, traffic wander- ing, and temperature distribution were considered using the procedure contained in the MEPDG. Because the modulus growth model and durability models could not be included in the current MEPDG or the Pavement ME Design software, a MATLAB code was developed to calculate the compressive stress on top of the CSL and the tensile stress at the bottom of the CSL on a monthly basis. Step 4, Determination of Fatigue Life and Damage After the pavement responses were determined, the bottom- up and top-down fatigue life data were determined using the bottom-up tensile-fatigue and top-down compressive- fatigue–erosion models given by Equations 3-13 and 3-16, respectively. The accumulated damage was then calculated for the applied repetitions of each traffic load level and the corresponding pavement response using Equation 3-14. Material properties, such as the MOR and the UCS in the top-down compressive-fatigue–erosion model, are also subject to growth and degradation due to wet–dry and freeze–thaw cycles. The strength growth models and durability models were used to determine the change in strength during pave- ment life on a monthly basis. If the value for the MOR was not readily available, it was determined from the UCS–MOR correlation given by Equation 3-1. Step 5, Calculation of the Modulus after Damage The reduction of the modulus of the CSL caused by traffic repetition was estimated from Equation 3-15 for bottom-up tensile-fatigue and top-down compressive-fatigue–erosion damage. Step 6, Model Calibration The predicted moduli, after considering growth, durability, and traffic effects, were compared with the FWD backcalculated moduli. The parameters in the fatigue models and damage- modulus model were determined by regression analysis. A flow- chart of the durability and fatigue model calibration process is shown in Figure 4-3. Figure 4-2. Typical wet–dry cycles from MEPDG.

23 Shrinkage Cracking Different mechanistic models of shrinkage cracking were evaluated but none was deemed appropriate for use in pavement design and analysis (details are provided in Appendix C). Therefore, shrinkage cracking and width models were developed using dimensional analysis and the data collected. Calibration Results Durability and Fatigue FWD Backcalculated Moduli Ratios A FWD backcalculation evaluation was performed using the deflections of pavements with known layer modulus val- ues (details are provided in Appendix D). The evaluation used the FWD modulus ratio (i.e., the ratio of the modu- lus of each subsequent FWD measurement to the modulus of the first FWD measurement) to reduce the systematic error often associated with FWD backcalculation of layer moduli. Durability Models Because there was only a limited number of sections for which FWD tests were conducted between the wheel-paths (i.e., with no traffic loading), the durability models developed based on laboratory tests were used. Figure 4-4 presents the predicted modulus ratios versus the backcalculated modulus ratios; the results reflect the error associated with FWD back- calculation (details are provided in Appendix D). Fatigue Models The k2 and k3 parameters for the bottom-up tensile-fatigue life models (Equation 3-13) were determined from laboratory fatigue tests on different types of materials (Table 3-6). The parameter that accounts for field adjustment, k1, and the model Figure 4-3. Durability and fatigue model calibration process. Environment: Temperature, Moisture, RH, etc. Initial Material Properties, except CSM Material (not CSM) Moduli over Time; Number of Freeze–Thaw and Wet–Dry Cycles Durability Model Growth Model Material Properties: CSM Modulus and Strength Pavement Response of CSL on a Monthly Basis Tensile stress at bottom, Compressive stress on surface Fatigue Life of CSL: Bottom-up tensile-fatigue, Top-down compressive-fatigue–erosion Damage: Bottom-up tensile-fatigue effect, Top-down compressive-fatigue–erosion effect Modulus after Damage: Bottom-up tensile-fatigue effect, Top-down compressive-fatigue–erosion effect Regress fatigue and damage models parameters to match field FWD modulus Traffic MEPDG CSM Properties

24 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Pr ed ic te d M od ul us R at io Measured FWD Modulus Ratio Line of Equality Figure 4-4. Predicted versus backcalculated CSL modulus ratios. 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Pr ed ic te d M od ul us R at io Measured FWD Modulus Ratio Line of Equality Figure 4-5. Predicted versus measured modulus ratios for fatigue. parameters in the modulus-damage models were determined based on field calibrations. Figure 4-5 shows the predicted modulus ratios versus the backcalculated modulus ratios. Table 4-4 lists the model parameters for growth, durability, and fatigue. Shrinkage Cracking The shrinkage crack spacing and width models for the CSL were developed based on dimensional analysis (Palmer 2007). Because fine and coarse materials are known to show different shrinkage cracking behaviors (Kodikara and Chakrabarti 2001), model calibrations were carried out for the fine and coarse soil separately; these are presented by Equations 4-4 and 4-5, respectively. Figures 4-6 and 4-7 show the predicted versus measured shrinkage crack spacing in the stabilized fine and coarse layers, respectively. Figures 4-8 and 4-9 show the pre- dicted and measured shrinkage crack widths for the stabilized fine and coarse layers respectively. Tables 4-5 and 4-6 present Model Equation Parameter Value Growth 3-4 p1 1.59 p2 1.61 Wet–Dry 3-12 m1 2.58 n1 0.62 Freeze–Thaw 3-12 m1 6.68 n1 0.93 Bottom-Up Tensile-Fatigue Life 3-13 k1 1.07 k2 Table 3-6 k3 Table 3-6 Top-Down Compressive-Fatigue– Erosion Life 3-16 k4 10.85 k5 1.47 Bottom-Up Tensile-Fatigue Modulus Reduction 3-15 m2 3.10 n2 3.99 Top-Down Compressive-Fatigue– Erosion Modulus Reduction 3-15 m2 5.08 n2 2.01 Table 4-4. Durability and fatigue model parameters after field calibration. 0 2 4 6 0 2 4 6 Pr ed ic te d lo g (L ) ( in. ) Measured log (L) (in.) Regression Line Line of Equality Figure 4-6. Predicted versus measured shrinkage crack spacing for stabilized fine layers. 0 1 2 3 0 1 2 3 Pr ed ic te d lo g (L ) ( in. ) Measured log (L) (in.) Regression Line Line of Equality Figure 4-7. Predicted versus measured shrinkage crack spacing for stabilized coarse layers.

