Flexural Behavior of Lattice Girder Slabs with Different Connections: Experimental Study

,


Introduction
Precast concrete (PC) structures are advanced structural systems that save construction time, formwork cost reduction, labor work, and energy. Globally, the PC structure is becoming a prevailing structure, and its construction is an important part of green and sustainable development. Regarding the structural system, the PC structures can be classifed into precast frame system [1], precast shear wall system [2], and precast frame-shear wall system [3]. Te precast slabs are the essential moment-bearing structural components that are primarily subjected to static gravity loads, no matter in which structural system. Nowadays, there are various precast slabs, such as the hollow-core foor, composite slab, solid prestressed composite slab, and so on.
Lattice girder slab (LGS) is one of the most commonly used composite slabs (Figure 1). In the LGS system, the precast reinforced plank is temporarily propped in situ and serves as the permanent formwork for the fnal casting on the top. Te usual thickness of precast plank is about 50∼65 mm. Te bottom foor reinforcement was placed in the precast plank, where the well-spaced 3D lattice girder trusses protrude. Te truss element can ensure the stifness of the precast plank during transportation and hoisting and the capacity of the precast plank during in situ casting. It can also enhance the interface between the precast plank and the CIP concrete topping. Te precast reinforced plank can have the benefts of free-fromwork on-site and the lightweight for delivery, hoisting, and installation. Moreover, it can also make it convenient to insert mechanical and electrical components together with the top layer of reinforcement before casting the top layer.
Usually, the maximum dimension of precast slabs is restricted by the defection limit of the precast plank and the transportation capacity. Hence, it needs connections to complete the slab system in the structure with pieces of precast planks. Especially for two-way slabs, the connection bearing capacity is essential as the joints are inevitable in one direction of the two-way slab. Eurocode 2 does not speculate on connection details but emphasizes that the joint bending resistance is based on the capacity of the crossbar across the joint and the available lever arm between the crossbar and the concrete compression zone at the top of the cross section [4]. Connections in LGS are specifed in the German and Chinese codes. DIN1045 [5] suggested the straight bar lapping connection with two closely attached precast planks for the two-way slab (Figure 2(a)). Te truss-type reinforcement is evenly distributed within the lapping zone of standard splice length plus 100 mm. As the crossbar is right upon the surface of the precast planks, the joint bending resistance is reduced with the reduction of efective height. Hence, such a connection is not recommended for the transverse joint in the one-way slab in the Chinese code (GB/ T 51231-2016) [6]. In the Chinese code, it is suggested that the CIP concrete strip connection with the lapping system of two longitudinal bars protruding from the precast planks for the joint needs bending resistance (Figure 2(b)). However, the CIP concrete strip connection construction needs additional formwork on site. Terefore, the development of advanced free-of-formwork joints capable of load-bearing is essential for improving the construction efciency of the precast slab system.
In the existing studies, the fexural performance of LGS was a major concern and has been investigated from the perspective of the performance in the construction stage and the ultimate stage. During the construction stage, the stressing in the precast plank is complex as it experiences three diferent loading stages: (a) propped before pouring concrete topping, (b) propped while pouring concrete topping until the concrete reaches the required strength, and (c) non-propped after the concrete topping hardens [7]. To better understand the composite behavior during construction, Newell et al. monitored the early-stage performance of two-way LGS in a fve-story building [8,9]. Te connections with closely attached precast planks (Figure 2(a)) were adopted in the LGS system. Te oneyearreal-time monitoring detected the concrete strain profle before, during, and after pouring the structural concrete topping and demonstrated that the two-way LGS system with the connections could exhibit elastically without cracking at the stage of construction as expected.
Te majority of studies focused on fexural performance at the ultimate stage. Since the fexural resistance for twoway action in labs under the plastic hinge line mechanism can be derived from the corresponding one-way fexural behavior of the slab, most parametric studies on the jointed LGSs were conducted on the one-way fexural behavior. Lundgren evaluated the fexural performance in the fullscale LGSs [10]. Te author adopted the connections illustrated in Figure 2(a) but without truss elements within the lapping zone in the midspan. Te corresponding failure mode was the rupture of reinforcement with a single fexural crack, since the slabs had low reinforcement (0.15%). From the experimental and numerical analysis results, it can be found that the joint without reinforcement would raise the risk of premature brittle failures, as was obtained in similar tests by Tim Gudmand-Høyer [11]. Te author also concluded that the joint with the connection in Figure 2(a) is sensitive to fexural cracking and causes brittle failure. Kim and Shim [12] investigated the fexural behavior of halfprecast slabs (longitudinal reinforcement ratio between 0.4%∼0.67%) for bridges with the