LABORATORY FATIGUE MODELS FOR RECYCLED MIXES WITH POZZOLANIC CEMENT AND BITUMEN EMULSION

In recent years, due to technical and economical advantages, the production of pozzolanic cements have considerably extended. In the case of asphalt recycling, using Pozzolonic cements has several advantages. Lower initial stiffness and less shrinkage microcracks than Ordinary Portland Cement (OPC) are some benefits of pozzolanic cements which may reduce the probability of occuring premature cracking in recycled layer. In this reserach in order to investigate the effects of I (PM) pozzolanic cement on the fatigue cracking of recycled mixes with bitumen emulsion and develope fatigue models for these mixes, extensive indirect tensile fatigue and resilient modulus tests were performed at different temperatures and curing times. Test results showed that at high strain levels I (PM) cement specimens have superior fatigue behavior than OPC specimens. Furthermore, the effects of I (PM) cement on fatigue life of recycled mixes related to the initial strain level. Therefore a boundary strain level was determined. Above the boundary strain level, adding cement caused a reduction in fatigue life, whereas below that level the reverse was true. Finally based on laboratory testing results distinct models were established for different boundary strain levels.


Introduction
Use of pozzolanic compounds as a partial replacement with cement can has several technical and economical advantages. These compounds can modify hydrated cement microstructure due to the formation of additional calcium silicate hydrate (C-S-H) phases through pozzolanic reaction between the pozzolan and calcium hydroxide formed during the hydration of cement (Stefanović et al. 2004). Based on extensive research works adding pozzolan to cement as a partial replacement, improves the mix durability by reducing the void content, increases the initial mix workability, increases the ultimate strength and reduces the heat of hydration and volumetric changes (Pekmezci and Akyüz 2004;Stefanović et al. 2004;Aydin and Gül 2006;Bai and Gailius 2009). These properties encourage the use of pozzolanic cements in cold recycling of asphalt with bitumen emulsion. Besides several aforementioned properties, the effect of pozzolanic cements on the fatigue properties of recycled mixes is not clearly understood.
Fatigue cracking is one of the major distresses that influences the service life and quality of pavement. Since fatigue cracking could be initiated at each point of bound layers where the tensile stress exceeds the tensile strength, the fatigue behavior of each bounded material should be well recognized (Tayebali et al. 1992;Ziari and Khabiri 2007).
Cold recycled mixes which usually use as applied thick bound bases act more like slabs (De Beer 1990).
Applying a vertical load on the surface of a slab usually generates horizontal compressive stresses in the upper half of the slab and horizontal tensile stresses in the lower half. The strain resulting from these stresses (particularly the tensile strain at the bottom of the layer) ultimately leads to fatigue failure after too many load repetitions.
The results of field studies on cold recycled mixes with bitumen emulsion (CRME) without cement, have shown that the fatigue life of these layers are equal or even more than conventional hot mixes (Miro et al. 2004;ARRA 2001). Furthermore based on laboratory research works, for an equal air void, cold mixes have lower stiffness and higher fatigue life than conventional hot mixes (Miro et al. 2004; Iowa Highway Research Board and Iowa Department of Transportation 2007; Karlsson and Isacsson 2006;Shu et al. 2008). Fatigue behavior of these mixes have been evaluated as open graded hot mixes in which fatigue life is usually greater than the case of dense graded mixes (Asphalt Institute 1986;Suleiman 2002;SHRP-A-404 1994). Proper viscosity of bitumen emulsion makes it possible to mix cold mixes at ambient temperatures. This will reduce the initial aging of bitumen which may affect the long term flexibility of the mix. Moreover the beneficial effects of some emulsifiers and additives, used in bitumen emulsion (e.g. polymeric additives) were considered to be the reasons for the higher fatigue life compared with hot mixes (Asphalt Institute 1986;Suleiman 2002). Because of the crystalline nature of pozzolanic bonds, lightly cemented materials tend to be brittle, while strongly cemented materials may be prone to shrinkage, cracking and pumping (Suleiman 2002;Rita et al. 2001). In general, however, the addition of cement increases the resilient modulus, shear strength and permanent deformation resistance and decreases the moisture sensitivity but may reduce the ability of the material to sustain repetitive loading (Suleiman 2002;Salomon and Newcomb 2000).
As previously mentioned, there is not enough information about the effect of cementitious additives on the fatigue characteristics of cold mixes. The results of a study on fatigue behavior of cement stabilized cold mixes showed that the relationship between stiffness and fatigue life is related to strain level (Brown and Needham 2000). As reported, at high strain levels, the fatigue life of mixes will be reduced by adding cement (Brown and Needham 2000). This was thought to be due to brittleness of these mixes and occurrence of micro shrinkage cracks. These will cause specimens to fail very quickly once crack initiation occurs (Brown and Needham 2000;Miller et al. 2006).
Due to lack of information regarding fatigue response of cold recycled mixes with bitumen emulsion and pozzolanic cement (CRMEP), in this research, the response of CRMEP was analyzed under different conditions (testing temperature and curing time). The main objectives of this research were: 1. Investigating the fatigue behavior of mixes containing different amounts of pozzolanic cement. 2. Comparison between the effect of pozzolanic cement and Ordinary Portland Cement (OPC). 3. Investigating the effects of curing time, cement content and temperature on resilient modulus and determining the relationship between resilient modulus and fatigue life. 4. Determining the boundary strain levels based on testing temperature. 5. Determining laboratory fatigue models for CRMEP.

