Fire tests on full-scale FRP reinforced concrete slabs

Fire tests on full-scale FRP reinforced concrete slabs

Accepted Manuscript Fire Tests on Full-Scale FRP Reinforced Concrete Slabs Hamzeh Hajiloo, Mark F. Green, Martin Noël, Noureddine Bénichou, Mohamed Su...

3MB Sizes 0 Downloads 32 Views

Accepted Manuscript Fire Tests on Full-Scale FRP Reinforced Concrete Slabs Hamzeh Hajiloo, Mark F. Green, Martin Noël, Noureddine Bénichou, Mohamed Sultan PII: DOI: Reference:

S0263-8223(17)30243-X http://dx.doi.org/10.1016/j.compstruct.2017.07.060 COST 8715

To appear in:

Composite Structures

Received Date: Revised Date: Accepted Date:

23 January 2017 29 June 2017 19 July 2017

Please cite this article as: Hajiloo, H., Green, M.F., Noël, M., Bénichou, N., Sultan, M., Fire Tests on Full-Scale FRP Reinforced Concrete Slabs, Composite Structures (2017), doi: http://dx.doi.org/10.1016/j.compstruct. 2017.07.060

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Fire Tests on Full-Scale FRP Reinforced Concrete Slabs Hamzeh Hajiloo a, *, Mark F. Green a, Martin Noël b, Noureddine Bénichou c, Mohamed c Sultan a

Civil Engineering, Queen’s University, 58 University Avenue, Kingston, Canada

b

Department of Civil Engineering, University of Ottawa, 161 Louis-Pasteur Pvt., Ottawa, Canada

c

NRC-construction, National Research Council Canada, Ottawa, Canada

Address all correspondence to E-Mail: [email protected]; Abstract The fire resistance of two concrete slabs reinforced completely with glass fibre reinforced polymer (GFRP) bars was experimentally investigated through a full-scale standard fire test. The Canadian standard on design of FRP reinforced buildings (CSA S806) currently requires 60 mm of clear concrete cover in a FRP reinforced concrete slab to ensure 2 hours of fire resistance. The two concrete slabs, which were identical in every aspect except the types of GFRP reinforcing bars, were designed and fabricated with 60 mm concrete cover in this experiment to examine the fire resistance of the existing FRP reinforced concrete structures designed using CSA S806. The uniformly distributed loads were applied causing flexural moment of 45 kN.m, which was 55% of the ultimate moment resistance of the slabs at room temperatures, and sustained throughout the fire test. Both loaded slabs endured more than three hours of ASTM-E119 standard fire exposure. The results promised a more efficient and economic application of FRP reinforcing bars in concrete construction. The effects of concrete cover, unexposed length in providing adequate anchoring at the ends, and two different types of widely used GFRP reinforcing bar were studied in this research. Keywords: fire resistance; fibre reinforced polymer; concrete slab; bond strength; numerical

analysis 1

1

Introduction

Enormous potential exists for applications of FRP reinforcing bars to replace conventional steel bars, which are highly susceptible to corrosion. Some examples include multi-storey buildings, parking garages, industrial structures, and bridge decks. Outstanding characteristics of FRP materials such as high strength-to-weight ratio and resistance to corrosion make it suitable for structures built in corrosive and harsh environments. However, the fire performance of concrete components reinforced with FRP bars needs to be addressed thoroughly to ensure safe applications of FRP. The performance of an individual structure with several primary loadbearing members such as beams, columns, and connections during fires is a complicated matter which demands the nonlinear properties of material and geometry to be studied. However, breaking up the structural members of a building and studying an individual member is an effective starting point to understand the behaviour of FRP reinforced concrete structures. The current study investigated the behaviour of flexural members with a focus of simply-supported FRP reinforced slabs. Fire tests have been conducted on FRP reinforced flexural concrete members [1-5]. The bond deterioration of FRP reinforcing bars to concrete at high temperature was the main focus investigated experimentally by Weber [3]. These were the first reported FRP reinforced concrete fire tests consisting of slabs with lap-sliced bars in the middle. The slabs failed due to the loss of bond strength at the middle of the slabs because of the lap-spliced bars at midspan. Nigro et al. [4] studied the effect of cool anchor zones on the fire performance of GFRP reinforced concrete slabs and concluded that the collapse of FRP reinforced concrete slabs will occur at midspan by tensile rupture of the FRP if there is a sufficiently long region that is unexposed to fire at the supports. 2

