Mechanical Properties of Lightweight Aggregate Concrete Reinforced with Various Steel Fibers

The objective of the present study is to examine the effects of different types and content of steel fibers on the workability, mechanical properties, and ductility enhancement of lightweight aggregate concrete (LWAC). Thirteen LWAC mixtures were prepared using different types of steel fibers including conventional hooked-end macro steel fibers (SF(H)), straight (CSF(S)), crimp-shaped (CSF(C)), and hooked-end (CSF(H)) copper-coated micro steel fibers at the designed compressive strength of 40 MPa. The fiber volume fraction varied from 0% to 1.5% at an interval of 0.5%. Test results showed that CSF(H) type possesses a better potential than the other types in enhancing the tensile resistance and toughness at the post-peak branch of the load–deflection curve of beams. The bond characteristics of fibers with cement matrix were one of critical parameters that needs to be considered in designing the fiber reinforced concrete. Thus, the splitting strength, modulus of rupture, and modulus of elasticity of fiber reinforced LWAC were formulated as a function of volume fraction, aspect ratio, and bond strength of fibers from the regression analysis using the present test data.


Introduction
With the development of large and tall concrete structures, concrete is required to have better performance (e.g., workability, higher strength and toughness, and light weight) (Mays & Barnes, 1991). Lightweight aggregate concrete (LWAC) is superior to normal weight concrete (NWC) in terms of fire resistance, seismic resistance, and thermal insulation (Li et al., 2017). LWAC has a higher strength-to-weight ratio than NWC (ACI Balendran et al., 2002;Committee&213, 2014). Hence, relatively smaller cross-sectional area and lighter structural elements can be designed using LWAC. It the ductility of concrete even with a relatively low V f . Moreover, surface coating can help in long-term corrosion resistance. Kang et. al. (2010) reported that the tensile strength of ultra high performance fiber reinforced concrete increases linearly when the V f of the steel fibers is increased up to 5%. Dupont and Vandewalle (2005) pointed out that when V f exceeded 2%, the interaction between fibers increased. This could induce a balling effect during the mixing or placing process of concrete. Park et. al. (2012) and Kim et. al. (2011) showed that deformed macro steel fibers (twisted, crimped, and hooked-end shapes) could significantly improve the tensile and flexural performance of concrete compared to non-deformed steel fibers (e.g., straight shape). The deformed shape can increase the fiber pullout resistance in the cement matrix, and it could be directly related to the improvement of the bridging effect of the fibers (Wu et al., 2018). Visalvanich and Naaman (1983) indicated that there was a concern that the deformed shapes of steel fibers could act as a link to weave other fibers during the mixing process. This could inhibit the distribution of fibers. As a result, the geometry of the steel fiber can lead to contradictory results in assessing the workability, tensile capacity and mechanical properties of concrete. Hence, shape of fiber should be considered as a critical parameter along with V f in assessing the mechanical properties of concrete. Additional studies on fiber reinforced lightweight aggregate concrete (FRLWAC) are needed because of the variability in the characteristics of lightweight aggregates (LWAs). In LWAC, cracks and fractures first occur in the aggregates unlike NWC, where the performance of the concrete is mainly determined by the paste, because the aggregate is generally stronger than the paste (Sim et al., 2013). Lee et. al. (2019) reported that the low strength of the lightweight aggregates yields more rapid crack propagation and more brittle fracture of concrete, because the cracks mainly pass through the aggregate particles, resulting in localized failure zones. Therefore, a large amount of data on the mechanical properties of FRLWAC is needed to establish a design model for making detailed decisions (e.g., fiber type, V f and LWA type, density, and water absorption) to achieve the target performance of FRLWAC.
In this study, we evaluated the effects of various properties of different fibers on the mechanical properties of FRLWAC. Four steel fiber types with different geometrical shapes, L f , and d f were used in this study. Twelve FRLWAC mixtures were prepared for different V f . A fiberless mixture was used as a control group. The compressive strength (f c (t)), splitting tensile strength (f sp ), load-deflection relationship, modulus of rupture (f r ) of concrete, and bond stress-slip relationship of embedded individual fibers were evaluated in this study. In addition, equations were developed based on experimental data and regression analysis to comprehensively evaluate the various mechanical properties of FRLWAC.