Effects of pozzolanic compounds
As previously mentioned, there are some differences between the effects of pozzolanic cements and OPC. Pozzolanic cements are manufactured by incorporating different amounts (usually between 15% to 40%) of natural or artificial pozzolans to OPC clinker. Therefore, the properties of mixes containing pozzolanic cements are completely related to type and amounts of pozzolan. It is important to point out that the pozzolanic reactions, in respect to the OPC hydration reaction, are much slower reactions and occur in a noticeable extent after about two weeks (Stefanović et al. 2004). The mechanism of pozzolanic reaction in pozzolanic cements is not well recognized but the simple shaped of this reaction has been presented in Eq. (1): According to Eq. (1), in order to form a solid product, pozzolanic material require the presence of calcium hydroxide which is a by-product of hydration reaction (Chengzhi et al. 1996;Leung and Balendran 2003). Based on several research results, calcium hydroxide concentration is the main reason of different properties of pozzolanic and OPC cements (Cojbasic et al. 2005). As shown in Fig. 1, in the case of OPC, the concentration of calcium hydroxide which has negligible contribution in strength is steadily increased by increasing the curing time. For pozzolanic cement, at the beginning of curing period, Ca(OH) 2 has low concentration. Therefore, pozzolan particles have no considerable activity and exhibit the behavior of an inert material (Ćojbašić et al. 2005). Due to the increase of (OH) concentration, which is a consequence of the creation of Ca(OH) 2 , the environment becomes more and more alkali. During this period (i.e. after about 7 to 14 days), the pozzolanic reaction starts to develop and as shown in Fig. 1, during the progression of this reaction the calcium hydroxide concentration will continuously reduce (Stefanović et al. 2004;Ćojbašić et al. 2005). This reaction leads to production of the secondary cementitious compositions (e.g. C-S-H). Formation of C-S-H leads to decrease in air void and the final structure of pozzolanic cement mix will be more integrated than OPC. Production of the secondary cementitious compositions along with the reduction in air void have been known as the main reasons of considerable final strength increase for pozzolanic cements (Stefanović et al. 2004;Ćojbašić et al. 2005).

. Aggregates
Recycled Asphalt Pavement (RAP) materials were taken from Tehran West Depot. After crushing the lumps, the materials greater than 25 mm were removed and the gradation of the remaining materials was determined. Fig. 2 shows the gradation.
According to ASTM D2172 procedure the bitumen content of the RAP was determined. RAP materials contained 3.3% bitumen by total weight of the material.

