Properties of Foamed Asphalt for Warm Mix Asphalt Applications (2015)

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.

14 C H A P T E R 2 In order to use effectively foamed binder to produce WMA, it is important to understand the characteristics of the foamed binder that influence mixture workability and performance. The four broad steps to achieve this are: 1. Develop a method to precisely quantify the characteristics of the foamed binder. 2. Use this method to investigate the influence of factors such as binder source, water content, liquid additives, and shearing action on the foaming characteristics of binders. 3. Relate the binder foaming characteristics to foamed mix- ture coatability and workability. 4. Use the foamed mixture coatability and workability to estimate the optimum foamed water content and validate the performance of the foamed mixture. A. Test Methods and Metrics to Characterize Binder Foaming This section presents a summary of the work done to achieve the first step (i.e., to develop a method and metrics by which to precisely quantify the characteristics of the foamed binder). A review of the literature shows that a graduated dipstick is commonly used to characterize foamed binders for WMA and base stabilization applications (He and Wong 2006; Namutebi 2011). Most investigators have also regarded the maximum expansion ratio (ERmax) and HL of the foam as meaningful indicators for the quality of the foam (Abel 1978; Brennen et al. 1983; Jenkins 2000; Namutebi 2011; Namutebi et al. 2011). In fact, similar metrics (along with bubble size distribution, as discussed in Chapter 1, Section B) are typi- cally used to characterize foam in other industries, such as the food and polymer industries (Huang et al. 1997; Phillips et al. 1987; Wilde 1996). Some of the limitations of using the dipstick method to measure HL and ERmax are as follows: • The method is highly dependent on the operator as it is based on manual observation of foam height and time. • The approach is limited to only two points in describing the rate at which the foamed binder collapses. • The idea of using HL to describe foaming characteristics of the binder implicitly assumes that the foam collapses following an exponential collapse. Due to the limitations discussed here, the parameters ERmax and HL measured using the dipstick method are not suitable to characterize foamed binder. Instead, three different approaches were considered and evaluated in this study: (1) two different noncontact methods to measure foam expansion and col- lapse over time, (2) an image-based method to characterize foam based on bubble size distribution, and (3) an in-situ density measurement method to characterize foam density over time. Results from these methods were used to develop metrics to characterize binder foam. A.1. Noncontact Measurement of Binder Foam Expansion and Collapse Two different types of sensors were used to measure the change in height and corresponding change in volume of the foamed binder: (1) an ultrasonic sensor and (2) a laser-based sensor. The following sections briefly describe the use of the ultrasonic and the laser-based sensors to measure the change in the height and volume of the foamed binder. The ultrasonic sensor comprises a transmitter and receiver to measure the distance from the sensor to a surface based on time-of-flight measurement. The laser-based sensor com- prises an emitter and detector to measure the distance from the sensor to a reflecting surface based on the phase-shift principle. The main difference between the two methods is that the ultrasonic sensor measures the height of the sur- face by reflecting sound waves over a circular area of about 3.9 in. (100 mm) in diameter, whereas the laser sensor mea- sures the height of the surface by reflecting light of different wavelengths over a very small circular spot of about 1 mm in Research Approach

15 diameter. The ultrasonic sensor was able to collect data more frequently (about 10 points per second) but was susceptible to secondary sound reflections from the sidewalls of the con- tainer if not properly centered. The laser sensor collected data less frequently (about 1 point per second) but was more robust. Note that the limitation in the data collection rate for the laser sensor may be overcome using different hardware operating on the same principle. The following method was used to measure the height and corresponding change in volume of the foam using the two aforementioned sensors. The two sensors were mounted on a tripod and aligned to point directly into a 1-gallon can of binder. The sensors were at least 1 m away from the surface of the can to avoid damage due to splatter from the hot foam- ing binder. A tube was used to enclose the ultrasonic sensor and prevent the sound waves from spreading to a larger area before reaching the container. No such arrangement was nec- essary for the laser sensor. The sensors were then connected to a computer using their respective data acquisition systems. The distance of each sensor from the bottom of the 1-gallon can was measured. Since the bottom of the 1-gallon can was not perfectly smooth but corrugated to improve stiffness, measurements were made to calibrate the weight and volume of the binder to the height of the binder in the can. In order to measure the collapse of the foamed binder, a sample was dis- pensed into a 1-gallon container. The container was immedi- ately removed from underneath the foaming unit and placed under the sensors to measure the height of the foam as it col- lapsed over time. The ER was determined as the ratio of the volume of the foamed binder to the