loop connection in the midspan and demonstrated that the loop connection was an efective connection for the slab with two closely attached precast planks. Te cracking load, fexural capacity, and defection could be comparable to the slab without joints. Chen et al. [13] investigated the bearing performance of lapsplice connections with additional crossed bent-up rebars (as suggested in Lundgren [10]) in the LGSs (longitudinal reinforcement ratio between 0.61%∼1.45%). Teir fnding was that the bent-up rebar could enhance the interface and anchorage efectiveness of the lap-splice rebar. Hence, the joint could well transfer the force as the monolithic joint. Te above-reviewed research focused on the bridge slabs having a thickness of over 200 mm, and the corresponding CIP layer was much thicker than that in the LGSs for buildings. Among the studies on the LGSs with thin thickness, Liu et al. [14] adopted the connection detailing as in Figure 2(a) in the LGSs for building structures. Tey experimentally investigated the efect of slab thickness, truss element spacing within the lapping zone, and lap length of crossbar on the fexural behavior of one-way LGSs where the connections were in the middle. Te test results presented that the truss element within the lapping zone could improve the interface bond and avoid signifcant detachment as concluded in Stehle et al. [15]. However, the variation of truss element spacing, slab thickness, or lap length of crossbar had a limited efect on load capacity enhancement compared to the LGS without joints. Te highest load capacity was 18% lower than the load of the control specimen. Further improvement still needs more reinforcement which may complicate the on-site construction. Ding et al. [16] improved the CIP concrete strip connection in Figure 2 with the bent-up bar to decrease the CIP strip width from 300 mm to 60 mm. Te bearing capacity of the jointed slab was close to that of the monolithic slab. However, the formwork below the joint was still needed. According to the existing studies on the fexural behavior of jointed one-way LGSs, the load-bearing connection without formwork still needs improvement.
Moreover, several researchers studied the fexural behavior of two-way LGSs with joints. Zajac et al. [17] compared the fexural behavior of four diferent two-waysemiprecast concrete slabs under short-term and long-term loading schemes. Te common feature was the lack of joint reinforcement. Te test observation showed that the LGS without joint reinforcement signifcantly afected the cracking morphology and the maximum defections under the long-term loading conditions. Te typical defection shape for slabs of the longitudinal joints without reinforcement resulted in a diferent load distribution factor compared with the in situ slabs [18]. Furthermore, other researchers made eforts to fnd a way to achieve monolithic equivalence for jointed slabs. Zhang et al. [19] conducted a bending test on a full-scaletwo-way LGS (5 m × 5 m) under a uniformly distributed load. Te connection regulated in the DIN1045 code [5] (Figure 2(a)) was designed in the middle joint in one direction. Te test results showed that the cracking pattern in the LGS was close to that in the monolithic slab, and the numerical study showed that the crossbar across the joint had a slight efect on increasing the stifness of the slabs. Chen and Shen [20] introduced the lattice girder truss and the crossbar across the joint to further improve the fexural rigidity and capacity of two-way LGSs, while the performance was not comparable to the CIP slabs in the study. Moreover, other load-bearing connection types have been investigated in other slab systems. Irawan et al. [21] investigated the two-wayhalf-precast slab with two joints in one direction to decrease the bearing demand of connection. Each joint used the triangular rigid connection where a groove with triangular section was formed after assembling precast planks, and bent-up rebar was used as the crossbar. Such a connection has been demonstrated to be equivalent to the monolithic joint in [22]. Irawan et al. concluded that the jointed slab could reach a similar strength as the monolithic slab but had smaller deformability. All in all, there was limited study considering the connection type efect on the fexural behavior of two-way LGSs.
Among the existing studies, it is demonstrated that the improved connection types for LGSs were rather limited. Te most efcient way is adding truss elements or using bent-up bars within the lapping zone to avoid interface detachment, while the enhancement is rather limited as the sectional bending capacity of the joint has not been increased. Te jointed LGSs cannot be compatible with the monolithic slab or LGS without joints at the ultimate stage, and hence the span of jointed LGS will be limited. In this study, the efect of the number of low-capacity joints (straight bar lapping connection) on the fexural behavior of LGSs was investigated. Additionally, two additional loadbearing connection types (loop connection and straight bar lapping connection within the keyway) were proposed for the LGS. Accordingly, seven full-scaleone-way LGSs were designed to be tested under the static four-point bending test. Two of them were the control specimens, and the remaining fve specimens were designed to consider the two factors mentioned above and divided into two groups. Te overall response and failure mode were observed during the test. Afterward, the load versus midspan defection, defected shape, and characteristic load capacity were analyzed and discussed. Te results can help enrich the connection types for LGSs for construction convenience and mechanical effciency and provide reference for the design of twoway slabs.