. Bitumen emulsion
A cationic slow setting bitumen emulsion (CSS-1h) was used in the experimental program. The binder and water level of the bitumen emulsion was 60% and 35% of total weight of bitumen emulsion, respectively. Also the bitumen emulsion consisted of about 3.5% solvent and 0.5% HCL.
3.1.3. Cement I(PM) pozzolanic cement was used in this research. This type of pozzolanic cement is a combination of 85% OPC clinker and 15% natural pozzolan (tuff) which consisted of the components reported in Table 1. In this table, the composition of OPC and the physical properties of both cements are also presented. The amounts of cement added to the mixes were 1, 2 and 3% by weight of the mix.

Mix design
Several procedures could be used to determine the optimum bitumen emulsion and water content in CRME design (AASHTO AGC-ARTBA Joint Committee 1998; Mučinis et al. 2009). In this research modified Marshall method was used. Following this method, first, mixtures without cement were prepared containing 3% water (consisting of emulsion water, RAP water content (0.3%) and additional water added to the mixture). Bitumen emulsion was added to the mix at different amounts (varying from 2.5% to 4.5%) at 0.5% increments. The mixes were then compacted by Marshall hammer. Then, samples were cured in oven for 6 hours at 60 °C. While in the molds, cured samples were then kept for 48 hours at room temperature. The specimens were then tested for maximum and bulk specific gravities and Marshall stability. The design criterion for determining the optimum bitumen emulsion specifies the void content within 9-14% (AASHTO AGC-ARTBA Joint Committee 1998). Using this criterion and determining the maximum specific gravity and stability, optimum bitumen emulsion content was assigned to be 4%. In the second step, the optimum moisture content (OMC) was determined for specimens containing different amounts of I (PM) cement at optimum bitumen emulsion content. The air voids of specimens at all cement contents were more than 14%. Therefore, with determining the maximum specific gravity, maximum stability and minimum air void, OMC values were determined. The details of testing results are presented in Table 2. As seen in this table, OMC values were increased as a result of increased I (PM) cement content.

Laboratory testing
The main laboratory program consisted of resilient modulus (M r ) and fatigue tests. Both of these tests were performed using cylindrical specimens applying Indirect Tensile method. The tests were performed at -10, 5 and 25 °C by a UTM-14 testing apparatus. Specimens were tested after 7, 28 and 120 days curing at room temperature of about 25 °C. Before testing, specimens were placed in temperature control chamber of UTM at least for 5 hours until the target temperatures were attained.
The chamber was equipped with a reference specimen containing two thermometers to measure and record the core and skin temperatures. For each I (PM) cement contents (0 to 3% by weight of mixture) 53 specimens were prepared. These specimens were divided into three groups. The first and second groups consisted of 18 cylindrical specimens for indirect tensile strength (ITS) and M r tests. At each condition, ITS was repeated three times and the average of ITS values were reported. M r test was carried out based on ASTM D4123-04 (2004) standard testing method. Specimens were preconditioned by initial load cycles. The minimum preconditioning load cycles should be determined so that the resilient deformations become almost stable. Based on laboratory observations, 100 load repetitions were selected.
