Coefficient of Thermal Expansion of RCA Concrete Made by Equivalent Mortar Volume

The present study was conducted to experimentally verify if the coefficient of thermal expansion (COTE) of recycled aggregate concrete is proportional to the volume of the original virgin aggregate in the total recycled aggregate concrete mix. Three types of recycled concrete aggregate (RCA) were crushed from: railroad concrete sleepers; precast (PC) culverts; commercial recycling plant. RCA concretes were mixed using two concrete mixing methods: conventional mix method and equivalent mortar volume (EMV) method. And by varying the replacement ratio, three test series were made. Test results showed that at the same RCA replacement ratio of 68%, the COTE of RCA concrete prepared by the EMV mix design was over 6–7% lower than that of RCA concrete made with the conventional mix method. It was also similar to or 1–2% lower than that of the natural coarse aggregate concrete. This may be because the conventional mix method does not take into account the residual mortar content attached to RCA. This results in a decrease in the volumetric ratio of the original virgin aggregate and a relative increase in the volumetric ratio of the mortar (or cement paste).


Introduction
Cracks in concrete structures are caused largely by the load, but they may also be caused by material characteristics such as creep or environmental factors such as temperature and humidity. In particular, curling occurs in concrete pavement due to changes of temperature and humidity. In concrete overlay pavement construction, when the two materials have different coefficients of thermal expansion (COTE), delamination may be caused at the interface between existing concrete pavement and overlaying material due to thermal instability [1,2]. To prevent that, the Texas Department of Transportation sets the upper limit of COTE of concrete in its specifications. Continuous reinforced concrete pavement (CRCP) is often adopted as a form of concrete pavement, and the COTE of concrete is known to have an impact on the initial behavior of the CRCP [3]. Since the COTE of river gravel is higher than that of limestone by 30% on average, initial cracking can easily occur in CRCP prepared using river gravel [4].
Due to the recent impact of global warming, blow-up of road pavement is often reported in Korea. The Korean government has promoted for 2 years a study to analyze the causes of blow-up and to prepare countermeasures [5]. As possible causes of blow-up, it is analyzed whether the joint space of 6 m is too wide compared to the current concrete pavement thickness (300 mm) or whether the problem comes from the pavement mix [5,6]. Unfortunately, the efforts made by scholars to apply high-quality recycled aggregates to marble and limestone. Zhou et al. [24] predicted the COTE of concrete by varying the shape of aggregate (circle, ellipsoid, square, rectangular) and using a self-consistent method; however, the effect of aggregate shape was found to be almost negligible.
Hansen [26], ACPA [27] and NCHRPP [28] reported that the COTE of concrete increased to 30% with the increase of RCA replacement ratio. The test, performed with core specimens taken from 5 test sections of the Transportation Departments of four states in the United States, showed that the COTE of RCA concrete specimens was retained in 1 section but increased in 4 sections up to 9-23% [29,30]. Sadati and Khayat [31] reported that when RCA was used at a ratio of 40%, the COTE of the concrete increased to 10%. Because of this, paving with RCA concrete by the Transportation Department was discontinued in some states in the United States [29,31]. Our previous study also showed that the COTE of concrete with 100% RCA replacement ratio increased by 5-9% in comparison with that of reference concrete containing natural aggregates [25].
On the contrary, according to Won [32,33], concrete prepared by using RCA at a replacement ratio of 100% in which silicious river gravel was used as the original virgin aggregate showed a COTE that was similar to the COTE of the natural aggregate concrete in actual road sites. Khayat and Sadati [34] reported that RCA concrete of which replacement ratio was within 15-40% did not show a significant difference in the COTE compared to the conventional concrete. Bekoe et al. [35] also reported that the COTE of concrete prepared by using RCA remained similar at a replacement ratio up to 50%.
On the other hand, Smith [36,37] reported that the COTE decreased by up to 40% as RCA replacement ratio increased by 50%, and explained that the cause of the decrease was the use of high-quality RCA (with limestone as the original virgin aggregate). Pickel [38] showed that the COTE is dependent on RCA replacement ratio and the degree of RCA saturation; when RCA was fully saturated and the replacement ratio was 30%, the COTE of concrete was considerably reduced. According to Wen et al. [39], the COTE of pavement concrete mix decreased by 10% when RCA replacement ratio was 45% and fly ash was used at a ratio of 20%. Therefore, the present study was conducted to show that the COTE of RCA concrete is proportional to the volumetric ratio of the original virgin aggregate in the total mix. Three types of RCA (one crushed from railroad concrete sleepers; one crushed from precast (PC) culverts; and one obtained from a commercial recycling plant) were mixed using two concrete mix designs (conventional mix design and equivalent mortar volume (EMV) mix design) and by varying the replacement ratio relative to that of natural coarse aggregate, which was used as a reference. The experiment was performed in 3 test series. Even if the replacement percentage of natural coarse aggregate with RCA is the same among the coarse aggregates included in a certain mix, the volumetric ratio of the original virgin aggregate is larger in mixes prepared by modified EMV mix design than in those prepared by the conventional ACI mix design method. Therefore, one of the main purposes of the present study was to examine the difference of the COTE between the two mixes.