Conventional steel reinforced concrete slabs are designed to satisfy fire safety requirements by providing sufficient concrete cover. It is generally assumed that steel reinforcing bars lose 50% of their room-temperature strength at 593 °C, which is called the critical temperature [6]. Given the wide variety of FRP materials available in the market with various surface treatments and composition as well as the complicated FRP-concrete bond loss at elevated temperatures, establishing critical temperature for FRP bars is less practical [7]. However, CSA S806 [8] still uses the critical temperature concept to provide a simplified method to design concrete slabs with FRP reinforcement. The approach presumes the loss of load-bearing capacity of the slabs when the reinforcing bars lose half of their tensile strength with increased temperature. This approach demands a concrete cover of approximately 60 mm to achieve a 2 hour fire resistance rating. The accuracy of the design method suggested in CSA S806, in which a critical temperature is used to define the fire resistance of FRP reinforced concrete members, was investigated by Adelzadeh et al. [9]. In the study, analytical modelling was used to study the fire resistance of FRP reinforced flexural members, and the authors concluded that critical temperature was not effectively applicable to FRP reinforced concrete members because it represented a lower bound of the expected fire endurance. This was mainly due to the considerable reserve strength of FRP reinforced concrete members because serviceability criteria are the governing factors in the design [10]. In another study [1], GFRP reinforced concrete beams with 70 mm concrete cover were tested and attained 90 minutes of fire endurance. The recommendation was a minimum concrete cover of 70 mm for the fire design of GFRP reinforced concrete. These suggested values for the concrete cover are neither practical nor economical and result in an inefficient application of relatively expensive FRP reinforcements.

3

To fill the significant gaps with respect to the assessment of existing FRP reinforced concrete flexural members in potential fire scenarios, as well as to provide applicable guidelines in the design and construction of FRP reinforced slabs and beams, an extensive project has been launched at Queen’s University, Canada. The material properties of different GFRP reinforcing bars were investigated through tensile and pullout tests at high temperatures. The characteristics of the GFRP reinforcements (Rebar-A and Rebar-B) used in the fabrication of the slabs have been thoroughly studied through experiments [11, 12]. As a part of this project, two full-scale slabs were tested in the floor furnace at the National Research Council (NRC) facilities. The slabs were designed and fabricated to represent common FRP reinforced concrete slabs such as those found in a typical parking garage. Additional slabs will be designed and tested based on the outcome of the first tests. 2

Experimental program

The experimental investigation consisted of ASTM-E119 standard fire [13] tests on two GFRP reinforced concrete slabs, designated Slab-A and Slab-B, to represent simply-supported one-way slabs. Although the test frame was capable of axially restraining the slabs, the slabs were conservatively tested without end restraints. 2.1 Test Specimens Table 1 summarizes the details of the slabs fabricated for the fire test. As Kodur et al. [7] concluded, the most influential parameter in the fire resistance of GFRP reinforced concrete members is concrete cover, whereas the slab thickness and aggregate type had only a minor effect. Table 1 - Test specimens’ parameters Slab ID

Reinforcement

Aggregate

length 4

Thickness

Clear concrete

Unexposed

Slab-A Slab-B

Rebar-A Rebar-B

Carbonate Carbonate

3900 3900

(mm)

cover (mm)

200 200

60 60

zones at the ends (mm) 200 200

As shown in Fig. 1, the slabs were 3900 mm long and 1200 mm wide with a thickness of 200

mm, which is typical for slabs in parking garages. The lateral concrete cover for the longitudinal reinforcements was 25 mm, and the centre-to-centre clear spacings of the bottom and top longitudinal reinforcement were 100 and 220 mm, respectively. The spacing of the bars was 200 mm in both the bottom and top mesh. The sides of formworks were stripped the day after concrete casting and the slabs were covered with plastic sheets to maintain moisture for better curing for two weeks. The slabs were kept in an indoor warehouse (approximately 10 °C) to prevent the potential freezing effects since the slabs were cast in January. According to Hertz [14], one of the main reasons for concrete spalling in fire tests is the short time between the cast of concrete and the fire test so that the moisture content is high enough to cause excessive pressure in the concrete resulting in spalling. To reach a stable moisture condition in the slabs, the fire test was conducted 10 months after the concrete cast. #5(16 mm) @ 100 mm

#4(13 mm) @ 220 mm 60

#4(13 mm) @ 200 mm

3900

#[email protected] mm

60

200

Straight Longitudinal Reinforcement

1200 Transverse Section

Fig. 1. Reinforcement details and dimensions of the slab in longitudinal and transverse sections