Materials and Mix Proportion
The ordinary Portland cement used in this study had a density of 3.15 g/cm 3 and a specific surface area of 3260 cm 2 /g with other properties meeting the specifications of ASTM type I (2012) (ASTM C150/C150M, 2011). Commercially available artificially expanded granules manufactured by Korea South-East Power Co. were used as structural LWAs. The bottom ash and dredged soil used for the raw materials of the granules were calcined and allowed to expand in large rotary kilns at approximately 1200 °C. The main components of the LWA, confirmed by X-ray diffraction, were quartz and magnetite (Fig. 1). The expanded granules were produced in nominal sizes (e.g., below 2 mm, 2-5 mm, 5-10 mm, and 10-20 mm), and adjusted to satisfy the particle distribution curve specified in ASTM C 330 (ASTM C330/ C330M, 2017). The physical characteristics of the LWAs used are summarized in Table 1. In this study, four types of steel fibers with different shapes, surface coating conditions, L f , d f , and τ were used. In addition, different copper-coated fibers (i.e., straight (CSF(S)), crimped (CSF(C)), and end-hooked (CSF(H)) in addition to conventional end-hooked steel fibers (SF(H)) were utilized (Fig. 2). The manufacturer (SM KOREA Co., Anyang-si, Korea) reported that the tensile strengths of SF and CSF were 2240 and 1220 MPa, respectively. Additional fiber properties are shown in Table 2. Table 3 shows the mix proportion of LWAC used in the experiment. The water-to-cement ratio (W/C) was set to be 0.35, and the fine aggregate-to-total aggregate ratio (S/A) was set to be 0.45. Twelve FRLWAC mixtures and a counterpart control fiberless mixture were initially prepared considering the steel fiber types and their V f . There were three levels of V f (0.5%, 1.0%, and 1.5%) for each fiber type. Thus, the FRLWAC specimens were denoted using the steel fiber type and V f level. For example, CSF(C)-1.0 indicates an LWAC mixture reinforced using crimped steel fiber with V f = 1.0%. Specimen Co refers to the control fiberless mixture.

Methods
Steel fibers were added after wet mixing of the LWAC. To achieve the targeted minimum slump, a different contents of commercially available polycarboxylatebased superplasticizer was added to each mixture. The slump and air content tests of fresh concrete were conducted according to ASTM C143 and C231, respectively. All specimens were subjected to the vibration casting method specified in ASTM C31. The specimens were cured at a steady temperature and relative humidity of 23 ± 2 °C and 60 ± 5%, respectively, until they were tested at the specified age. The compressive strength, dry densities, splitting tensile strength, and elastic modulus of concrete were measured using 100 × 200 mm cylindrical specimens, as specified in ASTM C39, C138, C496, and C469, respectively. The relationship between the modulus of rupture and load deflection was measured using prismatic beams having dimensions of 100 × 100 × 400 mm under four-point bending according to ASTM C78 and C1018, respectively. Fig. 3 shows the fiber pullout test setup used to evaluate the bond stress for each steel fiber type. A load cell with a maximum capacity of 3 kN was used to measure the applied pullout load in the jigs. A dog-bone specimen embedded with one individual fiber was fixed between two jigs, and the pullout test was performed at a speed of 0.02 mm/s (Kim & Yoo, 2020). The pullout of each fiber was commonly induced to occur at the side of the specimen with a 10 mm fiber embedment length. The applied load-slip relationship was evaluated by assuming that the stroke displacement was the slip of the fiber. The average bond stress of the fiber (τ) was calculated by dividing the maximum pullout load by the initial bonding area between the fiber and matrix (Eq. 1): where P max is the maximum pullout load, and L E is the initial embedded length of the fiber. The mean of the three specimens was applied to all the test results. The compressive strength was measured at 3, 28, and 91 days, and all other tests were performed at 28 days.