RCA Productions
Three types of RCA were used in the present study: 'RR' from Railway PC sleepers ( Figure 1a); 'RP' from PC culverts (Figure 1b); and 'RA' from the IS recycling plant in Korea. The PC sleepers, having a compressive strength of about 35 MPa, were manufactured according to the Korean Railway Standards [40] by steam curing at a maximum aggregate size of 20 mm and water-cement ratio of 35% or lower. The PC sleepers were then crushed in a plant into 'RR', having a maximum aggregate size of 20 mm. The second type, 'RP,' was obtained from crushing PC culverts having a maximum aggregate size of 20 mm and a compressive strength of 35 MPa. 'RA,' obtained from a recycling plant, was prepared by crushing old concrete from the same source into a size of 20 mm. It should be noted that the old railway PC sleepers were used about for 20 years, while the old PC culverts were laid out in the open area at best for a year and old concrete structures for 'RA' more than 20 years.

RCA Productions
Three types of RCA were used in the present study: 'RR' from Railway PC sleepers ( Figure 1a); 'RP' from PC culverts (Figure 1b); and 'RA' from the IS recycling plant in Korea. The PC sleepers, having a compressive strength of about 35 MPa, were manufactured according to the Korean Railway Standards [40] by steam curing at a maximum aggregate size of 20 mm and water-cement ratio of 35% or lower. The PC sleepers were then crushed in a plant into 'RR', having a maximum aggregate size of 20 mm. The second type, 'RP,' was obtained from crushing PC culverts having a maximum aggregate size of 20 mm and a compressive strength of 35 MPa. 'RA,' obtained from a recycling plant, was prepared by crushing old concrete from the same source into a size of 20 mm. It should be noted that the old railway PC sleepers were used about for 20 years, while the old PC culverts were laid out in the open area at best for a year and old concrete structures for 'RA' more than 20 years.
Old concrete structures were crushed using a hydraulic breaker, and foreign materials such as rail pads and wires in the railway PC sleepers and reinforcing rebars in the PC culverts and other structures were collected by separation. Ton bags of initially crushed RCA were then transported to the recycling plant and treated by four, six, and three additional crushing process steps (combinations of jaw crusher and impact crusher) for 'RR', 'RP', 'RA,' respectively. Finally, RCA products were obtained. The original coarse aggregate sources of the old railway sleepers and the old PC culverts were granite.   Table 1 shows the specific gravity, and water absorption of 3 types of RCA, 2 types of natural coarse aggregate (NCA), and 2 types of fine aggregate. The specific gravity and the water absorption of the RR produced from concrete sleepers were 2.48% and 4.53%, respectively, and those of the RP, produced from concrete culvert, were 2.60% and 2.62%, respectively, and those of the RA, produced from a plant, were 2.52% and 3.82%, respectively.