This test was performed at above-mentioned conditions at two different stress levels (15% and 30% of ITS) to determine M r values at each condition (testing temperature and curing times).
Fatigue tests were conducted on the third group which for each condition consisted of 27 specimens according to EN-12697-24 standard testing method. Three different stress levels were assumed and similar aforementioned conditions of temperature and curing periods were applied.
The loading frequency was set equal to 0.66 Hz, consisting 0.25s loading and 1.25s recovery time. Loading continued until complete splitting of the specimen.
In ITFT the maximum tensile strain at the center of specimen is calculated using Eq. (2): If the Poisson's ratio (ν ) is assumed to be 0.35, Eq.
(2) will be summarized as in Eq. (3): where d is diameter of specimen (mm); 0 ε is tensile strain ( µε ) at the center of specimen; ∆Η is horizontal deformation, measured using two LVDTs (mm). The results indicate that M r values of CRMEP were increased steadily by decreased temperature and increased cement content and curing period. The relationships between testing temperature and curing time were correlated with exponential and logarithmic functions, respectively. These resulted in high R 2 values (at most cases greater than 0.9) which indicates that unlike the case of CRME the effects of temperature on M r value is more sensible than curing time in CRMEP. Based on several research works, cement considerably accelerates the setting process of emulsified systems (Oruc et al. 2006).
Scanning analysis with electron microscope showed that the cementitious phase in cold mixes is dispersed within the bituminous binder which could have considerable effect on stiffening the organic binder at the initial days of curing (Brown and Needham 2000). Therefore, this phenomenon will reduce the slope of M r increments during the curing period. Due to high temperature sensibility of bitumen emulsion mixes adding cement at high levels (greater than 3%) usually results in a significant decrease in temperature susceptibility of cold mixes (Oruc et al. 2006). With reference to Fig. 3A, the slopes of M r diagrams have not appreciably changed upon adding I (PM) cement. This minor effect of cement on temperature sensibility of studied mixes might be as a result of low bitumen emulsion content compared with conventional cold mixes that usually contain more than 7% bitumen emulsion (Batista 2004;Brown and Needham 2000). Fig. 3C shows the ratio of resilient modulus gained for I (PM) cement specimens to OPC. Based on the results, for 7-day specimens, at most cases the resilient modulus ratio is between 0.75 to 0.95. This ratio increased with increased curing time and after 120 days curing for all specimens the resilient modulus ratio is more than 1.0. According to Table 1, I (PM) cement contains about 15% natural pozzolan. As previously mentioned, at the beginning of curing period the pozzolan particles have no activity and act like filler (Stefanović et al. 2004;Ćojbašić et al. 2005). By increasing the concentration of Fig. 3. CRMEP resilient modulus tests results calcium hydroxide, the pozzolan reaction starts to develop and the secondary cementitious materials will be produced. These compositions which have higher volume than initial structures lead to reduced void content (Stefanović et al. 2004). This phenomenon named pore refinement effect. Production of secondary cementitious compositions and pore refinement after long curing time in pozzolanic cement structures are known as the main reasons for attaining higher stiffness than OPC structures (Chengzhi et al. 1996).