Fiber-Matrix Bond Stress
The increase in the mechanical performance of concrete owing to fiber reinforcement is strongly related to the fiber-matrix interaction and bond strength. Kim and Yoo (2020) reported that the pullout load of the fiber increased significantly when the mechanical anchorage effect was induced owing to the geometric deformation of the steel fiber. Fig. 4 shows the bond stress-slip relationship for each steel fiber type. Although the L f of CSF(C) and CSF(S) was 13 mm (Table 2), steel fibers with L f of 20 mm or more were used in this test to fix the L E to 10 mm. The bond stress-slip curve was substantially different depending on the type of steel fiber (Fig. 4). Particularly, crimped and hooked-end shapes showed relatively  high τ values owing to their additional mechanical anchorage effects. In contrast, the straight shape was chemically attached to the adjacent cement matrix. This can also be confirmed by the very steep slope of the descending section, i.e., after the peak in the relationship curve of SCF(S). The test results showed that CSF(H) and SF(H) had excellent pullout resistance with maximum τ values of 18.68 and 16.2 MPa, respectively, followed by SCF(C) (11.2 MPa) and SCF(S) (8.6 MPa) ( Table 2). Moreover, the results indicated that if other engineering characteristics were excluded and only the influence on the bridging effect between matrices was considered, transforming it into a hooked-end shape can be advantageous. Kim and Yoo (2020) also reported that τ increased in the order of hooked-end, twist, and  Type straight shape, and the averages were 18.6, 17.3, and 10.9, respectively, which was similar to the results seen.

Workability
In the case of LWAC, a lighter mixture may induce a lower slump (S l ) owing to the effect of gravity between slump tests (Kockal & Ozturan, 2011). Mehta and Monteiro (2017) reported that an SI of approximately 50-75 mm could be sufficient to ensure good workability of LWAC. The test results of the FRLWAC workability showed that the addition of steel fibers tended to lower S l (    was increased from 0.5% to 1.5%, the ratio of superplasticizer content (R sp ) of SF(H), CSF(C), and SCF(H) mixtures increased by 0.15%, 0.4%, and 0.35%, respectively. However, the measured S l values tended to decrease. On the other hand, for the CSF(S) mixture, when V f increased from 0.5% to 1.5%, R sp increased by 1.5%, while S l decreased by 30 mm. The result indicated that the deformed steel fibers were more unfavorable than the straight-shaped in maintaining the workability of LWAC, because the geometric shape deformation of fibers could accelerate the fiber balling effect owing to the geometry of fibers. The slump value of fresh concrete can be determined mainly by V f and R sp (Fig. 5). The regression analysis results showed that the slump value was inversely proportional to V f and directly proportional to R sp . Therefore,  V f is the volumetric fraction of fibers, R sp is the ratio of the superplasticizer content relative to the cement mass content, S l is the slump of fresh concrete, A c is the air content of fresh concrete, and ρ c , f c (t), f sp , f r , and E c are the dry density, compressive strength at age t, splitting tensile strength, modulus of rupture, and modulus of elasticity of hardened concrete, respectively. it is possible to predict workability using parameters, such as V f and R sp , in mixtures, where the same fibers are utilized. In securing the workability of FRLWAC, SF(H) and SCF(S) seem to be relatively effective. This might be attributed to the fact that SF(H), with a greater d f and longer L f , produces fewer individual fibers, which can degrade the slump of concrete compared to fibers with a smaller diameter. For V f = 1.5% specimens, the compressive strength of SF(H), CSF(S), CSF(C), and CSF(H) increased by approximately 23.1%, 39.8%, 11.3%, and 15.8%, respectively, compared to the Co mixture. In CSF(S) and SF(H) groups, the addition of fibers had a relatively positive impact on strength improvement, which was closely related to the dispersibility of the fibers. Huang and Zhao (1995) indicated that fiber distribution is very difficult in FRLWAC with a high V f value. This could result in incomplete compaction and decreased compressive strength. The results of this study showed that CSF(C) and CSF(H), which have a higher quantity of deformed individual fibers, exhibited relatively low strength, i.e., the strength of V f = 1.5% specimens was lower than that of V f = 1.0%. The relative oven-dry density (ρ c ) of FRLWAC increased linearly with the V f value, and this increasing rate was independent of the steel fiber type. All the ALWC mixtures exhibited a parabolic curve of the compressive strength development. The ratios of 3-day compressive strength relative to 28-day strength ranged between 0.69 and 0.78, 0.68 and 0.76, 0.7 and 0.82, and 0.72 and 0.85 for SF(H), CSF(S), CSF(C), and CSF(H) groups, respectively. The corresponding rages obtained at 91 days are between 1.03 and 1.14, 1.06 and 1.14, 1.01 and 1.08, and 1.02 and 1.12, respectively, for such groups. Hence, the compressive strength development of LWAC was marginally affected by the type, shape, and volume fraction of steel fibers.