Specimens. Seven one-way
LGSs with identical dimensions were designed based on a multi-rise residential reinforced concrete building prototype, which was designed according to the Chinese concrete structure code (GB 50010-2010). Te slab length, width, and thickness were 3000 mm, 780 mm, and 140 mm, respectively ( Figure 3). Te thickness of precast planks was 60 mm. Te surface of the reinforced plank was roughened to achieve a strong interface bond between the precast plank and 80 mm concrete topping. Te concrete of precast planks and CIP toppings were designed as the same concrete grade (C30) with the same mix. Hence, one group of concrete cubes for obtaining the concrete strength and modulus was prepared after casting CIP topping, with consideration of the load bearing of slab largely related to the compression strength at the compression zone. Te average cubic compressive strength and elastic modulus  Figure 2: Two types of connections for LGS. (a) Closely attached precast planks [5]. (b) CIP concrete strip [6].

Advances in Civil Engineering
were 33 MPa and 32.1 GPa, which were tested during the test. Te longitudinal and transverse reinforcement at the precast plank was D8@150 (reinforcement diameter was 8 mm with a spacing of 150 mm). Te same reinforcement ratio was adopted in the CIP layer. Te reinforcements were HRB400 rebars having yield strength of 454MP, ultimate strength of 612 MPa, and elastic modulus of 206 GPa, obtained by the standard tensile test. It is worth to note that the reinforcement design also meets the requirements of ACI 318 and Eurocode 2. Seven specimens were divided into three groups. Te frst group was the control group, including LGS01 and LGS02.
LGS01 was the LGS without a joint. Its fexural behavior is the reference of LGS with joints.
LGS02 was designed as LGS with the CIP strip connection in the middle. Te CIP strip connection had a 300 mm width CIP strip and adopted straight bar lapping with the lap length of l a (Figure 4(a)), which is the typical monolithic joint with load-bearing capacity for the LGS as regulated in the Chinese code (GB/T512321-2016) [3]. It is the most commonly used joint detailing in engineering practice in China, and its behavior is the direct reference for the LGS with other types of connections. Except for two control specimens, the remaining two groups were the LGSs with connections of closely attached precast planks. Tey were divided into two groups. In Group A, LGSA1 and LGSA2 used the straight bar lapping connections with the lap length of 1.4l a (Figures 4(b) and 4(c)). Teir diference was the location of joints. LGSA1 had the joint in the midspan, similar to the confguration in the existing studies, while LGSA2 had two joints, each at one-third of the slab length and away from the loading point. Group B includes LGSB1, LGSB2, and LGSB3. All of them had the midspan joints but had diferent connection detailing. LGSB1 had a rectangular loop connection ( Figure 4(d)). Tis connection has been successfully used in the precast concrete shear walls as the vertical connection. Te lap length could be reduced to 200 mm, 70% of l a , which complied with the relevant codes for loop connections [4,5,23,24]. Te bent-up legs in the loop could enhance the monolithic behavior of the composite slab by providing shear resistance between the precast and CIP plank interface as the lattice girder trusses. Terefore, the number of truss elements within the joint is reduced by half, resolving the clashing rebar issue. For LGSB2 and LGSB3, the connections were improved by introducing keyways evenly spaced within the lapping zones of each precast plank (Figures 4(e) and 4(f)). Each keyway was 30 mm in depth, 80 mm in width, and 400 mm in length on one side of the precast plank. When two precast planks were assembled in place, two crossbars (D10, its yield strength and ultimate strength were 395 MPa and 614 MPa) were placed in the keyway. For the joint detailing in LGSB2 and LGSB3 (concrete cover for crossbar is 30 mm), it results in the increase of level arm of the crossbar in the keyway which will increase the bending capacity of the joints. In this study, the number of keyways on each precast plank was considered the test variable, that is, two keyways for LGSB2 and three for LGSB3 on one precast plank, respectively. Teoretically, when the keyways are right above the longitudinal reinforcements in the precast plank, all of the level arms of crossbars will increase. However, it will cause the edge of the precast plank to be damaged during transportation and construction. Terefore, the limited number of keyways on each precast plank was considered in the study. It is to be noted that the reinforcement ratio of LGSB2 is the same as that in the control group. Te ratio of LGSB3 is higher but with the same ratio in one keyway as that in LGSB2. Detailed information of all the specimens is listed in Table 1.