Aggregates Properties
Crushed granite was used as NCA; its specific gravity and water absorption were 2.65% and 0.70%, respectively, in test series 1, and 2.71% and 0.37%, respectively, in test series 2. River sand was used as fine aggregate; its specific gravity and water absorption were 2.56, 2.60 and 1.10, 0.95 respectively for fine aggregate 1 and fine aggregate 2. Table  2 provides details of the test series. Old concrete structures were crushed using a hydraulic breaker, and foreign materials such as rail pads and wires in the railway PC sleepers and reinforcing rebars in the PC culverts and other structures were collected by separation. Ton bags of initially crushed RCA were then transported to the recycling plant and treated by four, six, and three additional crushing process steps (combinations of jaw crusher and impact crusher) for 'RR', 'RP', 'RA,' respectively. Finally, RCA products were obtained. The original coarse aggregate sources of the old railway sleepers and the old PC culverts were granite. Table 1 shows the specific gravity, and water absorption of 3 types of RCA, 2 types of natural coarse aggregate (NCA), and 2 types of fine aggregate. The specific gravity and the water absorption of the RR produced from concrete sleepers were 2.48% and 4.53%, respectively, and those of the RP, produced from concrete culvert, were 2.60% and 2.62%, respectively, and those of the RA, produced from a plant, were 2.52% and 3.82%, respectively. Note: a . The first letter 'R' refers to RCA, and the second letter 'R' means its production from railway sleepers. b . The first letter 'R' refers to RCA, and the second letter 'P' means its production from PC culvert. c . The first letter 'R' refers to RCA, and the second letter 'A' means general aggregate produced from a plant. Crushed granite was used as NCA; its specific gravity and water absorption were 2.65% and 0.70%, respectively, in test series 1, and 2.71% and 0.37%, respectively, in test series 2. River sand was used as fine aggregate; its specific gravity and water absorption were 2.56, 2.60 and 1.10, 0.95 respectively for fine aggregate 1 and fine aggregate 2. Table 2 provides details of the test series. The thermal treatment method suggested by Juan and Gutierrez [41] was applied to calculate the RMC value of RCA. The prepared RCA samples were heated in an electric furnace at 500 • C for 2 h, and the heated samples were taken out of the electric furnace and immersed in ice water to apply a heat shock to separate the mortar. The measurements were inserted into Equation (1)

Aggregates Properties
In Equation (1), W RCA denotes the weight of RCA dried in a dryer after the first sampling; W OVA is the weight of the original virgin aggregate (OVA) obtained by drying the sample in a dryer after removing the RM.

Mix Design
The experiment was performed using Type I Portland cement, of which the specific gravity and specific surface area were 3.15 and 3380 cm 2 /g, respectively. The experiment consisted of 23 test series. In test series 2 and 3, Type C fly ash was added at a ratio of 10% relative to the cement. The chemical admixture used in the present experiment was a polycarboxylate-based air-entraining agent and a water-reducing agent from a South Korean vendor (Econex Co., Pyeongtaek, Korea). All the aggregates were prepared under saturated-surface-dry conditions.
The first mix series consisted of 5 mixes having a compressive strength of 35 MPa, which is the standard for manufacturing railway PC sleepers. The tests were performed using the 'RR' aggregates obtained by crushing old PC sleepers. The mix symbols shown in Table 2 are described below. The first letter 'C' means the conventional ACI mix design, and the letter 'E' means the EMV mix design. The following letters 'NR' mean the crushed natural aggregate used for railway sleepers. As described above, 'RR' means RCA produced from railway sleepers. CRR68 and CRR33 represent the mixes prepared by respectively replacing 68% and 33% of the total coarse aggregates with the RR recycled aggregates. 'ERR68 in ERR68 and ERR68-2 means that the mixes were prepared using EMV mix design by replacing 68% of the aggregates with the 'RR' recycled aggregates. In addition, '-2 is a scale factor meaning that 1/2 of the RM suggested by the modified EMV mix design was treated with aggregates, and the remaining 1/2 was considered as mortar [42].
The second mix series consisted of 5 mixes prepared using the 'RP' recycled aggregates produced from old PC culverts. The conventional ACI mix design and the EMV mix design were also applied to the second mix series. The CRP and ERP series of mix series 2 were prepared respectively by using concrete in which the RP recycled aggregate was coated with a cement paste dissociation agent for chloride penetration resistance, and concrete to which cement paste dissociation agent was not applied [43].
The third mix series consisted of 4 mixes. The first three mixes were prepared by applying the same mix design as in the second mix series, and the fourth was prepared by using the 'RA' recycled aggregate to compare the physical properties between the different types of aggregate.
A pan mixer that allows for mixing at a volume of 60 L was used to perform the experiment in a laboratory located at Hongik University, Korea. The coarse aggregate and fine aggregate were added to the mixer and mixed together for about 30 s. Then, cementitious materials were further added and stirred for 30 s. After that, admixtures, carefully diluted in water, were added and mixed for 2 min.