volume of the same mass of the binder without foaming. The volume of the foam (as a function of time) was calculated using the diameter of the can and height of the foam, which was measured using the sensors as it collapsed over time. The same weight of binder as used for foaming was placed into a similar can, and the height of the binder in the can was measured. The height of the binder and the diameter of the can were used to calculate the volume of the binder without foaming. Figure 2-1 shows the setup of the sensors, and Figure 2-2 illustrates the ER for a typical foamed binder using both the ultrasonic and laser sensors. In Figure 2-2, the electri- cal noise resulting from the ultrasonic sensor measurement was filtered. Based on test results collected in this study, both methods were promising in terms of their ability to provide a detailed history of the change in the ER of the foamed binder. However, the method using the laser sensor was preferred for two reasons. First, it requires minimal hardware and software for setup and use. Second, the laser can be pointed into the sampling container without interference from other parts of the foaming unit. This allows measurement of foam forma- tion and collapse as it is being dispensed into the sampling container. A.2. Image-Based Method to Obtain Binder Foam Bubble Size A preliminary method to obtain bubble size from digital images was developed. The methodology was subsequently revised, as will be discussed later. In the preliminary proce- dure, selected images of the foamed surface at different points in time were analyzed using an image processing and analy- sis program, ImageJ, to obtain the size and distribution of the bubble diameter on the surface. The following is a brief description of the preliminary method used to obtain this distribution. A digital camera with a flash pointing directly into the foamed container was used to photograph periodically the Figure 2-1. Ultrasonic and laser sensor test setup. Figure 2-2. Expansion ratio measurements using the two types of sensors.

16 surface of the foamed binder. Due to the spherical nature of the bubbles, light from the flash was reflected strongly at the center of the bubble and along the edges of the bubble. The highlight at the center and edge combined with the low reflection of the curved surface (between the bubble center and edge) created a distinct high-contrast annular ring for each bubble. The outer circumference of this annular ring was used as a measure of the bubble diameter. The image analysis was achieved in three steps. The first step was to con- vert the image to a black-and-white image using ImageJ; this step demarcated the bubble boundaries from the center of the bubble. The second step was to identify individual bub- ble boundaries on the image. This was achieved using Hough transformation, which is an algorithm typically used to iden- tify circular boundaries. The boundaries identified using this transformation were overlaid onto the image, manually com- pared, and corrected for any artifacts. The final step was to use the particle analysis feature in ImageJ to obtain the size (and location) of individual bubbles. Each image was cali- brated with the known internal diameter of the container to convert the image dimensions from pixels to millimeters. This analysis could only be conducted on images obtained after approximately 15 s of foaming, when the large, unstable bub- bles had collapsed and smaller semi-stable bubbles became clearly discernible. An alternative method to obtain the bubble size distri- bution is by using 3D X-ray computed tomography (CT) scanning. However, unlike taking photographic images, CT scanning is a time-intensive technique. Depending on the resolution and specimen dimensions, it can take approxi- mately 30 minutes or more to obtain a single 3D image. The previously described photographic method was preferred in lieu of CT scanning for the following reasons. In order to conduct CT scanning, the foamed binder must be cryo- genically frozen immediately after foaming by immersing a sample container in a liquid nitrogen bath. However, in order to freeze the foam in the shortest time, the diam- eter of the container with the foam sample must be mini- mized. In other words, the foam inside a larger-diameter container takes several seconds to freeze to its core even when immersed in liquid nitrogen. In contrast, the foam inside a smaller-diameter container freezes to its core more rapidly. However, a container with a smaller diameter also influences the quality of the foam since the average size of the bubbles several seconds after dispensing the foam is still on the order of several millimeters, as will be shown in the results later. Considering these two factors, it was concluded that although CT scanning would provide some insights into the bubble size distribution (albeit at a com- promise to the bubble structure itself), similar insights could also be gained by evaluating the bubble size distribu- tion on the surface of the foam. A.3. Measurement of Binder Foam Density In addition to the two methods described previously, the ultrasonic density method was considered for binder foam characterization. This method relies on the use of ultrasonic waves passing through a cross-section of the foamed binder to characterize the foam in real time. As the ultrasonic waves pass through the foam, the wave amplitude is damped. The mag- nitude of damping is directly proportional to the overall foam density (i.e., damping increases as the foam collapses over time). However, this approach was found to be very sensitive to changes in the viscosity of the