Test Program and Measurement
. Te fexural behavior was tested by the four-point bending test method. Te slab was simply supported with a span of 2750 mm. Moreover, the length of the pure bending zone was 1000 mm ( Figure 5(a)). Te test was frst controlled by the load with the constant increment of 2 kN before cracking. Afterward,  the load increment per step increased to 4 kN (each step interval was 10∼15 min varying due to cracking observation). Te test scheme follows the sequence as regulated in the Chinese standard for test method of concrete structures (GB 50152-2012). Te load rate was 0.3 kN/min. Te test program was changed to displace control (0.6 mm/min) when the displacement began to have a fast increment during the load increment. Te test would be terminated when the load decreased to 20% of peak load at least. Eight linear variable diferential transducers (LVDTs) were attached to the slab, as illustrated in Figure 5(b). Tree pairs of LVDTs detected the defection at two loading points and the midspan. Another two LVDTs monitored the deformation at the supports to calculate the defection within the span relative to the end. Te load and defections were recorded by the data log during the test synchronously. Meanwhile, the cracking development was observed with the assistance of the crack monitor to measure the crack width.

Test Results and Discussion
3.1. Overall Response. All of the specimens had the fexural failure mode. Te signifcant diference was cracking development which was elaborated as follows.

Control Group.
Initially, the defection of the slab, LGS01, slightly increases with the load. Te fexural cracks in the midspan were observed when the load increased to 12.3 kN. No detachment was displayed in the precast plank and CIP topping interface. Te fexural cracks developed upwards and crossed the interface with the load increase. More new cracks were generated and evenly distributed in the bending moment zone. After the load increased to 20.1 kN, the load control was switched into displacement control. No new fexural crack could be observed. Te existing cracks extended upwards, and the crack width increased with the load. Te observed maximum crack width was 1.2 mm in the precast plank. Te test was terminated when the load began to decrease at the displacement of 53.5 mm. Te fnal crack pattern is illustrated in Figure 6(a).
Te slab, LGS02, had a similar failure process as LGS01, except for cracking development. No crack was observed within the 300 mm CIP strip. Te initial crack was generated beside the CIP strip in the precast plank section, which further trigged the horizontal interface cracks. Afterward, more fexural cracks developed, similar to those in the LGS01 but away from the middle. Figure 6(b) illustrates the crack pattern of LGS02. Te maximum crack width was 1 mm. Te load began to decrease when the midspan defection reached 53.5 mm. At last, the test was terminated when the midspan defection was 54.5 mm.

Group A
(1) LGSA1. Te initial fexural crack was generated in the midspan of the CIP topping and above the gap between two precast planks when the load increased to 7.8 kN. Te  Advances in Civil Engineering cracking load was smaller than LGS01 and LGS02 due to the reduced efective cross section in the middle. At the load of around 12 kN, cracking occurred in the interface starting from the midspan. Afterward, an additional fexural crack was generated around the location of the lattice girder with the extension of the interface crack. Meanwhile, a new fexural crack was also observed in the precast plank (Figure 7(a)). With the load increasing, more cracks were generated in the precast planks out of the lapping zone and with a uniform crack spacing, which was much larger than that in LGS01 and LGS02. Meanwhile, three more fexural cracks with smaller spacing were generated in the CIP topping in the midspan where the interface bond deteriorated. Te defection developed faster when the load increased to 22 kN when the load program was changed to displacement control. Te interface opening at the midspan increased with the extension of the interface crack during the load (Figure 7(b)). At this stage, no new crack could be observed. Te fnal crack pattern is illustrated in Figure 7(c). Te maximum fexural crack (above the joint) width was 1.6 mm. Finally, the load reached the peak value when the defection increased to 75.35 mm.
(2) LGSA2. Te initial fexural crack was generated in the midspan of the CIP topping at the load around 8 kN; afterward, the interface crack was observed. At the load of 14.1 kN, two additional fexural cracks next to the location of truss girders were developed in the CIP topping of two connections after the extension of interface cracks (Figure 8(a)). Cracking development in the precast plank was observed when the load increased to 20 kN (Figure 8(b)). Te load control program was shifted to the displacement control when the load increased to 35 kN. Te slab began to have excessive development in defection and slow increment in load capacity. No new crack was generated, and the crack width increased faster in the connecting region than in the area between two connections. Te maximum fexural crack was at CIP layer above one joint. Te width observed was 1.2 mm, slightly smaller than that in LGSA1. Finally, the test was stopped when the load began to drop at the defection of 60.8 mm. Te fnal crack pattern is shown in Figure 8(c).
To sum up, LGSA1 and LGSA2 had a signifcant difference in the crack pattern compared with the two control specimens. In LGSA1 and LGSA2, the cracks in the CIP toppings within the lapping zones had three fexural cracks associated with interface cracking between two adjacent cracks, and the spacing of adjacent fexural cracks was determined by the location of the truss girder in the lapping zone. Due to smaller cross section stifness within the lapping zone, the crack spacing was smaller than that out of the lapping zone. Te connection moving apart from the midspan alleviated the cracking in the pure bending zone during the loading and reduced the cracking opening of interface crack at the connections. Terefore, the change of connection location resulted in LGS02 having lower curvature in the pure bending zone, the defected shape of which is demonstrated in Section 3.3.

Group B
(1) LGSB1. At the load of 9.6 kN, the interface crack was observed next to the panel joint. Te initial fexural crack was observed in the CIP topping at the midspan when the load increased to 11.3 kN (Figure 9(a)). Te development of cracks in the precast planks had not been recorded until the load increased to 16 kN (Figure 9(b)). Tese cracks extended into the CIP topping, as observed in the other composite slabs. After no new crack was observed, the defection developed faster, associated with the crack width increase. At this stage, the displacement control controlled the test until

Advances in Civil Engineering
the load dropped at the defection of 65.5 mm. Te crack pattern is illustrated in Figure 9(c). Te critical crack was the fexural crack in the midspan. Te maximum crack width was 1.4 mm. Compared with LGSA1, LGSB1 had a similar crack pattern but a relatively shorter length and width of interface crack.
(2) LGSB2. Te initial fexural crack was frst observed in the CIP topping at the midspan at the load of 7.4 kN. Simultaneously, the interface cracking was also observed (Figure 10(a)). When the load increased to 16 kN, more fexural cracks were generated in the precast plank. Also, these cracks crossed the interface and developed in the CIP topping when the load increased to 20 kN. Meanwhile, a few cracks were generated within the shear span. After the load reached 26 kN, no new crack was observed, and the load mode was changed to displacement control mode. Meanwhile, a signifcant extension of fexural crack was observed around the loading points ( Figure 10(b)). After the load began to stop at the defection of 69.1 mm, the fnal crack pattern was recorded, as shown in Figure 10(c). Te maximum fexural crack was located in the midspan. Te crack width was around 2 mm.
( 3) LGSB3. Te specimen had an identical failure process as LGSB2 but had diferent critical load values corresponding to the specifc stage. Te initial fexural and interface crack load was around 9.6 kN (Figure 11(a)). Te fexural cracks in the precast plank were observed at the load of 18 kN and developed into the CIP topping when the load reached 24 kN (Figure 11(b)). Similar to LGSB2, a few cracks were observed within the shear span. After the load reached 31 kN, the load program was controlled by the constant displacement rate until the failure of the specimen at the defection of 60.0 mm. Te fnal crack pattern is illustrated in Figure 11(c). Te maximum crack was the fexural crack under one loading point. Te crack width was around 1.4 mm. All in all, three diferent connection detailing measures had not caused a signifcant diference in the crack patterns. Similar to the specimens in Group A, the "z" shape crack pattern was observed at the joints. Te development length of interface cracking in LGSB1, LGSB2, and LGSB3 was rather smaller than that in LGSA1. Especially, the interface cracking was signifcantly restrained in LGSB1 due to the relatively signifcant dowel efect from the rectangular loops. Moreover, there were fewer fexural cracks in the specimens of Group B than LGSA1. Te setting of keyways signifcantly reduced the fexural cracks as its efect enhanced the efective height and stifness, which also improved the fexural resistance. Terefore, a few shear-fexural cracks were generated in LGSB2 and LGSB3. Te number of keyways had no signifcant efect on the crack pattern but had a signifcant efect on the load-defection behavior. A detailed explanation will be presented in the following sections.

Load versus Midspan Defection. Te fexural behavior of
LGSs with diferent connections was evaluated and compared by the load-defection curves (Figure 12), where the load referred to the data from the load cell under the hydraulic jack (Figure 5), and the defection was the midspan deformation corrected by deformation at two supports. All Advances in Civil Engineering the curves exhibited three-stage behavior. Initially, the uncracked slab had the linear load versus defection relationship, determining the initial stifness. From the fgures, it could be found that diferent connection confgurations had a minor efect on the initial stifness. After it cracked, the slope of the load-defection curve began to reduce. Tis stage stopped at the yield point, determined by the graphical method (see Section 3.4). After that point, the curve development stepped into the third stage when it had excessive defection and minor load increment. Two control specimens had identical load-defection curves. Terefore, the CIP strip (LGS02) connection enabled the slab to work as the monolithic LGS. Among the LGSs (Groups A and B) with the connections of closely attached planks, LGSA1 had the lowest fexural stifness and capacity Te reduction of cross section height in the midspan caused the sectional fexural stifness to be reduced by 81.3% (cross section height reduced from 140 mm to 80 mm), which further induced the excessive curvature increment in the joint and detachment along the interface initiated from the gap (Figure 13). Such a phenomenon has also been observed in other references [7,16,20]. Te interface detachment would reduce the bond transfer length and limit the stressing in the crossbar. Comparatively, LGSA2 had a smaller defection and sectional curvature in the joint. It reduced the interface normal stress and detachment length as shown in Figure 8(a). Te ultimate stress in the crossbar could be less restricted by the interface detachment, which increased the ultimate fexural capacity compared to LGSA1. However, the capacity of LGSA2 was still less than the monolithic slab as the efective height of the joint section was smaller.
Te load-defection performance of the LGSs in Group B was much better than that of LGSA1. For LGSB1, the detailing of the rectangular loops improved the interface resistance through the dowel efect, which further restrained the interface cracking and alleviated the deterioration of load transfer in the lapping system. Hence, the connection's fexural capacity could be improved compared to LGSA1. In comparison, the detailing in LGSB2 was designed to increase the efective cross section height at the joint with the keyway and increase the load capacity. Te results showed that two types of connections could make the LGSs have close loaddefection curves and similar enhancement efects. However, the improvement was not as much as that in LGSA2. Te load capacity was still lower than that of control specimens. Furthermore, the design of LGSB3 considered the efect of increasing cross section area in the transverse direction with the increase of keyways. Te comparison between LGSB2 and LGSB3 showed that increasing keyways which resulted in the increase of reinforcement ratio could signifcantly enhance fexural performance. Te load-defection curve of LGSB3 could be comparable to that of LGS01 and LGS02. Moreover, the enhancement could also be attributed to the increment of the reinforcement ratio from 2.8% to 5.7%. Te increment was to ensure the same detailing in each keyway between LGSB2 and LGSB3.

Defected Shape of the Slabs.
Based on a group of LVDTs installed at diferent positions ( Figure 5), the defected shape of the slabs could be drawn at a specifc load step, as shown in Figure 14. Herein, two load stages were selected for comparison, the yield load and the peak load stage. Te former was regarded as the ending of the elastic stage, where no signifcant non-linear deformation occurred. Te determination of the yield point has been elaborated in Section 3.4. Te defection ratio of the average defection at 1/3 and 2/3-span to the midspan defection was defned to evaluate the defected shape diference (Table 2). Accordingly, the closer to 0.63 the ratio is, the more linear the defection curve between the support and midspan is, demonstrating the midspan section forming a hinge.
At the yield load stage, two control specimens had much larger defection and higher defection ratios compared with other specimens. In Group A, LGSA1 had the lowest defection ratio closer to 0.63. Te excessive deformation concentrated in the middle was caused by the fexural stifness of the joint section which was much smaller than that of the section out of joint, which took efect at the service load stage. When the joint moved to the 1/3-span at each side, the excessive deformation within the joint was signifcantly alleviated. Te defection of LGSA2 at the joints turned closer to the defection of LGS01 and LGS02 at the same location. On the contrary, the defected shape pattern of the specimens in Group B was similar to that of the control specimens. Although the defection values were smaller than those in LGS01 and LGS02, the defection ratios were closer to the ratios in the two control LGSA1 LGSA2 LGSB1 LGSB2 LGSB3 LGS01 LGS02 yield load

12
Advances in Civil Engineering specimens. Te defection ratio of LGSB1, LGSB2, and LGSB3 was 9.5%, 5.9%, and 4.8% lower than the ratio of LGS01, respectively. Te connection detailing improvement had enhanced the joint fexural stifness and alleviated the stifness diference along the span before the rebar yield. At the peak load stage, the defection diference among LGSs in Groups A and B and control groups turned out to be minor compared with the service load stage performance. Two control specimens, LGS01 and LGS02, had identical defected shapes with the defection ratios of 0.82 and 0.83, respectively. Te ratio had no signifcant diference for specimens at service load and peak load stage. It further demonstrated that the CIP strip joint between precast planks could enable the jointed LGS to achieve monolithic behavior as the LGS without joints. Regarding the jointed LGS with closely attached planks (LGSA1 and LGSA2), the defection at the midspan was higher than that of LGS01 and LGS02.
LGSA1 had the highest defection at the midspan and the lowest defection ratio. Te joint location adjustment in LGSA2 could alleviate the excessive deformation in the midspan by setting two joints away from the midspan. Its defection ratio was improved by 25% and was 3.6% lower than that of LGS01. In Group B, the connection detailing improvement had increased the defection ratio of LGSB1, LGSB2, and LGSB3 by 7.9%, 19.0%, and 25.4% compared with LGSA1, respectively. Te most efective improvement was LGSB3, which had a defected shape comparable to LGS01 and LGS02. In other words, the efective cross section depth increase was the most efcient way to recover the fexural stifness comparable to that in the LGS without joints. LGSA1 LGS01 LGSA2 LGS02 LGSB1 LGS01 LGSB2 LGS02 LGSB3 LGSA1 LGS01 LGSA2 LGS02 LGSB1 LGS01 LGSB2 LGS02 LGSB3

Characteristic Load and Corresponding Defection.
As described in Section 3.1, the load-defection performance could be evaluated by the cracking load, yield load, peak load, and corresponding defection. In this study, since the rebar strain gauges were out of work during the test, the yield load was determined by the farthest point method proposed by Feng et al. [26] (illustrated in Figure 15). It can be found in Figure 12 that the predicted yield point was also the turning point which was consistent with the feature observed in other studies by the yield strain in the rebar [27][28][29]. In general, all the characteristic values of the specimens are summarized in Table 3.

Cracking Load.
Te cracking loads of LGS specimens in Groups A and B were between 8 and 10 kN, 22 to 40% and 37 to 51% lower than those of LGS01 and LGS02, respectively. It can be attributed to the reduction of efective cross section height and the fexural stifness in the specimens of Groups A and B. LGSB1 and LGSB3 had a slightly higher cracking load (fexural cracking associated with interface cracking) due to the dowel efect of loop reinforcement and higher reinforcement ratio, respectively. In general, the diferent connection confgurations for closely attached planks had a minor efect on the cracking load.