25 the model parameters for crack spacing and crack width for stabilized fine and coarse materials, respectively. ( ) ( ) ( )     = µ ρ     ω ρ     ∆ ρ     ε ε   ε        − − log % COTE UCS RH% (4-4) 2 28 2 2 top ult top IDT 9 1 2 3 4 5 6 7 8 L H H t c T H t E S l l l l l l l l t l i i i i i i i i i i i i i i i i i i i i i i i i i i i i % COTE UCS RH% (4-5) 2 28 2 2 top ult top IDT 9 1 2 3 4 5 6 7 8 W H t c T H t E S w w w w w w w w t w ( ) ( ) ( ) = µ ρ     ω ρ     ∆ ρ     ε ε   ε        − − where L = crack spacing, in. W = crack width, in. H = thickness of CSL, in. µ = coefficient of friction, psi/in., lab from Figure 3-21, field from Table 4-3 r = dry density, lb/ft3 t = age when crack survey conducted, days c% = calcium content, % w = water content, lb/ft3 DT = average daily maximum temperature variation, °F COTE = coefficient of thermal expansion, /°F, from Table 3-8 UCS28 = 28-day UCS at 68°F and 100% RH RH = average atmosphere relative humidity, % eult = ultimate drying shrinkage, calculated by Equa- tion 3-17 etop = shrinkage on the top surface, step-by-step calcu- lated by Equation 3-18 with daily environmental conditions SIDT = IDT strength, calculated by Equation 3-11a, psi Et = IDT modulus, calculated by Equation 3-11b, psi li = regression parameters for crack spacing model, i = 1, 2, 3, 4, 5, 6, 7, 8, and 9 wi = regression parameters for crack width model, i = 1, 2, 3, 4, 5, 6, 7, 8, and 9 0 0.1 0.2 0.3 0 0.1 0.2 0.3 Pr ed ic te d W (i n.) Measured W (in.) Regression Line Line of Equality Figure 4-8. Predicted and measured CSL shrinkage crack widths for stabilized fine materials. 0 0.02 0.04 0.06 0.08 0.1 0 0.05 0.1 Pr ed ic te d W (in .) Measured W (in.) Regression Line Line of Equality Figure 4-9. Predicted versus measured shrinkage crack widths for stabilized coarse layers. Parameters Fine Materials Parameters Coarse Materials l1 (<0) −1.19E−01 l1 (<0) 0 l2 (>0) 5.98E−01 l2 (<0) −1.39E−01 l3 (<0) −7.78E−01 l3 (<0) −1.36E−04 l4 (<0) 0 l4 (<0) 0 l5 (>0) 0 l5 (<0) −1.46E−01 l6 (>0) 0 l6 (>0) 2.11E+00 l7 (<0) −2.20E−03 l7 (<0) 0 l8 (<0) −2.53E−01 l8 (<0) 0 l9 8.74E+00 l9 3.85E+00 ( ) contains the parameter constraints. Table 4-5. Parameters for the calibrated CSL shrinkage crack spacing models. Parameters Fine Materials Parameters Coarse Materials w1 (>0) 7.81E−03 w1 (>0) 0 w2 (<0) −1.20E+00 w2 (>0) 0 w3 (>0) 7.67E−01 w3 (>0) 1.34E+00 w4 (>0) 0 w4 (>0) 1.76E−05 w5 (<0) 0 w5 (>0) 3.63E−02 w6 (<0) 0 w6 (<0) 0 w7 (>0) 6.69E−01 w7 (>0) 0 w8 (>0) 4.71E−01 w8 (>0) 5.36E−02 w9 8.63E−04 w9 1.78E−01 ( ) contains the parameter constraints. Table 4-6. Parameters for the calibrated CSL shrinkage crack width models.