binder due to small changes in temperature. Consequently, this approach was deemed appropriate when working with larger volumes of foamed sample under tightly temperature-controlled conditions that were less susceptible to temperature fluctuations. A picture of the ultrasonic density test setup is shown in Figure 2-3. A.4. Metrics to Characterize Binder Foam The previous sections presented different methods by which different characteristics of the foam can be measured in real time. The next step is to use these methods to under- stand the process of foam expansion and collapse and to extract a metric or metrics from the measured properties to describe quantitatively the foamed binder. In the case of foamed binder produced by water injection, a foaming unit homogenously combines very fine droplets of water with the binder at elevated temperatures. The water droplets turn into steam that expands and takes the form of bubbles within the binder, resulting in the formation of the foamed binder. The foamed binder has increased work- ability and aggregate coatability at temperatures lower than Figure 2-3. Ultrasonic density test setup.

17 conventional HMA temperatures. The overall expanded vol- ume and stability of the foam are important for WMA appli- cations. It is hypothesized that foam with a higher expansion has a lower overall viscosity and easily coats aggregate par- ticles. In addition, foam with a lower collapse rate has a longer effective time to coat the aggregate particles. Jenkins (2000) also developed an exponential collapse model (Equation 2-1) for binder foam as a function of time, ERmax, and HL. This model was adapted from the isotope col- lapse model: ( ) = − ∗ ER ER (2-1)max ln 2 HLt e t p Where: ER(t) = foam volume at time t to the volume of the binder after the foam collapses. t = time starting from end of foam dispensing. The parameters HL and ERmax were measured using a dipstick. As noted before, the use of a dipstick to measure these parameters is highly dependent on the operator since it is based on manual observation of foam height and time. This approach is also limited to only two points in describ- ing the rate at which the foamed binder collapses. Further- more, observations made during this study show that HL is typically in the range of 1 to 4 s; consequently, manual mea- surements of HL are prone to high variability. Data collected in this study using the more precise and faster methods (as described in the previous sections) to measure expansion also indicate that the binder foam does not collapse following an exponential collapse form. The following two mechanisms of bubble growth and col- lapse are proposed based on a review of such mechanisms for binder and other materials in the literature (Saye and Sethian 2013; Schick 2004; Schramm 1994; Sunarjono 2008) as well as observations made during this study. A.4.1. Unstable and Short-Life Bubbles Initially, foamed binder consists of a cluster of bubbles sep- arated by a thin layer of liquid binder (as shown in Figure 2-4), as confirmed by Hailisilassie et al. (2014). The thickness of the liquid binder layer is a function of ERmax. As ERmax increases, the ratio of liquid to gas (steam) volume decreases, resulting in a decrease in the initial thickness of the binder film layer. Unstable bubbles cause a sharp decrease in the volume of the foam in a few seconds. The following are the possible mecha- nisms of collapse of the unstable bubbles: 1. Liquid flow: Immediately after the foam is dispensed into a can, the liquid binder flows down through the inter- connected network of channels between the bubbles. At the same time, the bubbles also move up due to the buoyant force. As the liquid binder moves down and the bubbles move up, the film layer thins out, and finally the bubbles collapse. As the bubbles on the surface collapse, the liquid between these bubbles redistributes to the nearby bubbles. 2. Excessive steam pressure: In the case of larger water drop- lets (or coalescence of many fine droplets) that turn into steam, the vapor pressure inside the bubble causes the bubble diameter to grow rapidly to a point where the bubble becomes unstable and collapses. 3. Drop in the temperature of steam: For bubbles that are in direct contact with the atmosphere, the temperature of the steam may drop and cause the foam to collapse (implode). 4. Rising velocity: Larger bubbles rise to the surface at much higher speeds (proportional to the square of the diameter) and ultimately collapse. As will be shown later, this phe- nomenon explain why foams with higher water contents and higher expansion ratios are typically less stable than foams with lower water contents and expansion ratios. A.4.2. Semi-Stable and Long-Life Bubbles As the unstable bubbles collapse over time, the volume of liquid binder that separates the bubbles from each other increases. The increase in liquid binder volume, by itself, increases the relative stability of the bubbles. In the case of smaller water droplets that turn into steam, the vapor pres- sure inside the bubble causes the bubble diameter to grow rapidly, as before. However, the bubble diameter reaches an equilibrium size since the surface forces of the bubble bal- ance the internal pressure of the steam and allow the foam to remain stable for a finite amount of time. The equilibrium bubble diameter or radius was given by Laplace, as shown in Equation 2-2 (Pellicer et al. 2000). As before, such bubbles migrate toward the surface of the binder. However, the bubble velocity is much lower due to the smaller size, and the ratio Figure 2-4. Stability and shape of binder foam as function of time.