Specimen Preparation
Concrete specimens were prepared using plastic molds and a predetermined consolidation method [44]. The specimens were demolded after 24 h. All the specimens were cured for at least 1 year before being used in the experiment. The size of the cylindrical specimens was a base diameter of 100 mm and a height of 200 mm.
The COTE of the concrete specimens was measured according to the AASHTO TP-60 method using CDP displacement transducer [18]. As shown in Figure 2, an invar material, rarely affected by temperature change, was used as a support frame, and was mounted on the base plate. The CDP was installed at the top and connected with a data logger to measure the real-time variation of the length of the specimens depending on the temperature. Three semispherical supports were arranged on the base plate at equal intervals so that the specimens could be firmly settled. The internal temperature of concrete was controlled by fabricating 4 temperature specimens into which thermocouple wires were inserted. The temperature specimens were installed at each corner of the water bath and connected to the data logger to measure the real-time temperature. The mean values from the temperature data were used in the analysis of the specimen temperature. In addition, to avoid the effect of vapor on CDP, the lid of the water bath was removed during the experimentation. The water level was controlled by applying a water level control device to prevent changes of water level in the water bath.
As shown in Figure 3, the water bath temperature was gradually increased from 10 • C, the initial temperature, to 50 • C at 5 h, and kept at 50 • C for 3 h. Subsequently, the temperature was gradually decreased to 10 • C over 5 h and kept at 10 • C for 3 h. This temperature cycle was repeated.
The COTE of the concrete specimens was corrected using SUS-304 as a correction specimen. The COTE of the individual concrete specimens was corrected using Equation (2):  Experimental apparatus for measuring COTE and system setup [45].
As shown in Figure 3, the water bath temperature was gradually increased from 10 °C, the initial temperature, to 50 °C at 5 h, and kept at 50 °C for 3 h. Subsequently, the temperature was gradually decreased to 10 °C over 5 h and kept at 10 °C for 3 h. This temperature cycle was repeated. The COTE of the concrete specimens was corrected using SUS-304 as a correction specimen. The COTE of the individual concrete specimens was corrected using Equation (2):

Basic Properties
The COTE of the SUS-304 specimen, the reference specimen applied to Equation (2) for calculating the COTE of concrete specimens, was measured using a thermomechanical analyzer (TMA). For analysis of the experimental results, the COTE of the mortar, aggregate and cement paste, used for the composite model, was also measured using the TMA. The specimens employed by the TMA had cylindrical shapes of 8 × 25 mm. Figure 4 shows photos of the specimens applied to the TMA, and Table 3 shows the results of the COTE measurement.  Experimental apparatus for measuring COTE and system setup [45].
As shown in Figure 3, the water bath temperature was gradually increased from 10 °C, the initial temperature, to 50 °C at 5 h, and kept at 50 °C for 3 h. Subsequently, the temperature was gradually decreased to 10 °C over 5 h and kept at 10 °C for 3 h. This temperature cycle was repeated. The COTE of the concrete specimens was corrected using SUS-304 as a correction specimen. The COTE of the individual concrete specimens was corrected using Equation (2)

Basic Properties
The COTE of the SUS-304 specimen, the reference specimen applied to Equation (2) for calculating the COTE of concrete specimens, was measured using a thermomechanical analyzer (TMA). For analysis of the experimental results, the COTE of the mortar, aggregate and cement paste, used for the composite model, was also measured using the TMA. The specimens employed by the TMA had cylindrical shapes of 8 × 25 mm. Figure 4 shows photos of the specimens applied to the TMA, and Table 3 shows the results of the COTE measurement.

Basic Properties
The COTE of the SUS-304 specimen, the reference specimen applied to Equation (2) for calculating the COTE of concrete specimens, was measured using a thermomechanical analyzer (TMA). For analysis of the experimental results, the COTE of the mortar, aggregate and cement paste, used for the composite model, was also measured using the TMA. The specimens employed by the TMA had cylindrical shapes of 8 × 25 mm. Figure 4 shows photos of the specimens applied to the TMA, and Table 3 shows the results of the COTE measurement.   As pretreatment for analyzing the COTE results of concrete specimens using the composite model, mortar specimens were prepared using the mix applied to mix series 1 and the one applied to mix series 2 and 3, and the elastic modulus of the mortar specimens at the age of 28 days was acquired, as shown in Table 4. As pretreatment for analyzing the COTE results of concrete specimens using the composite model, mortar specimens were prepared using the mix applied to mix series 1 and the one applied to mix series 2 and 3, and the elastic modulus of the mortar specimens at the age of 28 days was acquired, as shown in Table 4.  CRR68, CRR33, ERR68 and ERR68-2 specimens were 8.91, 8.93, 8.38 and 8.28 × 10 /°C, respectively. The error of the reference CNR specimen and ERR68-2 was large in comparison with the mean values. Figure 5b shows the COTE values relative to the COTE of the reference CNR specimen. The COTE values of the CRR68 and CRR33 specimens were 5.2% and 5.4% higher, respectively, in comparison with the COTE of the reference specimen. On the contrary, the COTE values of the ERR68 and ERR68-2 specimens were 1.1% and 2.3% lower, respectively. Therefore, compared with the CR-mixes prepared by the conventional mix design, the COTE of the ER-mixes was lower by 6-7%.