Fatigue
On performing fatigue tests, horizontal deformations were recorded and the deformation diagrams were plotted. A sample fatigue data is shown in Fig. 4. As it can be seen, the diagram of horizontal deformation can be divided into three periods. At the initial load cycles, due to occurrence of plastic deformations, the rate of deformation increment is relatively high. At the second period, the rate of deformation increment gets stabilized and the fatigue curve shows a straight trend. During this period, microcracks which formed during the first and second periods will progress (Thiago et al. 2008). The combination and progress of these cracks lead to complete fracture of specimen.
There are two definitions for fracture life in ITFT method. As it can be seen from Fig. 4A, in the first definition, failure is considered to occur when the constant rate of deformation increase, is replaced by a faster rate. After that point, the microcracks present in the specimen are combined into macrocracks and the specimen is broken into two pieces (Thiago et al. 2008).
According to EN-12697-24 standard, the fracture life shall be determined as the total number of load applications that causes a complete fracture of the specimen. This definition is shown in Fig. 4B. In this research due to less scattered data, the fracture life was considered as in the second definition.  The results of fatigue tests, conducted at 25 °C are presented in Fig. 5. As it can be seen, the slopes of the fatigue lines have decreased by addition of I (PM) cement content. The reduction in slope of the fatigue line with increased cement content is the sign of changing the characteristics from a typical asphalt mixture to a cement treated material.
According to Fig. 5, except for specimens containing 1% I (PM) cement, approximately at above 200 microstrain level, the addition of cement caused a reduction in fatigue life, whereas at below 200 microstrain, the results were reversed.
In order to compare the effects of I (PM) cement and OPC, fatigue lines obtained for specimens tested at 5 °C and 25 °C are shown in Figs 6A and 6B, respectively. As seen in figure 6A, at above 380 microstrain, the fatigue life obtained from specimens with no cement is more than two other groups. Below 380 microstrain, the   (Huang et al. 2005) The results indicated that in comparison with OPC and CRME (i.e. specimens with no cement), the I (PM) cement specimens have the mean fatigue behavior. At all strain levels the fatigue behavior of CRMEP was more similar to CRME. In order to better investigate the obtained results, the toughness index (TI) parameter was determined for compared specimens. This parameter is calculated from the stress-strain curve obtained from ITS test and is a good indication for determining the brittleness of a material. Fig. 7, presents a typical normalized indirect tensile stress-strain curve. The TI parameter is defined as Eq. (4): where TI it is toughness index; A ε is area under the normalized stress-strain curve up to strain ε ; p A is area under the normalized stress-strain curve up to strain p ε ; ε is strain at the point of interest; p ε is strain corresponding to the peak stress.
For an ideal brittle material with no post peak load carrying capacity, TI equals to zero (Huang et al. 2005). In Fig. 8A, the stress-strain curves obtained from ITS test for 28 days specimens at 25 °C is shown. According to this Figure, the CRME specimen has the highest fracture strain. Furthermore, the fracture strain of I (PM) cement specimen is slightly higher than that of OPC. Fig. 8B shows the TI parameter for specimens containing different amounts of cements tested at similar condition (i.e. after 28 days at 25 °C). Based on this figure, the maximum TI parameter was attained for CRME. By increasing the cement content TI parameter was steadily decreased. This indicates the more brittle behavior of specimens containing both cements than CRME specimens. As shown for all amounts of cements the CRMEP specimens have less TI than OPC. Results obtained at other testing conditions (i.e. other temperatures and curing times) were somehow similar to above mentioned condition. Therefore, it could be concluded that there is a meaningful relationship between the toughness index and fatigue behavior of studied specimens. Specimen with higher toughness index (e.g. CRMEP in comparison with OPC) is more flexible and has higher fatigue life at high strain levels.
The results indicated that at heavy loading conditions (high initial strains), I (PM) cement has preferable fatigue characteristics than OPC. Fig. 9 shows the CRMEP fatigue lines at 5 and -10 °C. Based on the results the slopes decreased steadily by decreasing the test temperature. Laboratory observations confirmed that at low temperatures minor differences in specimens or testing conditions (e.g. strain level) leads to very different numbers of cycles to crack initiation. Therefore considerably more specimens were needed to achieve narrow distribution of the results.
The main factors that influenced the resilient modulus of CRMEP in this research were I (PM) cement content, curing time and testing temperature. Presented results show that, by adding cement, while the other parameters are fixed, the resilient modulus of the specimen will be increased and the fatigue line slope will be decreased. The same results were achieved upon the testing temperature was reduced. In contrast, as shown in Fig. 10, comparison of the fatigue life at a constant strain level showed no considerable changes in fatigue line slope with the addition of curing time. This is to some extent comparable to the results obtained from some other similar researches (Batista 2004). Therefore, the results achieved at different curing periods (7, 28 and 120 days) were grouped into one fatigue line in Figs 5, 6 and 9. Similar to the results obtained at 25 °C, the fatigue life of CRMEP specimens at 5 and -10 °C were also related to strain level. According to Fig. 9A, at 5 °C, for an initial strain greater than about 400 microstrain, adding I (PM) cement content will result in a reduced fatigue life. For strain levels below 400 microstrain, the reverse was resulted. With reference to Fig. 9B, at -10 °C similar results were achieved at about 300 microstrain.