18 of liquid binder to air (steam) volume is higher, resulting in an increase in the shell thickness of the individual bubbles. The relationship between bubble velocity and bubble diameter is given by Stokes’ law in Equation 2-3 and clearly shows that the bubble velocity is directly proportional to the square of the bubble diameter and inversely proportional to the viscosity of the fluid. Finally, when such bubbles reach to the surface of the binder, the shell thickness decreases due to liquid binder flow, and the vapor pressure inside the bubbles reduces due to cooling, triggering an unstable reduction in the bubble diameter and collapse. − = γ2 (2-2)bubble atmP P R Where: Pbubble = Pressure inside the bubble (Pa). Patm = Atmospheric pressure (Pa). g = Surface tension of the binder (N/m). R = Bubble radius (m). 18 (2-3)D V gf b( )= µ ρ − ρ Where: V = Rising velocity of the bubble (m/s). r = Density of the binder and the bubble (rf, rb) (Kg/m3). m = Viscosity of the binder (Pa.s). In summary, the largest bubbles that contribute most to the expansion of the binder are also the most unstable and short-lived. This effect was observed during the first few sec- onds as the foamed binder collapsed, and was exaggerated at higher water contents (i.e., higher air to liquid binder ratio) that are likely to result in larger water droplets and larger bubbles (see Figure 2-5). The collapse of the larger unstable bubbles was followed by a gradual rise and collapse of rela- tively less unstable or semi-stable smaller-diameter bubbles. This effect was relatively clear at lower water contents where bubbles continued to rise and collapse with the smaller bub- bles taking the longest to rise and collapse (see Figure 2-6). Based on the aforementioned understanding, the follow- ing approach was used to characterize the foamed binder. The exponential collapse model developed by Jenkins (2000) does not reflect the initial sudden collapse of foams in the first few seconds. Based on experimental observations, the following function was found to fit best the data obtained from testing the different binder, water content, and foaming equipment combinations: ER 1 ER 1 (2-4)maxt ae a ebt ct( )( ) = + + − −− − Where: ER(t) = The expansion ratio at any time t, a, b, and c are constants. ERmax = The maximum expansion ratio that was directly measured during the foaming process. Based on the form of this equation, it may be tempting to conclude that the overall collapse observed in the foam is the sum effect of two different collapse processes proceeding at different rates. However, Equation 2-4 was used only for data reduction purposes, and it is inappropriate to interpret the phenomenon purely based on this mathematical form. In fact, observations of the foam collapse process suggest that it may be more appropriate to model the collapse in the semi-stable foam after the first few seconds of foaming when the foam is in the semi-stable stage. This will be discussed in more detail. Figure 2-5. Binder N6 with 3.0% water content.