2.2 Materials The slabs were fabricated with calcareous aggregate concrete, and the average 28 day compressive strength of the concrete calculated from 5 cylinders was 28.9 MPa. The concrete did 5

not gain more compressive strength beyond 28 days because the average of 6 cylinders before the fire test was 27.9 MPa. The average splitting tensile strength of three cylinders at the time of testing was 2.8 MPa. The reinforcement consisted of #4 and #5 straight GFRP bars in transverse and longitudinal directions, respectively. Slab-A and Slab-B were reinforced with Rebar-A and Rebar-B reinforcing bars, which were produced by two Canadian companies using the pultrusion process in which continuous glass fibres pass through a sink of resin and then are pulled through a die with specific diameter. The bars contained 80-84% by weight continuous E-glass fibres bonded together with vinylester resin. The glass transition temperature of the bars was calculated using differential scanning calorimetry (DSC) method according to CSA S807 [18]. Table 2 summarizes the material properties that were reported by the manufacturers based on tests on the same batch of bars that were used to construct the slabs. Table 2- Reinforcing bars properties evaluated by the producer

Rebar-A Properties Nominal Diameter Effective Diameter (including coating) Nominal Cross Sectional Area Cross Sectional Area (including sand coating) Minimum Guaranteed Tensile Strength Nominal Modulus of Elasticity Ultimate Elongation Fibre content (by weight) Glass transition temperature

Units (mm) (mm) 2 (mm ) 2 (mm ) (MPa) (MPa) (%) (%) °C

Rebar-B

#4

#5

#4

#5

12 14.0 126 187 1300 69600 2.5 84 118

16 18.5 198 269 1180 64000 2.7 80 118

13 14.3 129 161 980 61100 2.1 82 113

15 17.9 199 252 1125 62600 2.1 82 113

GFRP reinforcing bars with different surface treatments such as deformed ribs and sand coated, with and without helical braided fibres, are available in the market. The type of surface treatment is one of the main factors influencing shear stress transfer between a FRP bar and the surrounding concrete. Rebar-A reinforcing bars have a sand coating on the surface which is applied on the reinforcing bar following the pultrusion process. The adhesive used to bond the 6

core of the bar to the sand coating is a key factor in providing bond resistance at elevated temperatures. In addition to sand coating, Rebar-B incorporates two braids of fibres helically wound on Rebar-B. The tightly wrapped glass fibres around Rebar-B reinforcing bars create convex protrusions on the surface of the bars. 2.3 Instrumentation To study the temperature profiles along the reinforcing bars and through the depth of concrete sections, 34 Type K thermocouples (TCs) were attached to the bottom and top three reinforcing bars at several locations. As shown in Fig. 2, the ends of slabs were protected form fire, and these unexposed zones are also called anchorage zone or cool zone in literatures. Because of the great importance of thermal fields at anchor zones in the failure of slabs, 24 out of 34 sensors were placed in the unexposed zones at the ends of the slabs. A vertical unit of TCs was also placed in the midspan of each slab to record the thermal gradient through the concrete depth. In total, 7 thermocouples were placed at depths of 20, 40, 60, 90, 120, 150, and 180 mm from the heated surface. In addition, each slab was instrumented with 6 strain gauges on top of the main longitudinal reinforcing bars and two 50 mm long strain gauges on the top surface of the concrete. Strain gauges on the reinforcing bars were installed 1.0, 1.95, and 2.7 m from the left end of the slabs. On top of the concrete, one strain gauge was installed 1.0 m from the left end, and the second gauge was installed at the midspan. Strain gauges were intentionally installed on top of the reinforcing bars to delay heat-induced damages so that data can be obtained for a longer time from initiation of fire. The functional temperature range of the strain gauges was -75 to +95 °C.