Concrete COTE Test Results
(a) measured COTE (b) relative COTE  Figure 6a shows the mean COTE values of the concrete specimens of mix series 2. The specimens of mix series 2 were prepared using concrete mix that had been used for manufacturing PC culverts, and RP recycled aggregate obtained by crushing old PC culverts was applied.
In mix series 2, the mean COTE of the CNR specimen prepared using the reference natural coarse aggregate was 8.54 × 10 −6 /°C, and the mean COTE values of the CRP25 and CRP50 specimens, prepared using the conventional concrete mix design and RP at 25% and 50%, were 8.66 and 8.59 × 10 −6 /°C, respectively. The COTE values of the ERP25 and ERP50 specimens, prepared by EMV mix design, were coincidently the same at 8.45 × 10 −6 /°C. The value of the CNP specimen is the mean of 3 specimens, and the values for the CRP25, CRP50, ERP25 and ERP50 specimens are each the mean of 6 specimens.  Figure 5b shows the COTE values relative to the COTE of the reference CNR specimen. The COTE values of the CRR68 and CRR33 specimens were 5.2% and 5.4% higher, respectively, in comparison with the COTE of the reference specimen. On the contrary, the COTE values of the ERR68 and ERR68-2 specimens were 1.1% and 2.3% lower, respectively. Therefore, compared with the CR-mixes prepared by the conventional mix design, the COTE of the ER-mixes was lower by 6-7%. Figure 6a shows the mean COTE values of the concrete specimens of mix series 2. The specimens of mix series 2 were prepared using concrete mix that had been used for manufacturing PC culverts, and RP recycled aggregate obtained by crushing old PC culverts was applied.
Appl. Sci. 2021, 11, 8214 10 of 16 As described above, the CRP and ERP series of mix series 2 were prepared respectively by using concrete in which RP recycled aggregate was coated with a cement paste dissociation agent intended for chloride penetration resistance, and concrete to which the cement paste dissociation agent was not applied. Assuming the coating agent did not affect the COTE, the COTE of the CRP series and ERP series specimens were each calculated as the mean of 6 COTE values. The errors of the CRP25 and ERP50 specimens were larger than those of the other specimens. Figure 6b shows the COTE relative to the mean COTE of the reference CNP specimen. The COTE values of the CRP25 and CRP50 specimens were 1.3% and 0.6% higher, respectively, than the COTE of the reference specimen. On the contrary, the COTE values of the ERP25 and ERP50 specimens were 1% and 1.1% lower, respectively.   Figure 7a shows the mean COTE of the concrete specimens of mix series 3. The specimens of mix series 3 were prepared using the same PC culvert concrete mix as used in mix series 2, and by adding RA as another recycled aggregate. In mix series 3, the mean COTE of the CNR specimen prepared using the reference natural coarse aggregate was 8.17 × 10 −6 /°C, and the mean COTE values of the CRP50 specimen, prepared by applying the conventional concrete mix design and using 50% RP, and the ERP50, prepared by applying EMV mix design and using 50% RP, were 8.40 and 7.96 × 10 −6 /°C, respectively. The mean COTE of the ERA specimens prepared by applying the EMV mix design and using 50% RA was 8.11 × 10 −6 /°C. Figure 7b shows the COTE values relative to the COTE of the reference CNP specimen. Compared with the COTE of the reference specimen, the COTE of the CRP50 specimen, prepared by applying the conventional mix design, was 2.8% higher, and the COTE In mix series 2, the mean COTE of the CNR specimen prepared using the reference natural coarse aggregate was 8.54 × 10 −6 / • C, and the mean COTE values of the CRP25 and CRP50 specimens, prepared using the conventional concrete mix design and RP at 25% and 50%, were 8.66 and 8.59 × 10 −6 / • C, respectively. The COTE values of the ERP25 and ERP50 specimens, prepared by EMV mix design, were coincidently the same at 8.45 × 10 −6 / • C. The value of the CNP specimen is the mean of 3 specimens, and the values for the CRP25, CRP50, ERP25 and ERP50 specimens are each the mean of 6 specimens.