19 Figure 2-5 and Figure 2-6 illustrate typical ER measure- ments versus time for a particular binder (denoted as N6) with 1.0% and 3.0% water content. The discrete points illus- trate the raw data, and the continuous line shows the fit. The expansion at the high water content shown in Figure 2-5 is extremely short-lived, and the collapse of foam is faster as compared to the lower water content presented in Figure 2-6, where the foam collapse is slower. The figures also illustrate a digitally inverted image of the surface of the foamed binder with bubbles at different points in time. (Bright annular bands indicate locations of bubbles.) As discussed before, the foamed binder with 3.0% water content initially expands much more than the foamed binder with 1.0% water content does. However, the foamed binder at the lower water content is more stable or semi-stable and clearly shows the migration and collapse of bubbles at the surface with the diameter of the bubbles decreasing with time. Selected images of the foamed surface at different points in time were analyzed as explained in Section A.2 using an image processing and analysis program, ImageJ, to obtain the size and distribution of the bubble diameter on the sur- face. Figure 2-7 and Figure 2-8 show the size distribution of bubbles at three different points in time on the surface of the N6 binder foamed with 3.0% and 1.0% water content, respectively. The figures show that the bubble diameter at the surface becomes smaller as the foam continues to collapse Figure 2-6. Binder N6 with 1.0% water content. Figure 2-7. Change of bubble size distribution with time for binder N6 with 3.0% water content.

20 over time. This is consistent with what can be expected from Equation 2-3, in that the larger-diameter bubbles rise to the surface faster. In addition, the effect of water content and time on the mean diameter of the bubbles is clearly visible from the two figures. These observations substantiate conclu- sions made by Ozturk and Kutay (2014a, 2014b). The proposed standard method for measuring expansion/ collapse via laser-based sensor is presented in Appendix B. Based on preliminary results, the following metrics were con- sidered to characterize foamed binders: 1. Maximum expansion ratio – ERmax: The ratio of the maxi- mum volume occupied by the foamed binder to the vol- ume occupied by the same mass of the binder without any water or foam in it. 2. Rate of collapse of semi-stable foam – k-value: Based on the results obtained thus far, a significant portion of the unstable bubbles collapsed during the first few seconds after foaming. The rate of collapse of the semi-stable foam was determined as the parameter k obtained by fitting the ER versus time to an exponential curve: ER(t) = 1 + ce-kt. Note that the ER after 10 s of foaming was used to deter- mine this parameter. 3. Foamability Index – FI: The area under the ER curve at selected times (Jenkins 2000). 4. Bubble size distribution: Bubble size distribution from digital image acquisition and analysis was also consid- ered as an important metric to characterize binder foams. Bubble size distribution through image acquisition and analysis was used in some cases to better understand the mechanisms that drive foam expansion and collapse. A revised image-based method to obtain bubble size, includ- ing the ratio of the surface area of the foaming bubbles at selected times to the surface area of the binder without any water or foam in it, or SAI, is presented in Appendix B. A similar approach based on bubble size distribution cal- culated from Stokes’ law was described by Ozturk and Kutay (2014b). Note that the HL for most combinations of binder, water content, and foaming equipment was in the range of 1 to 4 s. This parameter was not very repeatable and was associated with the more turbulent collapse of larger unstable bubbles in the foamed binders. Consequently, this metric was not used to characterize foamed binders. The slope of the expan- sion ratio at selected times and the drop in expansion ratio between selected times were also explored as alternative met- rics but did not provide meaningful trends. The FI proved to be a good indicator of the differences in binder foaming characteristics, and SAI was able to provide a link between the bubble size distribution (obtained from dig- ital image analysis) and the expansion ratio (obtained from laser measurement analysis) results. The applications of these two metrics are discussed in Chapters 3 and 4. B. Test Methods and Metrics to Characterize Foamed Mixtures One of the major unknowns in the application of foamed binders for WMA is that, to date, there have been no estab- lished relationships between binder foam characteristics to mixture workability, coatability, or performance. Foaming is Figure 2-8. Change of bubble size distribution with time for binder N6 with 1.0% water content.

