2.1 Reinforced Materials

In this study, bi-directional GFTG (grid mesh size: 25.4 mm × 25.4 mm, clear mesh size: 22.0 mm × 21.0 mm) was employed as the reinforced material, and steel wire mesh (SWM mesh size: 20.0 mm × 20.0 mm) was adopted as the reference strengthened material, as shown in Figure 2. GFTG yarn, a sort of composite material, is composed of fibers and resin; the yarn constitution depends on the material modulus and volume ratio, which implies that if all constituent materials are independent and noninteracting, the relevant composite-material theorem should correspond to the calculation of the elastic modulus of the textile warp and weft. Practically, the interaction between fibers and other materials such as the coating resin inevitably considerably deteriorates the mechanical properties of yarn, such as its modulus and tensile strength. Thus, the parameters of reinforced materials used in this paper study were tested experimentally, as presented in Table 1; the tests included those pertaining to the fineness of the fiber bundle [35], yarn tension [36] and wire tension [37].

Figure 2

Reinforced materials: (a) Glass fiber textile grid and (b) Steel wire mesh

2.2 Ceramsite Concrete

Structural lightweight aggregate concrete, which provides functions of both insulation and structure as adopted. The size of textile grid, penetration, flowability and water demand of concrete should be considered comprehensively; thus, lytag aggregates within a maximum diameter of 5 mm were adopted. The concrete constituents used in the experiment are listed in Table 2, Portland cement is composed of four oxides: calcium oxide (CaO), silicon dioxide (SiO₂), aluminum oxide (Al₂O₃) and iron oxide (Fe₂O₃). The total amount of these four oxides accounts for 95% in clinker, and the mixture design is defined in Table 3. The compressive strength and elastic modulus were determined by performing tests on specimens according to ASTM C39/C39M [38]. The mechanical performances of concrete are summarized in Table 4.

2.3 Core Foam

Polystyrene foam (EPS, density: 22 kg/m³) was chosen as the core material of panel. Usually, the core in PCPs and NCPs functions as the key coordination material for face plates to transfer shear; however, in this study, the structural contribution of core foam is neglected, because of the presence of the continuous lightweight TRC web, whose stiffness is far higher than foam, even when rigid foam is applied.

3.1 Specimen Details

The composite panel section is depicted in Figure 3. The inside corners of panels were designed to be arcs with a radius of 10 mm to reduce the stress concentration. SWM or GFTG as the reinforced material was simply placed in the middle of the face plates and webs. To avoid the complicated anchoring of the textile grid, approximately one plate-width GFTG was overlapped to utilize the bonding force between the fiber yarn and concrete, as shown in Figure 3b. For comparison, steel wire mesh reinforced-concrete composite panels (SMRCCP) specimens were produced in the same manner.

Figure 3

Sandwich panel section and reinforcement: (a) Section and (b) Reinforced material overlapping anchorage

3.2 Specimen Fabrication

Polystyrene foam was served as the inner mold for casting concrete directly. However, because of its high softness, the surface of the foam was plastered with a layer of high strength mortar as a hard shell. After hardening, mortar lumps were stuck onto the shell surface to control the position of the GFTG or SWM. Subsequently, the GFTG or SWM was wrapped up outside the mortar lumps tightly and fixed temporarily on the surface. Next, as the core of the specimen, the entire unit was placed in the center of the mold to be poured over by concrete, vibrated until compaction and cured naturally. The main steps are shown in Figure 4.

Figure 4

Main process for fabrication of panels: (a) high strength mortar layer with mortar lump; (b) wrapping with GFTG; and (c) casting concrete

3.3 Experimental Scheme

The test details are presented in Table 5. A single-layer reinforced material was set for withstanding tension at the bottom, and the overlapped double-layer was served as anchorage. Here, AL and TL refer to the axial and transverse loads respectively. The pseudo-static loading velocity was 0.5 mm/min during the loading progress for the axial and transverse load tests.

The axial load test scheme is shown in Figure 5. The geometric imperfections of panel ends were repaired with high strength mortar to ensure uniform stress. A steel hat and boot with both a thickness of 20 mm were placed at both ends for well-distributed axial force. The connections among the loading frame, actuator and load plate were spherical hinges. Steel bolts were used between the test bed and steel boot. Three units of strain gauges (SGs) were set on each plate surface along the axial direction at the top, middle, and bottom. Five linear variable differential transformers (LVDTs) were set up to monitor the panel's displacement: two for the vertical displacement of the top end, and three for the lateral displacements of the top, middle, and bottom correspondingly. It should be noted that in this study, just mean values of the two top LVDTs were on behalf of vertical displacements. To obtain the sheer panel displacements, the top and bottom lateral LVDTs were attached to the edges of the steel hat and boot, very close to the actual top and bottom shifts, so that the displacements of the steel hat and boot themselves could be excluded.

Figure 5

Axial load test scheme: (a) front view and (b) side view

The four-point transverse-load scheme is shown in Figure 6. Four LVDTs were installed to monitor the vertical displacement of specimens: one for the bottom-plate vertical deflection; and three for the top plate vertical displacements, among which one corresponded to the top deflection, and the other two pertained to the corresponding support sites. The load-displacement data were recorded by a universal test machine with the capacity of 5000 kN.

Figure 6

Transverse loading scheme: (a) front view and (b) side view

4.1 Axial Test

The failure modes, load versus vertical and lateral displacement (L-D) curves, and load versus two-face-plate vertical strain (L-S) curves for SCA1, SCA2, FCA1, and FCA2 under axial load are presented in Figures 7–12. In terms of face-plate test strains, all the face plates were under pressure. The failure modes of SCA and FCA specimens are summarized in Table 6. The lateral top-end displacements of the panels in response to ultimate load (UL) are shown in Table 7, where e is the lateral displacement of the top end and h is the specimen thickness (200 mm). Although the lateral displacements at the top and middle sections were larger than those for the bottom, as shown in Figures 7–12, all the lateral displacements are nearly negligible, and the extra eccentricities due to the axial load can be elided. Considering the extremely low lateral end displacements, the boundary of both the panel-ends can be regarded as simply-supported. Table 8 presents that the key characteristic loads during the axial-load testing process and the corresponding vertical displacements until the ultimate breaking point is reached.

Figure 7

SCA1 failure modes: (a) concrete crushing at the bottom and (b) concrete crushing at the top

Figure 8

FCA1 failure modes: (a) concrete crushing at the bottom and (b) concrete crushing at the top

Figure 9

Load-displacement and strain curves for SCA1: (a) load-displacement curves, (b) load-strain curves

Figure 10

Load-displacement and strain curves for SCA2: (a) load-displacement curves, (b) load-strain curves

Figure 11

Load-displacement and strain curves for FCA1 (a) load-displacement curves, (b) load-strain curves

Figure 12

Load-displacement and strain curves for FCA2: (a) load-displacement curves, (b) load-strain curves

The same type of L-D curves demonstrated similar features. The vertical top L-D curves of the panel are selected for the subsequent discussion pertaining to curve features; the characteristic load points are labeled as uppercase letters A-D in Figures 7–12. At the inception, all the measured-point displacements increase linearly with the load augmentation; next, point A accompanied by a low sound, corresponds to the first deviation load; subsequently, the L-D curves vary with irregular fluctuation. Generally the lateral displacements reverse against the initial state from A to B, which could lead to crossing of the origin point and change the sign of the displacement in the figure; an inflection point even occur in some cases. Before the UL reached point B or E, short plateaus are included in partial curves, in which the displacements increase while the loads are constant until point C, D or F. For loads to the UL, the panels lose their bearing capacities suddenly at point C with a loud sound, and the curve fluctuation strongly; in these cases, the ductility is not significant and points C and F may even be non-existent.

As shown in Table 6 and Table 8, the failure modes and ultimate bearing capacity of SCA were almost identical to those of FCA. Significant concrete crushing at the top and bottom ends along with some vertical cracks, was noted, while the specimen ductility was not conspicuous. Although the deformation capacity of glass fiber textile was considerably larger than that of steel wire, the experimental deformation of specimens (both SCA and FCA) was controlled by the concrete, not the reinforced-material elongation deformability on the condition of axial pressure.

4.2 Transverse Test

Figure 13 shows the failure modes, and Figures 14–15 show load vs. midpoint deflection (L-MD) and L-S curves of the SCB and FCB specimens under the transverse load. The bottom midpoint deflections are marked considering the characteristic points with the uppercase letters A, B, and C for the succeeding discussion. Before A, the L-MD curves of the top and bottom midpoint are almost linear with their slopes decreasing slightly. After A, which corresponds to the first deviation load, the L-MD curves decline rapidly and fluctuate with a crackling sound till the maximum UL at load point B. Next, the load dwindles with the further deflection, and the panels collapse and lose their bearing capacities at C. It should be noted that the L-MD curve of SCB1 demonstrates a strong violation stage significantly between B and C, owing to the partial steel wire fracture occurring in batches. From 0 to A, the face-plate displacements exhibit a trend of slight reversal. Before point C, all panels exhibit obvious nonlinear deformations. Most cracks were perpendicular to the bottom edges of specimens; however, a few cracks were slightly tilted, extending across the entire bottom plate and up to most of the height along the side webs. After the influence from the top plate-end rigid-body displacements is eliminated, the deflections of the bottom are much larger than those of the face. SCB and FCB both exhibit excellent cooperation between webs and face plates, as shown in Figure 15.

Figure 13

Transverse test failure: (a) side crack in SCB1; (b) continuous crack in SCB1 bottom; (c) side crack in FCB1; and (d) side flexural-sheared joint crack in FCB2

Figure 14

Load-deflection curves: (a) SCB1; (b) SCB2; (c) FCB1; and (d) FCB2

Figure 15

Load-strain curves: (a) SCB1;(b) SCB2; (c) FCB1; (d) FCB2

In summary, the main failure modes under the transverse four-point load were wide flexural cracks and broken wires and yarns at the bottom of the web, mingled with certain traits of shear failure. But no evident concrete crushing, spalling, or overlapping-area pull-out occurred. As textile-yarn elastic modulus is far lower than that of steel wire, the deflections of FCBs were larger than those of SCBs. FCB exhibits sufficient bearing capacity more than that of SCB overall. The ductility of FCB was approximately 2–3 times that of SCB, even though it was a type of pseudo-ductile. Table 9 presents that the ultimate loads and maximum displacements under transverse load test.

5.1 Finite Element Analysis

Finite element analysis is a powerful method to simulate the nonlinear behavior of concrete members subjected to all types of load actions. ANSYS software was employed to simulate the loading process on TRCCP with numerical model referred to the abovementioned working conditions. Solid65 element was utilized for concrete, and Link180 was used for steel wire and textile yarn. The element mesh sizes were controlled to match the actual sizes of actual wire and textile, and each element node had three translational degrees of freedom. The boundary conditions were simply-supported. To circumvent the unnecessary calculation divergence, the panel materials, except for polystyrene foam, were considered to be nonlinear during the loading process. Elements’ sizes were 15 mm~30 mm. Hongestad constitutive model and multilinear kinematic hardening rule were adopted for concrete, while bilinear kinematic hardening rule was used for steel wire. Linear-elastic model was used for core foam, whose structural action was neglected because its stiffness and strength capacity were an order-of-magnitude lower than those of concrete, steel wire and GFRP.

5.2 Analysis Method

TRCCP that is highly similar to the traditional RC slab in possess of the RC webs, similar to steel bars, GFRP can withstand higher tensile stress and can have higher adhesion to concrete, fortifies the cooperation between the two face plates and attains full composite in structure formation. Thus, it's expected that analytic method under RC theory can be introduced to estimate TRCCP performance. On the basis of the abovementioned experiments, axial and flexural carrying capacities are discussed, while the shear capacity is excluded because of no occurrence of shearing failure phenomena. The simplified section model for panel calculation is shown in Figure 16, in which the two side webs are combined into one middle theoretical web.

Figure 16

Simplified section of TRCCP for calculation

6.1 Axial load

The face-plate strain curves of specimens, shown in Figures 9–12, indicate that both sides of the test panels were under compression in most cases. Numerical analysis on the load-displacement responses of the sandwich panels was conducted through Finite Element (FE) modeling using software ANSYS. In such finite element model, composite, and core were modeled using Continuum 3-D, 8 node (C3D8R) solid element, and the nodes at the joint are coupled assuming no sliding between the composite materials and the core. Concentrated loading is applied on the model by defining the displacement on the loaded node. The boundary conditions are fixed at both ends. The FEM failure modes for both TRCCP and SMRCCP were material strength failure under a small slenderness ratio, as shown in Figure 17, wherein circles represent the damaged areas. All specimens failed due to concrete crushing and cracking at both panel ends in the FEM analysis, matching the actual test failure modes presented in Table 3 and Figures 7–8. Throughout the entire test process, GFTG could also constrain the concrete and maintain the cooperation of the two face plates and integrity of the panel as similar to the SWM's function.

Figure 17

FEM failure for axial load

The axial UL of panels SCA1, SCA2, FCA1 and FCA2 obtained from experimental tests, FEM simulation and analytic results are presented in Table 10. Figure 18 depicts the axial load vs. vertical displacements for the FCAs, pertaining to the test and FEM results. The ULs predicted with experimental data of SCA1 SCA2 and FCA1 are higher than those with FEM by 2.9%, 9.3% and 6.1% respectively. However, the predicted UL for FCA2 is lower than that obtained from the analytical FEM model by 2.8%. The results for the SMRCCP and TRCCP did not exhibit any substantial differences. The ULs obtained from the analytic method were higher than those from the experimental data by 22.4%, 27.9%, 25.2% and 17.5%, whereas if the design strength of concrete was adopted for comparison, the ULs obtained from analytic method were considerably lower than the test values and could be used to predict the ultimate bearing capacity directly after being fine-tuned. The FEM-obtained L-D curve in Figure 18 exhibits well-defined elastic and plastic stages; however, the curve pertaining to the tests does not exhibit notable yielding plateau and nonlinear stage, but is more analogous to nonlinear elastic behavior instead. The difference between the features of L-D curves obtained using FEM and tests imply that the initial vertical stiffness of specimens, as obtained from tests, is smaller than those obtained from FEM, probably due to the occurrence of the local crushing in advance on the end unevenness at the touching faces between the specimens and experimental set-up.

Figure 18

Axial UL vs. top vertical displacement for FCA

6.2 Transverse load

The FEM failures of SCB and FCB under transverse ultimate loads are shown in Figure 19. The element cracks surrounded the three faces of the panel, which were integrated with in the most height of two-side webs and the entire bottom plate. Among them, one or two key macro-cracks occurred, which extended to the top upper area of the web akin to the practical failure: a type of flexural failure. The FEM failure modes of TRCCP and SMRCCP were remarkably similar in general, which is due to the structures were dominated by flexural shearing failure, the failure load mainly depends on the shear strength of the new material and the flexural cracks near the load position matched with the experimental damage. However, some issues in the case of FEM should be noted: (a) The crack elements were dispersed within a certain width on the webs primarily due to the shear effect. (b) The crack elements at the SCB bottom plate were at a larger angle than those at the FCB plates due to the lower elastic modulus of GFTG yarn. (c) The ductile phase of FEM analysis is shorter than that in the experiment, which is a numeric divergence due to the considerably large displacement experienced during the strong nonlinear stage after cracking.

Figure 19

FEM failure of SCB and FCB under transverse load: (a) SCB side; (b) SCB bottom; (c) FCB side; and (d) FCB bottom

The relevant L-MD curves and ULs from experiments and FEM analysis are shown in Table 11 and Figure 20. The experimental curve for FCB1 displays self-evident intense load fluctuation at the nonlinear stage, while the corresponding ones for the others are not so strong. At the primary loading stage, the loads are correlated linearly with the displacements, and the experimental deflections are generally consistent with the FEM values. With load increasing continuously, the physical panel behaved clearly nonlinearly and exhibited ductile deformation. However, this type of ductility is supposed to be pseudo ductility induced mainly by fiber yarn elongation, especially in the key crack area at the ultimate limit stage when GFTG is still at the elasticity phase. FEM analysis did not simulate the late part of the plateau or nonlinear increment well owing to the oversized translation resulting in numeric divergence. GFTG or SWM both ensured the panel integrity before and after ultimate failure.

Figure 20

Bottom deflections and loads

It should be noticed that the SCB specimens lost their bearing capacity abruptly at the maximum loading, while FCB specimens exhibited a gradient. It can thus be deduced that some reinforced wires broke at the ultimate failure stage, yet fiber yarns did not, as their maximum stresses were considerably different from the tensile strength. Thus, the ultimate failure mode for FCB was controlled mainly by excessive displacement, which also hints at a non-uniform load effect occurring at the same cross section, a type of shear lag effect on the compressed flange of the concrete member and reinforcements.

To evaluate approximately the foregoing effective width within the width of flange, the participation ratios of the tensile flange reinforcements and the concrete stress ratios of the compressed flange surface under the ultimate limit stag were defined on the basis of FEM results. The reinforcement participation ratio (RPR) depicts the force proportions undertaken by tensile line elements on behave of steel wires and textile yarns including the webs and bottom plate at the same cross section. The concrete participation ratio (CPR) indicates the stress distribution on the top surface. The analyzed positions, A-B-C and D-E-F shadow areas for SCB, as shown in Figure 19, were selected in the serious crack regions at the web and bottom plate. The corresponding RPR and CPR results are shown in Figure 21. The participation ratios of tensile wires in SCB were apportioned by several wires near the webs, in which case the wire proportion contained in the webs was approximately 75%. The textile yarns in the FCB had the same traits; however, the bearing proportions of the yarns near the webs were lower than in the case of SCB, slightly over 45%, while the bearing proportions of those in the bottom plate were higher, which means the GFTG stress distribution between the longitudinal yarns is higher than the corresponding values for SWM. The surplus bearing capacity and excellent deformability of GFTG leads to the FCB failure modes, which makes the tension force redistribution likely. If summed along with the bottom plate width within the 80mm, this portion of RPR can be amount to 80%. Similarly, the concrete web nodes on the top plate also withstand a higher compression stress ratio over 4 times as much as the mid compressed flange does. However, the CPR variation rate is considerably smaller than the RPR at the bottom, which indicates the comparatively good cooperation among the compressed concrete areas. Nonetheless, there exists no evident difference between SCB and FCB in terms of CPR. In other words, the area around the web bears the maximum load action, which provides an explanation for the sudden loss bearing capacity of SCB.

Figure 21

CPR and RPR distribution in the panel

Thus, a simplified model for the ultimate limit stage can be defined by the analytic method in combination the test data and FEM results. The analytic method postulates that concrete crushing or bar yielding are the usual prerequisites, which does not represent the actual condition according to the abovementioned discussion. The crushing area in the top concrete plate flange within some width near the web is assumed, and the reinforcement-stress states in the webs and bottom plates are not calculated owing to the considerably lower than the tensile strength. Considering the actual crack extended into the compressed flange area, the limiting-flange height and web width can be respectively considered as the ultimate equivalent depth and width of the compressive zone during flexural analysis. Table 11 lists the ULs obtained from the tests, FEM and analytic method. As above, under flexural ultimate limit stage, the compression flange height can be assumed as the equivalent compression depth, that is, a flange width of 0.5–0.75 in the concrete compression area or GFTG tensile area can participate completely.