7

NORTH

600 180 Strain Gauge on FRP Rebar

R C

Thermocouple on FRP Rebar

200

150

S SC

120 90

Strain Gauge Top Surface of Slab

60

40

Thermocouple in Concrete Depth

20

Thermocouples in the midspan shown in slab's cross section

R33,34

R29,30

R25,26

R31,32

C

75 75

R23,24 R21,22

B

R27,28 850

950

R7,8

1050

R19,20

R17,18

R15,16

S6

S5

S4

R9,10 A

750

R11,12 R13,14

R5,6 S3

C

75 75

B

R1,2

S2

R3,4

S1

A

Instruments on the bottom layer of reinforcement

60

R24 R22

R20

R18

R23 R21

R19

R17

200 Unexposed

SC1 950

R16 R15

3500 Exposed

1000

R14 R12 R13 R11

200

SC2

Section B-B

200 Unexposed

Fig. 2. Thermocouple, strain gauges, and labelling system

Substantial care was taken in installing TCs to ensure that the tips of the sensors were situated in the specified position because any slight deviation of the sensors’ tips might result in reading temperatures of a wrong location. Fig. 3(a) shows TCs placed at the anchor zone of Slab-A, and Fig. 3(b) shows the installed TCs on the lower reinforcement. The collection of 7 TCs mounted on a piece of rod to measure temperatures throughout the depth of the slab is shown in Fig. 3(c) with red arrow. The slabs were cast at Sherbrooke University and cured indoors for 6 months before being transported to National Research Council (NRC) for the fire test. As a widely known fact, the age of concrete influences the fire performance of concrete, especially if it is less than 6 months.

8

(a)

(b)

(c)

Fig. 3. Instrumenting and casting of the slabs: (a) TCs placements at the end unexposed zone of Slab-A; (b) TCs installed on reinforcing bars; (c) Fully instrumented Slab-B

2.4 Specimen design When a structural element is loaded in flexure, the quality of the bond influences the crack spacing, deflection characteristics, and the ultimate load capacity of these elements. Since FRPconcrete bond is highly susceptible to deterioration at high temperature, the anchorage properties of the reinforcing bars are of particular interest. For this reason, extensive instrumentation was provided in the anchorage zones at the ends of the slabs. Another key challenge is selecting an appropriate sustained load level to represent the likely loads to be present during a building fire. For concrete slabs reinforced with steel, ASTM E119 [13] and ULC [15] recommend the service load level to be determined based on the ultimate strength of the slab. However, the design of FRP reinforced concrete slabs is generally governed by serviceability criteria such as deflection and crack control. Therefore, selecting the load level based only on ultimate strength would give an unreasonably high and impractical load for FRP reinforced concrete slabs. For these slabs, the ultimate moment resistance was calculated as 84 kN.m according to CSA S806. Based on ultimate strength alone, the full service load level would correspond to an applied moment of 61 kN.m. To select an appropriate sustained load 9

level to be applied on the slabs during the fire, serviceability criteria were checked with three common FRP reinforced concrete structures codes and guidelines including ACI 440.1R-06 [16], CSA S806-12 [8], and ISIS Manual 3 [17]. Table 3 summarizes the predicted results for deflection and crack widths at 5 levels of applied loading (Ma = 27, 40, 45, 52, and 61 kN.m). Table 3- Serviceability conditions design for Slab-A

Total Deflection (mm) Crack Width (mm)

27 3 13 3 0.43 0.52

ACI 440-1R-06 ISIS No. 3 CSA-S806 ACI 440-1R-06 ISIS No. 3

Applied moment (kN.m) 40 45 52 10 14 21 27 32 39 31 36 43 0.64 0.72 0.84 0.78 0.87 1.01

61 30 47 50 0.97 1.18

For the 3.8 m long span, the live load deflection limit is ln/360 = 3840/360 = 10.7 mm. If the live load deflection is assumed to be equal to the dead load deflection, then the deflection limit would be exceeded for a total deflection of 21 mm. CAN/CSA S806 allows crack widths of 0.5 mm for exterior and 0.7 mm for interior applications. From the serviceability calculations, the service limits for cracking and deflection were exceeded for moments above 40 kN.m. As a compromise between the ultimate load and service load requirements, the test load was chosen as a distributed load of 24.9 kN/m giving a total moment of 45 kN.m at midspan corresponding to 55% of the ultimate flexural capacity of the slabs. Given that the self-weight of the slab is 5.8 kN/m, the applied superimposed load before the fire test was 19.1 kN/m. The stress in the FRP at this load level is approximately 160 MPa which is well below the allowable service stress limits of 340 MPa for the reinforcement in Slab-A and 255 MPa for Slab-B. However, the strain is approximately 2500 microstrain which is well above the limit of 2000 microstrain recommended by CSA S806 to control cracking.