21 intended to improve mixture workability and coatability from reduction in binder viscosity due to the binder volume expan- sion. Therefore, it was important to focus on the evaluation of workability and coatability of the foamed asphalt mixtures. Most of the research conducted so far has been on binder foam characterization or mixture evaluation. As previously mentioned, Jenkins (2000) suggested using the area under the ER versus time curve to determine the optimum foaming water content. In this procedure, the optimum water content corresponds to the point where the area under the ER curve is at a maximum when plotted against water content. However, other studies found that the area under the ER curve versus water content did not have a maximum point for some binder foams (Sunarjono 2008). Also, the area under the ER curve was not compared to mixture workability or coatability. Con- sequently, the impact of change in binder foam characteris- tics on mixture workability and coatability has been unknown until this study. Workability of asphalt mixtures is a property that describes the ease with which the mixture can be placed, worked by hand, and compacted. It is a function of temperature, binder properties (e.g., viscosity, grade, polymer modification), and aggregate properties (e.g., size, angularity), among other fac- tors. Coatability of asphalt mixtures is defined as the degree of coating of the aggregates by the binder. This parameter is important to the performance of asphalt mixtures since well- coated aggregates are likely to have a stronger bond between the particle and the binder and, thus, a better resistance to moisture damage and other distresses. The proposed method to evaluate workability of foamed WMA is based on the Superpave gyratory compactor (SGC) compaction data (i.e., shear stress versus number of gyra- tions). Densification of asphalt mixtures can be considered as a function of (1) reorientation of aggregate particles and (2) distortion due to the flow of the binder. Thus, resistance to shear in the asphalt mix is provided by the internal friction from a combination of the angularity and hardness of the aggregate and the cohesion provided by the asphalt binder (De Sombre et al. 1998). The maximum shear stress is pro- posed as a mixture workability parameter; mixtures with lower maximum shear stress are considered more workable. This method is based on work done at the University of Min- nesota by De Sombre et al. (1998) in quantifying the work- ability of asphalt mixtures with various aggregate gradations and binders to define optimum compaction temperatures. For coatability, an alternative method to AASHTO T 195 was explored. The AASHTO method is a very subjective approach to quantifying the asphalt coating of coarse aggre- gate. It relies on a visual assessment of the aggregate coating and may be subject to variability from a number of possible sources, including operator experience, aggregate coloring, and lighting. In this study, a modified procedure based on the aggregate absorption method originally developed by Velas- quez et al. (2010) was used to measure mixture coatability. The method is based on the assumption that a completely coated aggregate submerged in water for a short period (i.e., 1 hour) cannot absorb water because water cannot penetrate through the binder film surrounding the aggregate surface. On the other hand, a partially coated aggregate is expected to have detectable water absorption since water is able to pen- etrate and be absorbed by the uncoated particle. The coatabil- ity index (CI) is proposed as the mixture coating parameter. The proposed protocols for workability and coatability are detailed in Appendix B. C. Initial Foamed Mix Design Approach An initial foamed asphalt mix design method was developed to include the use of a laboratory unit to determine the foam- ing ability of the binder. As will be detailed later, laboratory foaming experiments indicate that certain binders have little foaming ability, possibly due to the presence of anti-foaming agents introduced during the crude refining or binder pro- duction process. Therefore, the development of a foamed asphalt mix design procedure that involves the validation of binder foaming ability was considered essential. As shown in Figure 2-9, after the binder foaming ability has been corrobo- rated, a traditional hot mix design procedure (i.e., AASHTO R 35) is performed to determine the optimum binder content. Afterward, the optimum foaming water content is estab- lished using the workability and coatability protocols described in the last section and Appendix B. Experimental results indi- cate that a considerable difference in foamed mixture work- ability and coatability can result from minor changes in the foaming water content (i.e., 0.5% by weight of the binder); thus, determination of the optimum foaming water content is essential. In the proposed procedure, the water content that yields the lowest maximum shear stress with a CI at least equivalent to the CI of HMA is considered the optimum. The last step in the proposed mix design method is to eval- uate the performance of the unfoamed and foamed asphalt mixtures at optimum water content. Standard tests, includ- ing MR per ASTM D7369, IDT strength per AASHTO T 283, and HWTT per AASHTO T 324, are suggested for perfor- mance evaluation. For HWTT, alternative parameters to the traditional rut depth at a specific number of load cycles are suggested, as detailed by Yin et al. (2014). If all performance parameters, namely wet IDT strength and tensile strength ratio (TSR), dry MR, HWTT stripping number, HWTT remaining life, and HWTT viscoplastic strain at stripping number, com- ply with established specifications, the mixture is accepted. Otherwise, changes in material selection or other mixture components should be considered and the mixture retested.

22 Figure 2-9. Proposed foamed asphalt mix design method.