Fatigue model
In order to present the fatigue models for CRMEP, first the general relationships between strain -fatigue life and M r -fatigue life were determined. Then based on the obtained relationships, final fatigue models were developed so that the fatigue life of CRMEP could be estimated using the M r and strain level values.

General strain-fatigue life relationship
According to Figs 5 and 9, the general relationship between initial strain and fatigue life will be as presented in Eq. (5) where a and b are constant values; ε is initial strain (microstrain).
For example, as shown in Fig. 5, the above relationship for specimens containing 2% I (PM) cement at 25 °C could be presented by Eq. (6):

General M r -fatigue life relationship
In order to determine the general relationship between M r and fatigue life, at constant initial strains, the curves of fatigue life variation with M r have been drawn. In Fig. 11, the curves obtained for 7 days cured specimens, tested at 5 °C, are shown. Based on these results, at different strain levels, various relationships were obtained. According to Fig. 11A, for initial strains below 400 microstrain, there is a polynomial trend between fatigue life and M r whereas at above 400 microstrain the fatigue life is steadily reduced by M r addition. Similarly for specimens tested at 25 and -10 °C the same trends were achieved at about 200 and 350 microstrain, respectively. Considering the distribution of results at different conditions (testing temperature and curing times), the initial strain of 250 microstrain was chosen as boundary strain level. Based on Eqs (7) and (8), two different relationships were determined at each strain level.
, ε 250µm/m where a, b and c are constant values; M r is resilient modulus (Mpa). For example the above relationship for 7 days cured specimens tested at -10 °C and at initial strain of 600 microstrain, has been presented in Eq. (9): where M r7 is resilient modulus value after 7 days curing (Mpa).
Similarly the final fatigue model for stain levels below 250 microstrain could be presented as in Eq. (11): The final calibrated models, for initial strain levels of above 250 microstrain are presented in Eqs (12) where ε is initial strain (microstrain); M r7 , M r28 and M r120 are resilient modulus after 7, 28 and 120 days curing (MPa). Fig. 12 illustrates the goodness of fit between predicted fatigue life by Eq. (12) and the measured values. As seen the coefficients of determination (R 2 ) for model is above 0.9 and the predicted values from models are equally distributed.
Similarly, the calibrated models for strain levels below 250 microstrain are presented by Eqs (15) to (17) Comparison between predicted fatigue life by Eqs (15) and measured values is shown in Fig. 13. Based on this figure the coefficient of determination indicates good prediction capability of model and the predicted values are equally distributed. In Table 3, the relationships between Measured (N) and predicted (n) values for all models are presented. As reported, all developed models resulted in high R 2 values. Furthermore, the slopes of the N-n relationships are close to 1 and the constant values are acceptable. High constant values for Eqs (16) and (17) are acceptable, because these models are presented for low strain levels which the fatigue life as high as 6 2 10 × loading cycles were achieved for these levels. With reference to the presented models at low strain levels (Eqs (15) to (17)), the fatigue life of CMREP steadily increased by increasing M r value. At high strain levels (Eqs (12) to (14)) the reverse was true and addition of M r led to reduced fatigue life. Therefore, it could be concluded that for pavement with heavy loading conditions, the addition of I (PM) cement content has a detrimental effect on fatigue characteristics of CRMEP.
Various research results have indicated that for CRMEP layers which are mostly used as base layer in a pavement system, strain levels rarely exceed 200 microstrain (Brown and Needham 2000). Therefore it can be concluded that adding cement (up to 3%) to mixes clearly extends their fatigue life.

Conclusions
The main objective of this research was to develop fatigue models for cold recycled mixes with bitumen emulsion and pozzolanic cement (CRMEP). Laboratory experiments, consisted of resilient modulus and indirect tensile fatigue tests which have been carried out at three different temperatures (25, 5 and -10 °C) and curing times (7, 28 and 120 days). Based on the results the following conclusions can be drawn: 1. Adding I (PM) cement content led to a reduction in fatigue line slopes. The least slopes were obtained for specimens tested at -10 °C.
2. There was a meaningful relationship between toughness index and fatigue behavior of studied mixes. Specimen with higher toughness index is more flexible and has higher fatigue life at high strain levels.
3. Initial strain level was identified to be as an important parameter affecting CRMEP fatigue life. At 25 °C and at above 200 microstrain, the addition of resilient modulus caused a reduction in fatigue life whereas for below 200 microstrain the results were reversed. The boundary strain levels for 5 and -10 °C testing temperature were about 400 and 300 microstrain, respectively. 4. Unlike cement content and testing temperature, at a specified strain level, there was no particular relationship between curing time and fatigue curve slope. Therefore the fatigue models were presented separately for each of the curing times.
5. The results indicated that at heavy loading conditions (high initial strains), the fatigue behavior of I (PM) cement is better than OPC.
6. Based on fatigue life-initial strain and fatigue liferesilient modulus relationships the initial strain of 250 microstrain was selected as the final boundary strain level for all testing temperatures. 7. For strain levels of below and above 250 microstrain, distinct fatigue models were presented. According to these models, increased I (PM) cement content resulted in increased fatigue life at low strain levels (below 250 microstrain). Reversely, at high strain levels the addition of I (PM) cement had detrimental effect on fatigue characteristics of CRMEP.