As described above, the CRP and ERP series of mix series 2 were prepared respectively by using concrete in which RP recycled aggregate was coated with a cement paste dissociation agent intended for chloride penetration resistance, and concrete to which the cement paste dissociation agent was not applied. Assuming the coating agent did not affect the COTE, the COTE of the CRP series and ERP series specimens were each calculated as the mean of 6 COTE values. The errors of the CRP25 and ERP50 specimens were larger than those of the other specimens. Figure 6b shows the COTE relative to the mean COTE of the reference CNP specimen. The COTE values of the CRP25 and CRP50 specimens were 1.3% and 0.6% higher, respectively, than the COTE of the reference specimen. On the contrary, the COTE values of the ERP25 and ERP50 specimens were 1% and 1.1% lower, respectively. Figure 7a shows the mean COTE of the concrete specimens of mix series 3. The specimens of mix series 3 were prepared using the same PC culvert concrete mix as used in mix series 2, and by adding RA as another recycled aggregate. In mix series 3, the mean COTE of the CNR specimen prepared using the reference natural coarse aggregate was 8.17 × 10 −6 / • C, and the mean COTE values of the CRP50 specimen, prepared by applying the conventional concrete mix design and using 50% RP, and the ERP50, prepared by applying EMV mix design and using 50% RP, were 8.40 and 7.96 × 10 −6 / • C, respectively. The mean COTE of the ERA specimens prepared by applying the EMV mix design and using 50% RA was 8.11 × 10 −6 / • C. The CNP, CRP50 and ERP50 specimens of mix series 2 and mix series 3 were each prepared by applying the same mix design, and thus the mean COTE values obtained from the two mix series were calculated and compared, as shown in Figure 8. Compared with the reference CNP specimen, the COTE of the CRP50 specimen, prepared by applying the conventional mix design, was higher by 2.3%, and that of the ERP50 specimen, prepared by applying the EMV mix design, was lower by 0.5% and thus are considered almost the same. Therefore, the results showed that the COTE of RCA concrete specimens prepared by applying the EMV mix design was lower than those of specimens prepared by applying the conventional mix design.  Figure 7b shows the COTE values relative to the COTE of the reference CNP specimen. Compared with the COTE of the reference specimen, the COTE of the CRP50 specimen, prepared by applying the conventional mix design, was 2.8% higher, and the COTE values of the ERP50 and ERA50 specimens, prepared by applying the EMV mix design, were 2.6% and 0.8% lower, respectively.
The CNP, CRP50 and ERP50 specimens of mix series 2 and mix series 3 were each prepared by applying the same mix design, and thus the mean COTE values obtained from the two mix series were calculated and compared, as shown in Figure 8. Compared with the reference CNP specimen, the COTE of the CRP50 specimen, prepared by applying the conventional mix design, was higher by 2.3%, and that of the ERP50 specimen, prepared by applying the EMV mix design, was lower by 0.5% and thus are considered almost the same. Therefore, the results showed that the COTE of RCA concrete specimens prepared by applying the EMV mix design was lower than those of specimens prepared by applying the conventional mix design. from the two mix series were calculated and compared, as shown in Figure 8. Compared with the reference CNP specimen, the COTE of the CRP50 specimen, prepared by applying the conventional mix design, was higher by 2.3%, and that of the ERP50 specimen, prepared by applying the EMV mix design, was lower by 0.5% and thus are considered almost the same. Therefore, the results showed that the COTE of RCA concrete specimens prepared by applying the EMV mix design was lower than those of specimens prepared by applying the conventional mix design.

Dependence of Total Coarse Aggregate Volume
In the present study, both RCA and NCA were based on similar granite aggregates, and thus the effect of the aggregate type was considered to be negligible. Assuming that concrete comprises coarse aggregate and mortar (water, cement, fine aggregate, air void), it was predicted that the COTE of concrete would increase as the volumetric ratio of mortar increased or that the COTE of concrete would decrease as the volumetric ratio of the coarse aggregate increased.
Using Equation (3), the total coarse aggregate volume in the mix was calculated using the data shown in Tables 1 and 2

Dependence of Total Coarse Aggregate Volume
In the present study, both RCA and NCA were based on similar granite aggregates, and thus the effect of the aggregate type was considered to be negligible. Assuming that concrete comprises coarse aggregate and mortar (water, cement, fine aggregate, air void), it was predicted that the COTE of concrete would increase as the volumetric ratio of mortar increased or that the COTE of concrete would decrease as the volumetric ratio of the coarse aggregate increased.
Using Equation (3), the total coarse aggregate volume in the mix was calculated using the data shown in Tables 1 and 2: where V TCA is the sum of the volume of natural aggregate ( W NCA SG NCA ×1000 ) and the volume of original virgin aggregate ( W RCA ×(1− RMC S ) SG RCA ×1000 ) and where SG NCA and SG RCA are the specific gravities of natural coarse aggregate and of RCA, respectively; RMC is residual mortar content from Equation (2) and 'S' is a scale factor, as explained before and in reference [42]. Figure 9 illustrates the quadratic relations of the mean COTE values of the specimens prepared by applying the individual mixes of mix series 1-3 and the volumetric ratio of coarse aggregate of the individual mixes. In test series 1, the volumetric ratios of the coarse aggregate of the reference CNR mix and the ERR68 and ERR68-2 specimens of the EMV mix were 0.34, 0.35 and 0.35, respectively. The volumetric ratios of the coarse aggregate of the CRR68 and CRR33 specimens of the conventional mix design were 0.24 and 0.29, respectively. Figure 9a shows that as the total volumetric ratio of the virgin coarse aggregate in the recycled aggregate mix increased, the COTE of concrete specimens decreased with a determination coefficient of 0.98. As also described in many previous reports, our results clearly show that the volumetric ratio of the aggregate is one of the factors influencing the COTE of concrete specimens.
In test series 2, on the contrary, the volumetric ratios of the coarse aggregate of the reference CNR mix and the ERP25 and ERP50 specimens of the EMV mix were all about 0.39. The volumetric ratios of the coarse aggregate of the CRP25 and CRP50 specimens of the conventional mix design were 0.36 and 0.34, respectively. As shown in Figure 9b, with the determination coefficient of 0.95, the COTE of concrete specimens still decreased as the total volumetric ratio of the virgin coarse aggregate in the recycled aggregate mix increased.
aggregate of the CRP50 specimen, prepared by the conventional mix design, was 0.34. Figure 9c shows that as the total volumetric ratio of the virgin coarse aggregate in the recycled aggregate mix increased, the COTE of the concrete specimens decreased with a determination coefficient of 0.99. However, it is reported by literature [10][11][12]21] that the COTE value of concrete is proportional to the volume ratio of total coarse aggregate. Further study is needed to verify if the COTE value of concrete is linear or nonlinear to the volume ratio of total coarse aggregate.

Predicted COTE Values
Since concrete is a composite including various materials, the C be determined from the COTEs of the individual components. Two m ally been presented for this: one is the parallel model; the other is the s [46] combined the two models and presented the following composit α = (α + α ) + (1 − ) α + α + where α , α = COTE of mortar and aggregate, respectively V , V = volume fraction of mortar and aggregate, respectively E , E = elastic modulus of mortar and aggregate, respectively = relative proportions of materials conforming with the up solution.
It should be noted that Hirsh's model becomes the series mod it becomes the parallel model when XX = 0. Tables 3 and 4 summarizes the input variables of the Hirsh mod of the CNR, CRR68, CRR33, ER68 and ER68-2 specimens at the age o 1 were 27.1, 24.8, 25.8, 26.1 and 26.9 GPa, respectively [42], and thos CRP50, ERP25 and ERP50 specimens in mix series 2 were 27.7, 27 GPa, respectively [43]. Since the elastic modulus of concrete and th in the present study, as shown in Table 4, were very similar to each o in the Hirsh model was neglected. Therefore, the condition of XX = Under the assumption that XX = 1, the input variables in Tab into the Hirsh model to predict the COTEs of the specimens prep conditions; the results of the curve fitting of the COTEs are shown 10a,c shows that the predicted COTE values fit well to the measur determination coefficient of 0.85 and 0.86, respectively by linear rela Mix 3. However, the determination coefficient was lower in mix seri On the other hand, in mix series 3, besides the specimens prepared using the same mix design as used for mix series 2, additional specimens were prepared by applying the RA aggregate and the EMV mix design. Figure 9c shows the experimental results obtained from the specimens as a function of the total coarse aggregate volume. The volumetric ratio of the coarse aggregate was about 0.39 in the reference CNP specimen and in the ERP50 and ERA50 specimens of the EMV mix design. The volumetric ratio of the coarse aggregate of the CRP50 specimen, prepared by the conventional mix design, was 0.34. Figure 9c shows that as the total volumetric ratio of the virgin coarse aggregate in the recycled aggregate mix increased, the COTE of the concrete specimens decreased with a determination coefficient of 0.99. However, it is reported by literature [10][11][12]21] that the COTE value of concrete is proportional to the volume ratio of total coarse aggregate. Further study is needed to verify if the COTE value of concrete is linear or nonlinear to the volume ratio of total coarse aggregate.