10

2.5 Test setup and procedure The testing apparatus consisted of a movable steel frame which was placed on top of the floor furnace. As shown in Fig. 4, the ends of the slabs were placed on the corbel of the steel frame with the same conditions at both ends. The slabs were simply supported and unrestrained both axially and rotationally when the load was applied. In order to maintain the unrestrained conditions throughout the fire test, sufficient gaps were provided at the ends of the slabs. Any gaps between the two adjacent slabs as well as the spaces between each slab and the lengthwise side of the frame were filled with ceramic insulating panels. The furnace was programmed to follow the ASTM E-119 [13] time-temperature curve, and the heat input of the furnace was controlled during the test using the readings of several shielded thermocouples located close to the soffit of the slabs. Heat-insulating layers were paled on the bottom surfaces of the slabs at the ends as shown in Fig. 4. With considering the support length, 200 mm of the ends of each slab were protected from fire. Since the ends of reinforcing bars were 25 mm from the ends of slabs, thus, 175 mm of each reinforcing bar was protected from fire at both ends.

Insulation

Insulation 200

3500

200

Fig. 4. Floor furnace test setup

The uniformly distributed load was applied by a loading system consisting of six jacks evenly located along the length of each slab (Fig. 5). During the fire test, the slabs were tested under a 11

sustained uniformly distributed load that was gradually applied in four increments of 25% of the total load allowing time for stabilization of deflection after each increment. Pre-loading of the slab and reaching the planned sustained loads was completed 30 minutes prior to the start of the fire, and the constant load level was maintained throughout the fire test. The slabs were once loaded the day before fire test to assess and evaluate the extent of cracks development as well as deflection.

(a)

(b)

(c)

Fig. 5. (a) NRC floor furnace facility; (b) Several jacks to apply uniform load on each slab; (c) slabs on furnace

3

Experimental results

Both concrete and reinforcing materials suffers from degradation of mechanical properties when subjected to fire. Although the heat-induced degradation of concrete and steel reinforcement in conventional reinforced concrete elements is a matter of concern, concrete degradation is not as much of a concern for FRP-reinforced members since the degradation of the FRP material takes place considerably earlier than that of concrete. In steel-reinforced structures, the bond strength variation at elevated temperature is quite similar to the variation in the compressive strength of the concrete [18]. However, the bond of FRP to concrete is the first mechanism of deterioration at elevated temperatures. Based on pullout tests, when the temperature at the concrete-FRP reinforcing bars interface reaches 170 °C, the remaining bond strength is only approximately 12

10% of the original bond strength at room temperature for both Rebar-A and Rebar-B bars [12]. The temperature distribution in the concrete and the FRP reinforcement, and the load, deformation, and strain behaviour are discussed in the following sections. 3.1 Temperature Variation The thermal distribution in the slabs was monitored and recorded with several thermocouples (see Fig. 2). Temperatures at various depths through the concrete at midspan are shown in Fig. 6. The curves are labeled with the representing distance in millimetres from the soffit of the slabs. In the presented curves, the locations of thermocouples in the slabs are shown next to the curves to locate the sensors’ position easily. At any depth within the cross section, the time-temperature curve temporarily reached a plateau when the temperature reached 100 °C. Depending on the depth where the measurement was carried out, the plateau lasted differently. This plateau lasted for 3, 17, and 23 minutes at depths of 20, 40, and 60 mm, respectively. This phenomenon coincided with the evaporation of the mixing water entrapped in the concrete. The closer to the exposed surface, the faster the evaporation occurred due to the higher temperature and also less distance for steam to escape the concrete environment. Temperatures at 60 mm in both slabs were similar and reached 370 °C by the end of fire test. The temperature at this level was comparable to the readings from the sensors installed on the bottom of the reinforcing bars (Fig. 7). The readings from the sensors at depths far from the soffit showed that the compression side of section did not experience extreme conditions, and the temperatures remained around 100 °C by the end of test. The deterioration in mechanical properties of concrete such as compressive strength and modulus of elasticity at 100 °C is negligible [19]. Normally, the tensile strength of concrete is ignored in strength calculations of flexural members. The temperature profiles in the

13

compression portion of the concrete slabs showed that there was no significant strength loss in concrete. The minor differences in thermal measurements between identical locations in the two slabs can be attributed to the possible displacement of the tips of the thermocouples during concrete casting. 800

600 500

20 mm 40 mm 60 mm 120 mm 150 mm 180 mm

Slab-A 700 600

Temperature (°C)

700

Temperature (°C)

