Mechanical Properties and Structural Performance of Recycled Aggregate Concrete: An Overview

Around 25 billion tons of concrete is used each year globally that means over 3.8 tons per person in the world each year [1]. Occupying about 60% to 70% of the volume, aggregate is a main ingredient of concrete. Nowadays, there is an increasing trend toward using sustainable concrete and reusing of aggregates from demolished concrete is one way to achieve this aim. Although using recycled concrete aggregate (RCA) does not lead to significant reduction of CO2 emission, but it reduces using natural resources (virgin aggregate) as well as helping to reduce dumping construction and demolition wastes in landfills. More than 900 million tons of construction and demolition waste is produced each year only in Europe, the U.S., and Japan [1]. The use of RCA to make a new concrete began in Europe after World War II. Also the building contractors’ society of Japan started research on RCA after the 1973 oil crisis [2]. Although comprehensive research has been done on fresh and hardened properties of recycled aggregate concrete (RAC), relatively limited studies have been performed on structural behavior of RAC. This paper compiles around 60 published studies (from 1977 to 2015) that studied mechanical properties (splitting tensile strength, modulus of elasticity, flexural strength, and fracture mechanic) as well as structural performance (shear, flexural, and bond) of RAC. The work presented here summarized only those studies that used RCA as a coarse aggregate replacement. Given the fact that incorporation of fine RCA is typically reported to result in inferior fresh and hardened properties, the present study summarizes the properties of RCA incorporating coarse RCA [3].


Introduction
Around 25 billion tons of concrete is used each year globally that means over 3.8 tons per person in the world each year [1]. Occupying about 60% to 70% of the volume, aggregate is a main ingredient of concrete. Nowadays, there is an increasing trend toward using sustainable concrete and reusing of aggregates from demolished concrete is one way to achieve this aim. Although using recycled concrete aggregate (RCA) does not lead to significant reduction of CO 2 emission, but it reduces using natural resources (virgin aggregate) as well as helping to reduce dumping construction and demolition wastes in landfills. More than 900 million tons of construction and demolition waste is produced each year only in Europe, the U.S., and Japan [1]. The use of RCA to make a new concrete began in Europe after World War II. Also the building contractors' society of Japan started research on RCA after the 1973 oil crisis [2]. Although comprehensive research has been done on fresh and hardened properties of recycled aggregate concrete (RAC), relatively limited studies have been performed on structural behavior of RAC. This paper compiles around 60 published studies (from 1977 to 2015) that studied mechanical properties (splitting tensile strength, modulus of elasticity, flexural strength, and fracture mechanic) as well as structural performance (shear, flexural, and bond) of RAC. The work presented here summarized only those studies that used RCA as a coarse aggregate replacement. Given the fact that incorporation of fine RCA is typically reported to result in inferior fresh and hardened properties, the present study summarizes the properties of RCA incorporating coarse RCA [3].