Splitting Tensile Strength
According to ASTM C 330, structural LWACs must have a splitting tensile strength (f sp ) of at least 2.0 MPa. The f sp of LWAC was significantly lower than that of NWC with the same compressive strength, because initial cracks can be induced inside the aggregates rather than in the ITZ between the aggregates and matrix (Sim et al., 2013). When V f was 0.5%, the f sp of CSF(C) and CSF(H) were relatively higher than that of the others (Fig. 7). When V f increases by 1.0% or more, f sp increases linearly. However, the increase in CSF(S) was relatively higher. Hassanpour et. al. (2012) reported that the f sp of FRLWAC was 16-61% higher than that of fiberless LWAC. f sp /(f sp ) N , representing the relative f sp value of FRLWAC compared Page 8 of 14 Yang et al. Int J Concr Struct Mater (2022) 16:48 to that of the fiberless specimen, was in the range of 1.13-1.92, where the subscript N refers to the fiberless control concrete. The f sp /(f sp ) N value ranged from 1.14 to 1.44, 1.13 to 1.92, 1.29 to 1.82, and 1.36 to 1.82 for the SF(H), CSF(S), CSF(C), and CSF(H) groups, respectively (Table 4 and Fig. 8).
Visalvanich and Naaman (1983) defined the characteristics of fibers influencing the performance of FRLWAC as the fiber-reinforced index (β) and expressed it as β = gτV f L f /d f , where g, τ, and L f /d f refer to the efficiency factor of discontinuous fibers, interfacial fiber bond strength, and fiber aspect ratio (S f ), respectively. It is difficult to evaluate the factor g independently as g is dependent on several parameters, such as fiber orientation and/or dispersibility, mean embedded length after matrix cracking, and contact area between the fiber and cement matrix. We devised a modified formula for β using S f , V f , and τ that measured in Sect. 4.1 for each fiber type. Fig. 8 shows the results of nonlinear multiple regression (NLMR) of f sp /(f sp ) N using modified S f , V f , and τ as parameters. The exponent for each variable shown on the X-axis represents the weight of the parameter. The V f and S f weights were greater than that of τ for the  Therefore, it is necessary to treat S f , which affects the number of individual fibers, as a variable as important as V f in determining the mechanical properties of FRLWAC. From NLMR analysis using test data sets (Fig. 8), the f sp value for FRLWAC can be optimized as follows:

Flexural Performance
Fig . 9 shows the typical load-deflection behavior of LWAC reinforced with different fiber types at different V f . There were no data of the control fiberless LWAC after the peak load, because it failed after the initial flexural crack appeared. The descending section of FRLWAC load-deflection curve in Fig. 9 shows the effect of each fiber on the tensile and ductility properties. Many researchers, including Meng and Khayat (2017) and Gesoglu et. al. (2016), have reported that the slope of the descending branch decreases as the bridging capacity (2) or V f of the fiber increases. The beam with V f = 0.5% showed a similar performance, because the effects of the fibers were small. On the other hand, when V f was 1.0 or 1.5%, the behavior of the descending section was significantly different depending on the fiber type. The CSF(H)-1.5 beam showed slight hardening or plastic flow performance after initial cracking, but the CSF(S)-1.5 beam showed a very short hardening stage. The CSF(H) group and V f = 1.5% group revealed that the ultimate load (P n ) was higher than the initial cracking load (P cr ) by 4.2-18.1%, and the hardening performance was observed after P cr . This results indicated that hooked-end shape fibers were most efficient in enhancing the ductility of FRLWAC. Fig. 10 shows typical images of the interaction between the cement matrix and steel fibers across a flexural crack, captured for Beams CSF(H)-1.0, SF(H)-1.0, and CSF(S)-1.0. All the beams failed along the primary crack developed at the mid-span, indicating no crack distribution. Most straight shaped fibers were pulled out from the cement matrix, resulting in a wider crack mouth opening displacement (CMOD) in the CSF(S) group, when compared to the other groups using fibers with a longer length. Hooked-end fibers across the crack effectively restricted crack propagation even at the post-peak response because of their good sliding resistance with cement matrix.  When evaluating the flexural performance of FRLWAC, the elastic section before reaching P cr is significantly affected by the strength of the concrete and τ of the reinforcing fibers. However, it is reasonable to assume that most of the τ between the fibers and matrix, which plays a bridging role in the fracture plane, is already lost after initial cracking. The geometric shape of the steel fiber is considered to have a significant influence on determining the ductility capacity and P n of FRLWAC after reaching P cr . The crack opening displacement of a specimen containing deformed fibers may increase more than that containing a non-deformed fiber owing to the straightening of the deformed shape. The f r of the CSF(C)-1.5 beam is believed to be relatively low because of this mechanism. Fig. 12 shows the NLMR analysis results of the flexural strength (f r /(f r ) N ) of FRLWAC, which was normalized based on the results of the fiberless control concrete. Similar to the previous analysis of f sp /(f sp ) N , the increasing rate of flexural strength of FRLWAC was predicted using the modified formula for β. The weights of V f , S f , and τ, set as parameters of prediction, were 0.6, 0.4, and 0.5, respectively. The f r values for FRLWAC can be optimized via NLMR analysis as follows: (3) f r = 1.53 V 0.6 f · S 0.4 f · τ 0.5 0.53 · (f r ) N .

Modulus of Elasticity
Owing to the low modulus of elasticity (E c ) of the aggregate, E c of LWAC was 25-50% lower than that of NWC with the same compressive strength. This tendency became more prominent as the LWA content increased. Mehta and Monteiro (2017) suggested that fiber reinforcement in LWAC is not an appropriate solution for improving E c , especially when V f is low. Some studies (Campione et. al. (2001);Shafigh et. al. (2011)) reported that E c could be increased by 6-30% when V f of FRLWAC was increased. Fig. 13 shows the NLMR analysis results for the normalized modulus of elasticity (E c /√f c (28)), where f c (28) represents the compressive strength measured at the age of 28 days. The experimental values of E c /√f c (28) for FRLWAC were between 2607 and 2713. When E c /√f c (28) values were predicted for LWAC using the equation given in ACI 318-19, the predicted values ranged between 1959 and 2443; E c = ρ c 0.043√f c (t), where ρ c is the oven-dry density of the hardened concrete. When comparing the test results of FRLWAC with the predicted values of LWAC, we determined that   (28) values of the V f = 0.5% group were similar regardless of the fiber type (Fig. 13). However, when V f = 0.5%, E c /√f c (28) values of CSF(S)-1.0 and 1.5 were considerably higher than those of the deformed shapes. It is believed that the effect of geometrical deformation of the fiber on the slope of the elastic section of the stress-strain curve is relatively small. However, it can help improve the tensility and ductility of FRLWAC. It is possible to optimize the E c value of FRLWAC through NLMR analysis (Fig. 13). This is more accurate for FRLWAC with a deformed shape: