Study on the Fire Resistance of Castellated Composite Beams with Semi-Rigid Restraints

Abstract

To study the fire resistance of castellated composite beams with semi-rigid restraints, temperature rise tests with constant loads were performed on two full-scale castellated composite beams with circular holes and semi-rigid restraints to compare the influence of whether stiffeners were set or not under semi-rigid restraints on the fire resistance of castellated composite beams. The results indicate that during the fire, the primary failure mode of castellated composite beams with semi-rigid restraints is the buckling failure of the web and lower flange in the negative moment zone at the beam end. Composite beams with stiffeners exhibited less buckling of the web and lower flange than those without stiffeners; for steel beams without stiffeners, the web and lower flange show overall lateral instability. Following the fire, the composite beams initially exhibit downward vertical deformation. After 5–10 min, when the web temperature is around 500 °C, it matures upward to the initial position. After 50 min, when the temperature of the web is around 800 °C, it starts to deform downward continuously. During the cooling stage, the end plates at the lower flange of the steel beam and the steel column show a separation phenomenon. By comparing the joint deformation and the mid-span displacement, the fire resistance performance of semi-rigid restrained castellated composite beams is better than that of hinged and rigid restraints. Numerical simulation analyses were carried out on the castellated composite beams. The simulation results showed a high degree of consistency with the test results, which effectively validated the accuracy and reliability of the proposed finite-element model.

1. Introduction

Castellated composite beams are castellated steel beams and concrete slabs combined through studs to form a unified load-bearing structure. The castellated steel beams are created by cutting along the web centerline and welding the sections out of alignment. These beams are characterized by their low self-weight and high bending capacity, along with the ability to route pipelines through the steel beam’s web openings. Thus, they have been widely adopted in engineering applications [1,2,3,4,5,6].

Building fires are frequent disasters. When they occur, people’s lives and property are greatly endangered. It is essential to investigate the fire-resistant capabilities of castellated composite beams for enhancing the structural safety of buildings, preventing building collapse, and reducing fire hazards. In studying the fire resistance performance of castellated composite beams, the primary forms of joints are simply supported, hinged, and rigid restraints [7,8,9,10,11,12,13,14,15,16,17]. Nadjai et al. [11] performed fire tests on castellated composite beams and found that the beam with a circular opening failed due to web-post buckling, while beams with elongated openings failed due to Vierendeel bending associated with the buckling of the web posts of the steel section. Naili et al. [12] carried out an experimental and numerical investigation into the behavior of three castellated composite beams featuring circular and elongated web openings under elevated temperatures. The study revealed that for an asymmetrical composite beam, flexural web-post buckling in the moment–shear interaction zone was the dominant failure mode. In contrast, for a symmetrical composite beam, Vierendeel bending accompanied by the buckling of web posts emerged as the main failure mechanism. Nadjai et al. [13] carried out fire-related experimental research on full-scale castellated composite beams, both unprotected and protected with intumescent coatings of varying sizes and opening shapes. The experimental results showed that the failure of these castellated beams was attributed to web-post buckling, as well as Vierendeel bending that was connected to the buckling of the web posts in the steel section. Sunar Bükülmez et al. [16] performed fire tests on four full-scale simply supported composite beams, demonstrating that the primary failure mechanism was web buckling between openings in protected and unprotected castellated composite beams. They also observed that more severe damage occurred near the supports than in the mid-span region. Li [17] performed fire tests on castellated composite beams with different hole shapes, thicknesses of concrete slabs, and beam-end restraint forms and found that the hole shape and slab thickness have little influence on the bearing capacity of castellated composite beams at high temperatures. The extent of the plate surface inclination of castellated composite beams with rigid restraints is smaller than that of castellated composite beams with hinged restraints; the web and lower flange at the first opening at the beam ends are severely buckled. Based on the above research findings, the failure of castellated composite beams is mainly caused by the web’s buckling, and the web’s buckling at the beam end is more severe. Therefore, in this paper, the castellated composite beams are designed without openings at the beam ends.

Studies have shown that the ductility of welded joints decreases significantly after a fire, and welding defects can severely affect their fatigue performance [18]. Compared with rigid connections, semi-rigid connections save on-site welding work. In light of the rotational characteristics of semi-rigid joints, several scholars [19,20,21,22] formulated multiple high-rise semi-rigid steel frames and rigid steel structures for performance evaluation. Through comparative analysis of the merits of rigid and semi-rigid joints, they explored the effect of connection rotational stiffness on the seismic performance of steel frame structures. The outcomes demonstrated that semi-rigid frames are more likely to maintain complete operational performance during an earthquake. The semi-rigid frame exhibits a lower base shear force and greater structural deformation. However, the inter-story drift ratio stays within the permissible limits, and it has a higher energy-absorption capacity. The rotational stiffness of semi-rigid joints can influence the performance of steel frames, and this factor should not be overlooked in the design and construction of steel frames. Semi-rigid restraint structures have more substantial stability and better seismic performance and are widely used in engineering [23,24]. The research on the fire resistance performance of castellated composite beams with semi-rigid restraints is still in the blank stage, and it is necessary to conduct fire resistance research on them.

In recent years, research on fire-resistance performance has gradually expanded from single components to the overall structural system. Studies have shown that the number of spans and the length of the structure have a significant impact on its fire-resistance performance [25]. Therefore, in this paper, a single-span full-scale component is adopted to conduct the fire-resistance research. We performed constant-load temperature rise tests on two circular-hole castellated composite beams with semi-rigid restraints. Both castellated composite beams had local openings (no openings within a certain distance from the beam end), and one beam web was provided with stiffeners. Through fire tests, we measured the temperature gradient, deflection conditions, and failure modes of the composite beams and compared the fire-resistance performance of castellated composite beams with semi-rigid, rigid, and hinged restraints.

2. Experimental Plan

2.1. Specimen Design

Two single-span, full-scale castellated composite beams were designed and fabricated. The beam-end restraint was semi-rigid. The composite beam without stiffeners was numbered L1, and the composite beam with stiffeners was numbered L2. The beams had a span of 4500 mm, flange width of 1500 mm, and the thickness of the plate was 100 mm. The concrete strength grade was C30, and the slab reinforcement consisted of HRB400 steel bars, which were 8 mm in diameter, spaced at 200 mm, and arranged in two layers. The studs had a diameter of 16 mm and height of 80 mm and were arranged in two rows with a transverse spacing of 100 mm and a longitudinal spacing of 200 mm.

The castellated steel beams were made of hot-rolled Q345 I-section steel, with a 70% opening rate (ratio of hole diameter to beam height). The circular openings had a diameter of 245 mm. The thickness of the transverse stiffener of L2 was 15 mm. The steel columns on both sides of the beams were hot-rolled Q345 H-section steel, with a height of 1200 mm. The base plate of the steel columns was connected to the foundation beam using six M20 anchor bolts. The column base was welded with rib stiffeners measuring 150 mm × 80 mm × 12 mm. The primary parameters of the castellated composite beams are depicted in Table 1. Taking L2 as an example, Figure 1a,b show the elevation and section views, respectively, of the composite beam. Figure 1c,d show detailed views of the semi-rigid joint and onsite views, respectively. Figure 1e,f provide side and top views, respectively, of the anchor bolt connections between the steel columns and the foundation beam (the labels in Figure 1, Figure 2, Figure 3 and Figure 4 are in millimeters (mm)).

Table 1. The primary parameters of the castellated composite beams.

Component	Size (mm)	Type	Arrangement
Concrete flange	4300 × 1500 × 100	C30	/
Castellated steel beam	350 × 200 × 8 × 12	Q345	/
column	200 × 200 × 8 × 12	Q345	/
Steel reinforcing bar	8@200	HRB400	Two-tier bi-directional layout
Stud	M16 × 65@200	Q345	Two-row layout

Table 1. The primary parameters of the castellated composite beams.

Component	Size (mm)	Type	Arrangement
Concrete flange	4300 × 1500 × 100	C30	/
Castellated steel beam	350 × 200 × 8 × 12	Q345	/
column	200 × 200 × 8 × 12	Q345	/
Steel reinforcing bar	8@200	HRB400	Two-tier bi-directional layout
Stud	M16 × 65@200	Q345	Two-row layout

Figure 1. Specimen size and structure.

Figure 1. Specimen size and structure.

2.2. Measurement Content and Measurement Point Arrangement

2.2.1. Arrangement of Temperature Measurement Points

K-type thermocouples were used to measure the temperatures at various section heights of the concrete slab, castellated steel beam, reinforcement bars, and studs. Four groups of temperature measurement points—labeled A, B, C, and D—were arranged as shown in Figure 2a. Groups AD and BC had the same arrangement along the section height. We give a detailed account of Groups C and D. Figure 2b,c show their temperature measurement points. Group C had 10 thermocouples: C1–C4 measured concrete temperatures along the section height, CG1–CG4 measured the temperatures of the castellated steel beam, and CS1 and CS2 measured the stud temperatures. Group D had 8 thermocouples: D1–D4 for the concrete temperature and DJ1–DJ4 for the reinforcement bar temperature. The temperature measurement points for the concrete and studs were spaced 20 mm apart. For the castellated steel beam, the temperature sensors were arranged at the lower flange, the web between holes, the edges of the openings, and the upper flange, as shown in Figure 2d.

Figure 2. Arrangement of temperature measurement points.

Figure 2. Arrangement of temperature measurement points.

2.2.2. Arrangement of Displacement Measurement Points

Displacement meters were employed to measure the vertical displacement of the composite beam and the horizontal displacement at the beam ends. A total of seven displacement measurement points were arranged on each specimen: points 1–3 measured the vertical displacement at the mid-point (1/2 position) and quarter-point positions of the composite beam; points 4 and 5—which were 375 mm from points 2 and 3 at the centerline of the flange—mainly measured the inclination of the concrete slab; point 6—which was located at the middle of the flange of the steel column at the beam end—measured the horizontal displacement; and point 7 was located at the edge of the flange of the steel column at the beam end. The layout of the displacement measurement points is presented in Figure 3.

Figure 3. Plan view of displacement measurement points.

Figure 3. Plan view of displacement measurement points.

2.3. Loading Method

Constant-load temperature-rise tests were carried out in the fire-test furnace at Shandong Jianzhu University’s fire laboratory. A uniformly distributed load of 3.5 kN/m² was applied to the concrete slab by means of cast-iron load blocks with a weight of 20 kg (the standard value of a live load for residential buildings, office buildings, etc. is 2 kN/m², and the dead load is 1.5 kN/m²). A total of 119 blocks were placed on the surface of the board. The distribution of the uniform load is shown in Figure 4. The temperature was raised following the GB/T 9978.1-2008 [26] standard temperature–time curve. The lower flange, web, lower surface of the upper flange of the castellated steel beam, and the lower surface of the concrete slab were exposed to fire. Prior to the tests, the columns and concrete foundation beams were fire-protected with rock wool.

Figure 4. Load arrangement.

Figure 4. Load arrangement.

3. Experimental Phenomena

3.1. Phenomena During Fire

L1: At 4 min of heating, fine arc-shaped cracks appeared in the concrete at the beam end accompanied by water stains (Figure 5a). At 6 min, fine transverse cracks appeared in the mid-span along the length of the plate, accompanied by water stains (Figure 5b). At 9 min, the arc-shaped cracks at the beam end further developed with gradually increasing water stains. Transverse cracks appeared at intervals of approximately 200 mm along the length of the plate (Figure 5c). At 37 min, the water stains at the cracks on the surface of the plate connected into large patches, forming puddles, and steam began to emerge (Figure 5d). At 69 min, the puddles were still present at the mid-span of the plate, accompanied by a large amount of steam, while the water stains at the plate end showed a trend of gradually drying up (Figure 5e). At 97 min, the mid-span deflection increased significantly and the surface inclined. The heating was stopped at 120 min (Figure 5f).

L2: At 6 min of heating, fine arc-shaped cracks appeared in the concrete at the beam end accompanied by water stains (Figure 6a). At 7 min, transverse cracks were seen on the plate’s surface. As the temperature rose, water stains appeared along the cracks (Figure 6b). At 12 min, the arc-shaped cracks at the beam end further developed with gradually increasing water stains. Transverse cracks appeared at intervals of approximately 200 mm along the length of the plate (Figure 6c). At 31 min, the water stains at the cracks on the surface of the plate connected into large patches, forming puddles, and steam began to emerge (Figure 6d). At 68 min, the surface of the plate gradually dried up, with puddles only existing at the mid-span (Figure 6e). The heating was stopped at 120 min, and the surface was almost dry (Figure 6f).

3.2. Post-Fire Phenomena

After natural cooling, the overall torsion of the L1 plate surface was not significant (Figure 7a). The web and lower flange at the beam end of L1 exhibited severe buckling. The web and lower flange were laterally unstable, and the end of the concrete slab was severely damaged. The webs between the holes showed buckling failure to different degrees. The webs near the beam end presented an S-shaped buckling failure, and the webs between the holes at the mid-span presented C-shaped buckling. This was formed due to the combined effect of the axial force formed by the restrained temperature rise expansion of the castellated composite beam and the external load of the castellated composite beam. After the fire was stopped, due to the cooling and contraction of the steel beam, a slight separation occurred at the lower part of the end plate at the joint and the steel column, but the bolts remained intact, and there was no shedding or fracture.

It can be seen from Figure 7e–i that the overall inclination degree of the plate surface of specimen L2 is small. The overall failure mode of the steel beam is similar to that of specimen L1, and the web at the beam end presents S-shaped local buckling. Due to the stiffeners’ setting, the web’s buckling at mid-span is not apparent. The joint remained intact.

By observing the failure patterns of L1 and L2, local buckling occurred at the webs and lower flanges of the steel beam ends of both specimens. Because stiffeners were set on the web of L2, the buckling at the beam end was reduced. The torsional deformation of the lower flange of L2 was decreased significantly, suggesting that the incorporation of stiffeners can enhance the anti-buckling capacity of the composite beam.

Figure 7. Test phenomena after the fire.

Figure 7. Test phenomena after the fire.

The crack distributions on the plate surfaces of the two specimens were similar (Figure 8). During the initial heating phase, inverted V-shaped cracks appeared at the beam ends (Figure 8a,c, ①), followed by transverse cracks as the temperature increased (Figure 8a,c, ②). During the later stages of heating, longitudinal cracks developed, which continued to propagate after the heating stopped (Figure 8a,c, ③). During the initial cooling phase, V-shaped diagonal cracks appeared at the beam ends (Figure 8a,c dashed line). Analysis of the causes of longitudinal cracks in the later stage of temperature increase shows that when the temperature rises, the temperature of the concrete increases, and the material properties deteriorate. Moreover, both sides of the plate surface are cantilevered. Due to the effect of the load, longitudinal fissures emerge on the surface of the plate. The heat conduction of concrete is slow. After the fire is stopped, the temperature of the concrete continues to rise, the material properties further deteriorate, and more longitudinal cracks appear on the plate surface. The V-shaped cracks on the concrete slab surface at the beam end occur because the coefficient of thermal expansion of the steel is larger than that of the concrete. At the beginning of the temperature rise, the castellated steel beam expands and deforms rapidly, while the concrete slab expands more slowly. As a result, the steel beam exerts a horizontal outward pulling force on the concrete slab, causing V-shaped cracks to appear at the end of the concrete slab. During the initial cooling phase, the reason for the appearance of V-shaped diagonal cracks in the same direction is that the cooling rate of the steel beam is greater than that of the concrete. The steel beam exerts a horizontal inward pulling force on the concrete, leading to the appearance of V-shaped diagonal cracks at the end of the concrete slab.

4. Temperature Field

Figure 9a presents the furnace temperature variation curves for the two tests compared with the standard temperature curve in GB/T 9978.1-2008. the furnace temperature variation curves of the two tests are basically consistent with the development trend of the GB/T 9978.1-2008 temperature curve. The heating time of L1 and L2 was 120 min.

The temperature field distributions of L1 and L2 were consistent. Figure 9b–e show the temperature variation curves at various measurement points on the concrete slabs of specimens L1 and L2. A significant temperature lag was observed in the concrete slabs during the tests—particularly in the upper part of the concrete. This is because concrete has a low thermal conduction rate. After the fire stopped, heat transferred to the upper concrete. Therefore, within 40 min after the fire stopped, the temperature of the upper concrete continued to increase. When the temperature of the concrete slab reached 100 °C, there was a temperature plateau due to the evaporation and heat absorption of liquid water in the concrete.

Figure 9f,g present the temperature variation curves of the steel beams in specimens L1 and L2. The temperature at the edges of the web openings increased the fastest in the initial heating phase. This is because, under fire conditions, the castellated composite beam formed a ring-shaped high-temperature region at the web openings due to thermal convection and radiation, which was gradually transferred to the web between the openings. The rate of temperature rise of the web between the openings was slightly higher than that of the lower flange, while the upper flange, which was in close contact with the concrete slab, heated the slowest. During the mid-heating phase, the temperature difference between the web and the lower flange was insignificant. The web edge temperature remained the highest, and the upper flange temperature the lowest. In the final heating phase, owing to the steel’s good thermal conductivity, the upper flange’s temperature aligned with that of the middle web and lower flange as the temperature increased. During a fire, due to the relatively high temperatures of the web and the lower flange, the compressive stress generated during their expansion and deformation is significant. The load creates a negative bending moment at the beam end, causing the lower flange at the beam end to be under compression. Under the combined effect of these two factors, severe buckling occurs in the web and the lower flange at the beam end.

Figure 9h,i show the temperature variation curves of the reinforcement bars in the two specimens. The lower reinforcement bars, closer to the fire-exposed surface, had significantly higher temperatures and heating rates than the upper reinforcement bars. At 100 °C, the upper reinforcement bars showed a temperature plateau. The temperature trend was consistent with the concrete at the same position.

Figure 9j,k show the temperature variation curves of the studs in the two specimens. The temperature at measurement point 1 was higher than that at point 2.

Figure 9. Temperature–time curves.

Figure 9. Temperature–time curves.

5. Vertical Displacement

Figure 10a,b show the two specimens’ vertical displacement curves at various measurement points. Specimens L1 and L2, within the first 5 to 10 min of heating, when the web temperature reached approximately 500 °C, deformed upward towards their initial position. After 50 min of heating, with the web temperature around 800 °C, they began to deform downward continuously. This was due to the interaction between the negative bending moment generated by the end constraints and the temperature gradient along the composite beam section height. After the cessation of fire, due to the concrete’s thermal inertia, the concrete slab’s temperature continued to rise, and the material properties continued to degrade, resulting in a continued increase in the displacement of the composite beams. Afterward, under natural cooling of the specimens, as the temperature decreased, the displacement of the specimens began to recover. After the recovery stabilized, the final displacement change was presented.

Data collection continued until 250 min. At that time, the maximum mid-span displacements of L1 and L2 were 162.9 and 149.3 mm, respectively, and the residual deformations after cooling were 150.7 and 133.7 mm, respectively. The deformation recovery ratios were 7.5% and 10.4%.

Under the combined action of high temperature and load, steel exhibits obvious creep deformation. As the temperature rises, the elastic modulus and strength of steel decline sharply, accelerating the creep rate of steel and thus influencing the overall performance of the composite beam. Part of the plastic deformation of steel is irreversible, resulting in significant residual deformation of the beam even after cooling.

Figure 10. Vertical displacement–time curves.

Figure 10. Vertical displacement–time curves.

6. Horizontal Displacement

Figure 11 presents the two specimens’ horizontal displacement–time curves, where inward shrinkage is negative (+) and outward expansion is positive (−).

After the fire test began, the composite beam rapidly expanded and deformed outward, and the horizontal displacement of L1 reached −25 mm. Due to the restraint effect of the steel columns at the beam ends, a large axial compressive force was generated inside the composite beam. As shown in the figure, when the test was carried out for 28 min, the axial compressive force of L1 reached the maximum value and lasted for 55 min. At 83 min, the composite beam began to contract and deform inward, and the axial compressive force decreased. At this time, the temperature of the steel beam web was between 900 °C and 1000 °C. Analysis of the reasons for the horizontal displacement contraction shows that as the temperature rises, the steel enters the plastic deformation stage, its bearing capacity further decreases, and the deflection increases, leading to the redistribution of internal forces, and the composite beam gradually changes from pressure to tension. When the temperature rose for 28 min, the horizontal displacement of L2 reached −24 mm and lasted until the end of the heating. After the cessation of the fire, the composite beam began to contract and deform. The final deformation value of L1 is −4.10 mm and the final deformation value of L2 is −7.70 mm.

Figure 11. Horizontal displacement–time curve.

Figure 11. Horizontal displacement–time curve.

7. Different Constraints Comparison

L1 has the same parameters as the castellated composite beams in reference [24], except for the constraint conditions. The failure modes of L1 are similar to those of the hinged and rigid restraint castellated composite beams in reference [27]. In all cases, the bottom flange experienced torsional deformation, and both the end web and bottom flange underwent buckling failure. For the hinged restraint castellated composite beam, the joint suffered severe damage, with significant deformation of the end plate connecting the steel column and beam and relative slip between the bolts and nuts (Figure 12a). The welds fractured at the joint for the rigid restraint castellated composite beam (Figure 12b). In contrast, the joint of L1 showed no significant deformation and remained intact. Figure 13 shows the mid-span displacement–time curves of castellated composite beams with different joint types. It can be observed that the mid-span displacement of the semi-rigid restraint castellated composite beam is significantly smaller than that of the rigid and hinged restraint beams under the same heating duration. In conclusion, the semi-rigid restraints castellated composite beam exhibits superior fire resistance compared to the hinged and rigid restraint beams.

8. Finite-Element Simulation

8.1. Modeling

Under standard heating conditions, the fire-resistance performance of a steel–concrete castellated composite beam was simulated through the application of the thermo-mechanical coupled analysis approach within the ABAQUS (2021) software. To be more precise, thermal analysis was conducted initially to acquire the temperature field distribution of the components. Subsequently, a thermo-mechanical coupled simulation analysis was carried out.

8.2. Temperature Field Models

During the heat-transfer analysis, the three-dimensional solid element DC3D8 was employed to model the castellated steel beam and the flange plate. The heat transfer connection element DC1D2 was used to model the reinforcement. The composite beam experienced three-sided fire exposure. For the surfaces affected by fire, both the impacts of thermal convection and radiation were taken into account. The convective heat transfer coefficient was set at 25 W/(m·°C). The radiation coefficient of the castellated steel beam stood at 0.95, while that of the flange plate was 0.5. The Stefan–Boltzmann constant was set at 5.67 × 10⁻⁸ W/(m²·K⁴), with absolute zero assumed to be −273 °C. The mesh was sized at 20 × 20 mm, and additional refinement was carried out in the region around the hole. As depicted in Figure 14, this shows the mesh division of the components. The material properties used in the simulation were obtained from EC3 (EN 1993-1-2) [28] and EC4 (EN 1994-1-2) [29].

8.3. Thermal–Mechanical Coupled Analysis Model

In the structural analysis involving thermal and mechanical forces, C3D8R elements were employed for both the castellated steel beam and concrete slab, whereas T3D2 elements were utilized for the reinforcement bars. The connection between the steel beam and concrete slab was established using a tie constraint, with the assumption of no shear slippage between them. Additionally, the reinforcement was considered fully embedded within the concrete slab.

The properties of the materials were specified in the following manner. The density of steel was set at 7850 kg/m³, and its Poisson’s ratio was taken as 0.3. In the case of concrete, its density was defined as 2400 kg/m³. The concrete slab was modeled to include plastic-damage characteristics. The values of its compressive strength, elastic modulus, and tensile strength were sourced from references [30,31]. The stress–strain relationship of concrete was determined in accordance with EC4 (GEN 1994). Regarding steel, the material properties were obtained from EC3 (GEN 1993).

8.4. Analysis of Simulation Results Compared to Experimental Results

Figure 15 illustrates the comparison between the experimental and numerical simulation results for temperature and displacement curves. It is evident that the trends in the simulation align closely with the experimental data. This consistency allows for further parametric analysis to investigate additional factors influencing the fire resistance of beam-end restrained castellated composite beams.

9. Conclusions

We investigated the fire performance of semi-rigid restrained castellated composite beams. The following conclusions were drawn.

For the semi-rigid restrained castellated composite beam without stiffeners, the web between the middle holes presented “C”-shaped buckling, and the web and lower flange showed overall lateral instability. There was no apparent buckling in the web between the middle holes and no torsional deformation in the lower flange for the semi-rigid restrained castellated composite beam with stiffeners. Setting stiffeners can effectively improve the anti-instability ability of the castellated composite beam. Therefore, in engineering projects, stiffeners can be arranged reasonably according to actual requirements.

After the fire, the composite beams first produce downward vertical deformation. After 5–10 min, when the temperature of the web is around 500 °C, it starts to deform upward to the initial position. After 50 min, when the temperature of the web is around 800 °C, it starts to deform downward continuously.

The deformation recovery ratios of the two semi-rigid restrained castellated composite beams after the fire were 7.5% and 10.4%, respectively. Whether to set stiffeners has little influence on the composite beams’ deformation recovery.

Under fire, the failure modes of semi-rigid restrained castellated composite beams are similar to those of hinged and rigid restraints, showing torsional deformation of the bottom flange and buckling of the web and bottom flange at the beam end. After the fire, the deformation of semi-rigid joints is minor, and their fire resistance performance is better than that of hinged and rigid joints. By comparing the mid-span displacements, the fire resistance performance of semi-rigid restrained castellated composite beams is also better than that of hinged and rigid restraints. For buildings with high requirements for fire resistance performance, it is recommended to adopt semi-rigid restraint joints.

When cooling down, the castellated composite beams calm and contract inwardly. The composite beams exert a sizeable tensile force on the columns through the action of bolts. Semi-rigid joints should have sufficient tensile strength to ensure that the beams and columns do not separate and cause the building to collapse.

Numerical simulation analyses were carried out on the castellated composite beam. The simulation results showed a high degree of consistency with the test results, which effectively validated the accuracy and reliability of the proposed finite-element model.

In this paper, only the constant-load heating tests were carried out on two semi-rigid restrained castellated composite beams. The sample size is relatively small, the variables and load forms are single, and there are no comparative tests on semi-rigid restrained castellated composite beams with different joint stiffnesses.