Predicted COTE Values
Since concrete is a composite including various materials, the COTE of concrete may be determined from the COTEs of the individual components. Two models have traditionally been presented for this: one is the parallel model; the other is the series model [10]. Hirsh [46] combined the two models and presented the following composite equation: where α m , α a = COTE of mortar and aggregate, respectively V m , V a = volume fraction of mortar and aggregate, respectively E m , E a = elastic modulus of mortar and aggregate, respectively X = relative proportions of materials conforming with the upper and lower bound solution.
It should be noted that Hirsh's model becomes the series model when XX = 1, while it becomes the parallel model when XX = 0. Tables 3 and 4 summarizes the input variables of the Hirsh model. The elastic moduli of the CNR, CRR68, CRR33, ER68 and ER68-2 specimens at the age of 28 days in mix series 1 were 27.1, 24.8, 25.8, 26.1 and 26.9 GPa, respectively [42], and those of the CNP, CRP25, CRP50, ERP25 and ERP50 specimens in mix series 2 were 27.7, 27.7, 26.3, 26.4 and 27.2 GPa, respectively [43]. Since the elastic modulus of concrete and that of the mortar used in the present study, as shown in Table 4, were very similar to each other, the parallel part in the Hirsh model was neglected. Therefore, the condition of XX = 1 was applied.
Under the assumption that XX = 1, the input variables in Table 3 were substituted into the Hirsh model to predict the COTEs of the specimens prepared under different conditions; the results of the curve fitting of the COTEs are shown in Figure 10. Figure 10a,c shows that the predicted COTE values fit well to the measured COTE values with determination coefficient of 0.85 and 0.86, respectively by linear relationships in Mix 1 and Mix 3. However, the determination coefficient was lower in mix series 2 than in mix series 1 or 3, because RCA replacement ratios were 33% and 68% in mix series 1 and 25% and 50% in mix series 2. As a result, the distribution of the total coarse aggregate ratio, represented by the x-axis of the plot shown in Figure 9, is closer among the specimens in mix series 2 than in mix series 1. The difference between the predicted values and the measured values may have resulted because the α a and α m values were obtained using a dilatometer type TMA, but the COTE value of concrete was obtained under saturated conditions in a bath in which RH = 100% according to the AASHTO-TP60.

Conclusions
The present study was conducted to experimentally verify that the COTE of RCA concrete is dependent on the volume of the original virgin aggregate in the total RCA concrete mix. Three types of RCA (one crushed from railroad concrete sleepers; one crushed from precast (PC) culverts; and one obtained from a commercial recycling plant) were mixed by two concrete mixing methods (conventional mix method and EMV

Conclusions
The present study was conducted to experimentally verify that the COTE of RCA concrete is dependent on the volume of the original virgin aggregate in the total RCA concrete mix. Three types of RCA (one crushed from railroad concrete sleepers; one crushed from precast (PC) culverts; and one obtained from a commercial recycling plant) were mixed by two concrete mixing methods (conventional mix method and EMV method) and by varying the replacement ratio relative to that of natural coarse aggregate. Three mix series were used in the experiment. The conclusions acquired from the experimental results are described below. It should be noted that because of limited test variables and test numbers in this study, further study is needed for the findings of this study to be generalized to other RCA concrete types (i.e., using RCA from other sources, or having different parameters and exposure conditions).
(1) At the same RCA replacement ratio (volumetric ratio of RCA in the total coarse aggregate) of 68%, the COTE of RCA concrete prepared by EMV mix design was over 6-7% lower than that of RCA concrete prepared by conventional mix method, and it was similar to or 1-2% lower than that of the reference natural coarse aggregate concrete. (2) For RCA concrete prepared by conventional mix method, which does not take into account the residual mortar content (RMC) attached to RCA, the volumetric ratio of the original virgin aggregate was relatively small and the volumetric ratio of the mortar (or cement paste) was relatively high, and thus the COTE may have increased. (3) The COTE of RCA concrete is represented in terms of the volumetric ratio of coarse aggregate of the individual mixes by the best fit quadratic curves. However, further study is needed to verify if the COTE value of concrete is linear or nonlinear to the volume ratio of total coarse aggregate. (4) Finally, the predicted COTE obtained by applying the composite series model was slightly higher than the measured COTE, but the determination coefficient was higher than usual, indicating that experimental results obtained in this study are reliable.