Yield and Peak
Strengths. Two controlling specimens had similar yield and peak strength with a 2 to 4% diference.
LGSA1 had the lowest yield and peak strength compared with other specimens. Te yield and peak values were 45.6% and 32.2% lower than the corresponding average values in the control group. When the connection confguration, the same as in LGSA1, was set at one-third and two-thirds of the span, the yield load and peak load were signifcantly increased by 65.9% and 36.9%, respectively. Moreover, the diference to that of controlling specimens was reduced to be within 10%. In Group B, LGSB1 and LGSB2 had close yield and peak strength with a diference of 6.5% and 0.9%, respectively. Tey were around 25% and 17% lower than the average yield and peak strength of LGS01 and LGS02. Compared with the connection type in LGSA1, the rectangular loop connection and keyway confguration enhanced the yield and peak load capacity by over 30% and 20%, respectively. Te enhancement of joint bearing in LGSB1 and LGSB2 can be attributed to the loop confnement on concrete and level arm increment in the crossbar. Among all the jointed LGSs, LGSB3 had the best performance. Te yield load and peak load were only 7% and 2% lower than the average yield and peak load of slabs in the control group, respectively. Moreover, the cross section analysis was conducted to compare the fexural capacity values between test and analytical values (Table 3). Te strength of the jointed slab is calculated after the determination of the arm level of the crossbar. It can be found that the controlling specimens and LGSB3 could be well predicted. Hence, LGSB3 behaved as a monolithic slab, while the fexural capacities of the other LGSs were underestimated.
LGSA2 and LGSB1's prediction diferences are much larger than the diference for LGSA1. It can further demonstrate that the measures in LGSA2 and LGSB1 can have a signifcant efect on alleviating interface detachment between CIP and precast layers. It is because the interface detachment efect has been fully considered in the analytical formulas, that is, the contribution of the precast layer and the reinforcement has not been considered in the fexural resistance. Further studies will be needed to establish the analytical method to adequately consider the interface performance in the jointed LGSs. Table 3, it can also be found that the connection with higher bearing capacity caused the slab with smaller ductility. In other words, the improvement measures had a more signifcant enhancement efect on fexural stifness, especially for the secant stifness corresponding to the yield point. However, all of the specimens in Groups A and B had ductility ratios larger than 3. Meanwhile, all the specimens in Groups A and B had relatively larger deformability than the control specimens and exceeded the slab defection limit (l/50 � 1/50 * 2750 � 55 mm). Terefore, the improvement of connection detailing had no signifcant adverse efect on the defection performance.

Conclusion
In this study, seven full-scaleone-way lattice girder slabs were designed to investigate the efect of connection position and connection type on the fexural performance of LGSs. Te conclusion is summarized as follows: (1) Te position adjustment of joints with the straight bar lapping connections could improve fexural capacity and defection development. When the joint was set at one-third of the slab length, the capacity could reach 91.1% of LGS01 and LGS without joints and was 30.6% higher than that of LGSA1. (2) Te loop connection could improve the interface shearing resistance through the dowel efect and alleviate the deterioration of load transfer in the lapping system. Consequently, the fexural capacity was improved by 22.5% compared with LGSA1. (3) Te measurement of the keyway could increase the efective height directly and hence the fexural capacity and deformability. Te test demonstrated that the three-keyway measurement with a higher reinforcement ratio could ensure that the slab was comparable to the LGS without joints. Te fexural capacity was 38% higher than that of LGSA1.
To sum up, both the position of joint and reinforcement detailing improvement could improve the fexural performance of one-way LGSs with closely attached precast planks. Te current study qualitatively investigated the fexural behavior of one-way LGSs with diferent improved connections. Numerical studies will be our further work to optimize the connection design for the LGSs by the parametric study from the fnite element model (FEM) with consideration of the non-linearity of concrete and steel and interface behavior between precast plank and CIP topping. Te FEM will further investigate the fexural behavior of two-way LGSs with optimized connections.

Data Availability
Te data used to support the fndings of this study are available from the corresponding author upon request.

Conflicts of Interest
Te authors declare that they have no conficts of interest.