Mechanical Properties
The following section compares splitting tensile strength and flexural strength as well as modulus of elasticity and fracture energy of RAC with the ACI 318, CEB-FIP, Eurocode 2 and JSCE provisions [4,7]. and ratio of the experimental-tocode values for splitting tensile strength based on the ACI 318 [5] and Eurocode 2 [7] provisions for previous studies . The average values of experimental-to-code ratio are 0.90 and 0.92 for the ACI 318 [5] and Eurocode 2 [7] provisions, respectively. Both provisions are conservative for about 22% of the data points. In other words, the investigated codes overestimate the splitting tensile strength for RAC ( Figure 1) based on measured compressive strength values. Furthermore, Regression analysis was conducted to draw the best fit and 95% confidence intervals for the investigated data points. The lower and upper 95% Civil Engineering Research Journal confidence intervals (L95 and U95) are established. Results of the statistical data analysis reveal the fact that even in the case of the mixtures made with 100% RCA replacement (RAC-100), test results fall within a 95% confidence interval of a nonlinear regression curve fit of the conventional concrete (CC) database ( Figure 1).     Figure 2 presents the flexural strength database of RAC mixtures compared to the provisions of both ACI 318 [5] and Eurocode 2 [6]. Based on the observed trends, it may be concluded that both provisions over predict flexural strength of RAC. However, the ACI 318 [4] is unconservative for 24% of the data points compared to overestimations observed for about 65% of the data points in the case of the EC 2 [6].  [6] provision. This reveals the fact that the ACI 318 underestimates the flexural strength of RAC by 15%, proposing ACI 318 as a reliable source for predicting the flexural strength of the RCA. Contrary to the ACI 318, the EC2 seems to overestimate the average flexural strength by about 8%. In addition, regression analysis of the CC data indicates that part of the RAC-100 test results don't fall within a 95% confidence interval of a nonlinear regression curve fit of the CC database. This result indicates that the RAC-100 flexural strength is lower than the CC for mixes with compressive strength around 35MPa and higher.    [6] provisions showed only 34% and 5% of results is higher than the codes predictions, respectively ( Figure 3). Almost all the previous studies reported lower modulus of elasticity for RAC compared to CC and it may be attributed to the lower stiffness and rigidity of the RCA particles which in turn decrease the rigidity and modulus of elasticity of the RAC [8].

Fracture energy
Limited literature is available in the case of the fracture energy of RAC. Results presented in [21,24,27,33] suggest decrease in fracture energy as a function of increases in RCA replacement ratio. Table 4 shows that the CEB-FIP [5] code provision is conservative only for 7% of the investigated data points The Japan's code [7] underestimates the fracture energy for 17% of the experimental results (Table 4). Likewise splitting tensile strength, statistical analysis of the data shows that the RAC-100 test results fall within a 95% confidence interval of a nonlinear regression curve fit of the CC database. That means the RAC-100 fracture energy is similar to the CC ( Figure 4).

Structural Performance
The following sections compare structural test results of RAC including bond, flexural, and shear strength with both codes provisions and CC databases.

Bond Strength
Three most common methods to measure bond strength between reinforcing steel and concrete are pull-out, beam-end, and splice specimens. Although pull-out test method is the most commonly used due to the ease of sample fabrication, but this method has the least realistic bond strength measurements. The reason is the bar at pull-out specimen is loaded in tension and the surrounding concrete is in compression, but in most practical applications, both the bar and surrounding concrete experience tensile stresses. For beam-end and splice specimens test methods, both the bar and surrounding concrete are in tension and provide a more realistic response [40]. The following section discusses the results of each test method separately. . The bond strength provision is a function of the inverse square root of the compressive strength of concrete for most of design standards [4,7,[14][15][16], but the ACI 408 [13] recommends the fourth root of the compressive strength of concrete instead of the square root. For this reason, Table 5(a-e) presents normalized bond strength of previous studies for both methods.  Mukai & Kikuchi [41] used both 15% and 30 % RCA replacement in their specimens and reported no reduction in bond between reinforcing steel and RAC. Ajdukiewicz & Kliszczewicz [11] replaced 100% RCA from seven years old structures for pull-out specimens based on RILEM 7-II-128 [42]. Their results showed that the RAC specimens possess 7% lower bond strength compared to the Xiao & Falkner [43] tested 36 pull-out specimens with 0, 50%, and 100% RCA replacement and reported similar load-slip curve for RAC specimens. However, 4% and 12% increase on bond strength of RAC was reported for the 50% and 100% replacements, respectively compared to the Breccolotti & Materazzi [44] used pull-out specimens with 0, 50%, and 100% RCA replacement and observed no significant difference between bond strength of the RAC and CC specimens.

Civil Engineering Research Journal
Kim & Yun [22] used RCA with the replacement ratios of 30%, 60%, and 100% to make 144 direct pull-out tests. Authors reported that the RAC specimens with the smaller aggregate size had a greater bond strength compared to the specimen with the larger aggregate size, while the CC specimen showed similar bond strength for both aggregate sizes. Furthermore, authors reported that the current ACI 318 [5] and CSA [45] codes provisions can be applied to RAC specimens. Steele [28] results showed that replacing both 50% and 100% RCA slightly improved the bond strength of reinforcing steel and RAC compared to the CC Civil Engineering Research Journal specimens. Prince and Singh [46] tested 30 pull-out specimens with 0, 25%, 50%, 75%, and 100 % RCA replacement ratios and reported identical load-slip relationships for the RAC and CC specimens. However, the bond strengths for 10mm diameter deformed bars in the RAC were higher than the CC specimens.  Table 6 presents dimensions and mechanical properties of materials used to test splice beams. The only difference is that Table 6 includes reinforcing steel stress at failure, fs, instead of average bond stress based on ACI 408 [40] recommendation. Steele [28] tested full scale beams with cross section of 300×460mm for both 50% and 100% RCA replacement levels and reported that the bond strength decreased around 10% for 50% and 20% for 100% RCA replacement.

End beam
Fathifazl et al. [47] used both conventional method (replaced 100% RCA) as well as the equivalent mortar volume (EMV) method for proportioning RAC. They reported 24% lower bond strength for the conventional method, while the EMV method bond strength was comparable (only 6% reduction) with the CC specimens. Butler et al. [17] tested 48 beam-end specimens with 100% RCA replacement from different resources, including regional sidewalk, curb, and gutter infrastructure, apron and terminal structures from Pearson International Airport in Toronto, Canada, and a concrete ready-mix plant. Their results showed 3% to 20% reduction in bond strength between reinforcing steel and the RAC compared to the CC beams. A summary of details for end beam tests can be found in Table 7 (a,b).     [41] tested beams measuring 150×150mm in cross section and 1.8-m in length with both 15% and 30 % RCA replacement ratios and reported no significant difference in ultimate moment. However, authors reported lower cracking moment for RAC beams. Yagashita et al. [48] used low grade, medium grade, and high grade types of recycled aggregate with 100% replacement for determining the flexural strength of reinforced concrete beams. Their results showed using high grade RCA slightly decrease (around 10%) the flexural capacity of RAC beams. Ajdukiewicz and Kliszczewicz [11] used partial or full recycle aggregate. All the beams were fabricated with rectangular sectons measuring 200×300mm and 2600mm in length with two longitudinal reinforcement ratios of 0.90% and 1.60%.

Flexural Strength
They reported that the RAC beams had slightly (3.5% in average) lower moment capacity and higher deflection compared with the CC beams. Sato et al. [12] tested 37 beams with three different longitudinal reinforcement ratios of 0.59%, 1.06%, and 1.65%. They used 100% recycled aggregate for their mix designs. Results of their study showed that the RAC beams had larger deflection compared to the CC beams. In terms of crack spacing no significant difference observed between the RAC and CC beams. However, the RAC beams had wider crack openings compared to the CC ones. They also reported almost the same ultimate moment for the RAC and CC beams.
Maruyama et al. [49] tested beams with 1% longitudinal reinforcement ratio and reported that the cracks propagated on the RAC beams were wider and spaced closer compared to those of the CC beams. The RAC beams had larger deflection, but authors reported no significant difference between the flexural capacity of the RAC and CC beams. Fathifazl et al. [15] incorporated the EMV method for their mix designs. They used both 64% and 74% RCA as a partial replacement of the coarse aggregate for their RAC mixtures. Their beams were cast with three different longitudinal reinforcement ratios ranging between 0.49% and 3.31%. Authors reported comparable and even superior flexural behavior for RAC beams at both service and ultimate states. They concluded that the flexural strength provisions of the current codes can be used for RAC beams. Bai & Sun [36] used 8-10 years old RCA with different replacement levels of 50%, 70%, and 100%. They observed similar crack pattern, but deflection and crack width increased with the increment of RCA replacement level. They also concluded that RCA replacement level doesn't significantly affect the cracking ultimate moment of beams. Ignjatovic et al. [20] studied nine full scale beams with 0, 50%, and 100% recycled coarse aggregate and 0.28%, 1.46%, and 2.54% longitudinal reinforcement ratio. They reported no noticeable difference between load-deflection behavior, service load deflection, and ultimate flexural strength of RAC and CC beams. But they observed that the beams with higher range of recycled aggregate showed higher level of concrete destruction at failure.
Kang et al. [25] used beams with longitudinal reinforcement ratio ranged between 0.5% and 1.8% with RCA replacement level up to 50% for both normal and high strength concrete. They observed greater number of cracks and lower cracking moment for RAC beams. They also reported no significant decrease in flexural capacity up to 30% RCA replacement level. Knaack & Kurama [32] tested 150×230mm cross section and 2000mm long beams. They used RCA from late 1920s foundation and with both 50% and 100% replacement level. They reported higher deflection for the RAC beams, but they concluded that the existing analytical models and code provisions can be used for the RAC beams. Arezoumandi et al. [50] replaced 100% RCA in full scale beams (cross section 300×450mm and 3050mm long) and reported similar crack pattern and flexural capacity, but lower cracking moment for the RAC beams.
As it can be seen from Table 9 (a,b), in terms of ultimate moment, when 100% RCA used as a coarse aggregate, the ACI 318 [5]and EC 2 [7] provisions are conservative for 82% and 92% of beams, while only 83% of the RAC beams showed yielding moment higher than the predicted values. Furthermore, 53% of the RAC beams showed cracking moment less than the ACI 318 [5] provision; however, it was 39% when it compared to the EC 2 [7] provision. The ACI 318 [5] and EC 2 [7] ultimate moment provisions are conservative for the beams used less than 30% RCA as a coarse aggregate, but in terms of cracking and yielding moment 30% and 60% of beams showed values less than code predictions, respectively.   Figure 4 presents the normalized flexural strength versus normalized longitudinal reinforcement ratio for the RAC beams as well as the wealth of flexural test data available in the literature for CC [51]. Figure 4 seems to indicate that the RAC and CC test results fall within the central portion of the data. Furthermore, statistical analysis (regression analysis) of the data indicates that the RAC and CC test results fall within a 95% confidence interval of a nonlinear regression curve fit of the database. This result indicates that the RAC flexural test values are very consistent with the wealth of flexural test data available in the literature.

Shear Strength
Mukai & Kikuchi [41] replaced both 15% and 30 % RCA instead of virgin aggregate and reported slightly inferior shear strength for the RAC compared with the CC beams. Han et al. [52] tested beams with different shear span to depth (a/d=1.5 to 4.0) and both washed and unwashed RCA and observed no significant difference in failure mode and shear capacity between the RAC and CC beams; however, the ACI 318 [5] shear provision was unconservative for beams with a/d greater than 3.0. Yagashita et al. [48] used low, medium, and high grade RCA with 100% replacement level. They reported 8% lower shear stress at failure for the RAC beams (with high grade RCA) compared to the CC. Sogo et al. [53] used RCA (10 to 40-year-old concrete foundation) for casting the beams and reported 20% lower shear strength for the RAC beams compared to the CC beams. Gonza´lez-Fonteboa et al. [54] tested beams with 50% RCA replacement and observed no significant difference on shear capacity between the RAC and CC beams. Etxeberria & Va´zquez [55] used different percentages of RCA (25%, 50%, and 100%) for casting specimens and concluded that a substitution of less than 25% RCA, scarcely affects the shear capacity of RC beams. Ji et al. [56] tested beams with a cross section of 170×270mm and a shear span-to-depth of 2 and 100% RCA replacement. Results of their study showed that existing codes can be used to predict shear strength of the RAC beams. Choi et al. [16] tested beams with different span-to-depth ratios (1.50, 2.50 and 3.25) and different RCA replacement ratios (30%, 50%, and 100%). They reported the lower shear strength for the RAC beams. Fathifazl et al. [57] used the EMV method with both 63.5% and74.3% recycled aggregate replacement as a coarse aggregate for their mix designs. They tested beams with four different shear spanto-depth ratios ranging between 1.5 and 4. They reported superior shear strength for the RAC beams. They also concluded that current code provisions for shear conservatively predicted the capacities of the RAC beams. Schubert et al. [30] tested slabs with 100% RCA replacement and concluded that existing codes provision can be used to calculate shear strength of the RAC slabs. Xiao et al. [58] tested push-off specimens with different percentages of recycled coarse aggregate replacement. They observed similar shear stress-slip curves and crack propagation path between the RAC and CC specimens. They also reported that more than 30% RCA replacement decrease the ultimate shear capacity. Arezoumandi et al. [59] replaced 100% RCA in full scale beams with different longitudinal reinforcement ratios and reported 12% lower shear strength for the RAC beams compared to the CC beams. Knaack & Kurama [32] replaced both 50% and 100% of RCA (from 90 years old building) to cast beams with cross section of 150mm×230mm. Their results showed small reduction on shear strength of the RAC beams, but still existing codes can be used to predict shear strength of the RAC beams.  Table 10 summarizes details of shear beams test including width, b, height, h, depth, d, shear span to depth, a/d, maximum aggregate size, dmax, longitudinal reinforcement ratio, ρ, compressive strength, f´c, and shear at failure, V. Furthermore, test results compared with shear provisions of both ACI 318 [5] and EC 2 [7]. Table 10 (a) shows that both ACI 318 [5] and EC 2 [7] shear provisions are conservative for more than 90% of data, even though for the majority of tests the RAC beams showed lower shear strength compared to the CC beams.
The four key parameters that affect concrete contribution to shear strength include depth of member or size effect (d), shear span to depth ratio (a/d), compressive strength of concrete (f´c), and longitudinal reinforcement ratio (ρ). To evaluate the effect of the aforementioned parameters on shear strength of the beams, the results of RAC beams are compared with the wealth of shear test data available in the literature for CC [60].  Figure 5 presents the shear stress versus f´c, ρ, d and a/d, respectively. Given the significant scatter of the database of previous shear test results, it is somewhat difficult to draw definitive conclusions on the current test values. Nonetheless, visually, Figure 5 seems to indicate that the RAC test results fall within the lower portion of the data -except results of Fathifazl et al. [56] -and follow the same general trend of the database. Furthermore, statistical analysis (using regression analysis to draw the best fit and 95% confidence intervals) of the data indicates that the test results fall within a 95% confidence interval of a nonlinear regression curve fit of the database. This result indicates that the RAC beams shear strength is lower than the shear strength of the CC.

Conclusion
The following conclusions can be drawn from literature review of mechanical properties and structural behavior of RAC specimens: a.
The existing codes provisions are unconservative for splitting tensile strength of RAC around 10%.
Moduli of elasticity of RAC are around 10% to 20% lower than the predicted values of the ACI 318 [4] and EC 2 [7] codes provisions. d.
Moduli of elasticity of RAC are around 10% to 20% lower than the predicted values of the ACI 318 [4] and EC 2 [7] codes provisions. e.
The CEB-FIP [5] and Japan's code fracture energy provisions are unconservative for 90% and 20% of RAC, respectively. f.
Pull-out specimens with different RCA replacement levels showed almost identical load-slip behavior with no significant difference in bond strength between RAC and CC specimens. g.
Both splice beam and end beam results showed up to 20% reduction in bond strength of RAC with 100% RCA replacement. h.
In terms of flexural strength, RAC beams showed wider cracks, lower cracking moment, larger deflection, but similar ultimate moment with CC. i.
The RAC beams test results fall within a central portion of 95% confidence interval of a nonlinear regression curve fit of a CC flexural test database.

j.
Similar load-deflection for shear behavior observed between RAC and CC beams, but the majority of previous studies reported lower shear strength for RAC beams (up to 20% when 100% RCA used instead of virgin aggregate) compared to CC.

k.
The RAC beams test results fall within a lower portion of 95% confidence interval of a nonlinear regression curve fit of a CC shear test database. l.
Using the EMV method avoids inferior mechanical and structural performance of RAC.