800

20 mm 40 mm 60 mm 90 mm 120 mm 150 mm 180 mm

400 300

500 400 300

200

200

100

100

0

Slab-B

0 0

30

60

90

120

150

180

210

Time (min)

0

30

60

90

120

150

180

210

Time (min)

Fig. 6. Temperatures within the thickness of the slabs

The initial thermal increase at the bottom of the lower reinforcement layer to 30 °C was observed 10 minutes after fire started. Notable differences between temperatures at the top and bottom of the reinforcing bars at the same locations are shown in Fig. 7 due to nearly 20 mm diameter of the bars. There was, for example, a difference of 185 °C between the readings of R19 and R20 thermocouples, which were located on bottom and top of the FRP bar, respectively. A sudden rise was recorded by R18 and R20 when Slab-A failed. This is believed to be the result of slippage of the bars that occurred upon failure and displaced sensors. Moreover, flames could penetrate deep in the section and reach the sensors on the bars upon failure. The results demonstrated small differences in the heat distribution along the span length of the slabs at the same depth in the concrete in the exposed zones. In Fig. 7, the thermal measurements

14

of the sensors in the exposed zones indicated reasonably close temperatures at the same depth at three locations: R15, R17, and R19. 1000 500

350

R15 R16 R17 R18 R19 R20

450

Slab-A

400

Temperature (°C)

400

R16 R15

R18 R17 500

R15 R16 R17 R18 R19 R20

450

Temperature (°C)

Section B-B

R20 R19

300 250 200 150

350 300 250 200 150

100

100

50

50

0

Slab-B

0 0

30

60

90

120

150

180

210

Time (min)

0

30

60

90

120

150

180

210

Time (min)

Fig. 7. Temperatures on top and bottom of the main middle rebar

Similarly, thermal measurements presented in Fig. 8 indicated that there was uniform temperature distribution in the transverse direction of the slabs. Six thermocouples in the midspan showed that temperature profile in the transverse direction was uniform. This result can save time and cost in numerical modelling of heat transfer and reduce the model from 3D to a planar 2D model.

15

1200 R18 R17

R6 R5 R5 R6 R17 R18 R29 R30

450

Temperature (°C)

400 350

500

R5 R6 R17 R18 R29 R30

450

Slab-A

400

Temperature (°C)

500

R30 R29

300 250 200 150

350 300 250 200 150

100

100

50

50

0

Slab-B

0 0

30

60

90

120

150

180

210

Time (min)

0

30

60

90

120

150

180

210

Time (min)

Fig. 8. Temperatures on top and bottom of reinforcing bars in the midspan

The most insightful thermal evaluation was performed in the unexposed zones at the ends of the slabs that were not exposed directly to the furnace. These areas of interest were heavily instrumented since the failure of the slabs was expected to be governed by FRP-concrete bond performance at elevated temperature as heat propagated into the anchor zone. Four thermocouples on each of three selected reinforcing bars were installed on the top and bottom of the bars at 75 and 150 mm from the end of each slab at both ends (Fig. 9). The graphs show approximately 150 °C differences between the temperatures on the bottom of the reinforcing bars at the locations 150 and 75 mm from the end of slab. While temperatures on the bottom of reinforcing bars reached 400 °C in the exposed zone (Fig. 8), temperatures remained around 100 °C at the locations monitored by sensors R1, R9, R11, and R23 in unexposed zones.

16

300

300

Temperature (°C)

250 200

R1 R2 R3 R4 R7 R8 R9 R10

Slab-A 250

Temperature (°C)

R1 R2 R3 R4 R7 R8 R9

150 100 50

200

Slab-B

150 100 50

0

0 0

30

60

90

120

150

180

210

0

30

60

Time (min)

300

120

150

180

210

150

180

210

300

200

R11 R12 R13 R14 R21 R22 R23 R24

Slab-A 250

Temperature (°C)

R11 R12 R13 R14 R21 R22 R23 R24

250

Temperature (°C)

90

Time (min)

150 100 50

200

Slab-B

150 100 50

0

0 0

30

60

90

120

150

180

210

Time (min)

0

30

60

90

120

Time (min)

Fig. 9- Temperatures at the anchorage zone on the bottom of the bars

The readings from some of the thermocouples at the unexposed zones at the bottom of reinforcing bars are demonstrated in Fig. 10 as a function of length from the end of a slab toward the exposed zone. The readings from two thermocouples (R21 and R23) at the unexposed zone plus the third temperature which is a representative of temperatures at 60 mm in the exposed

17

zone draw a thermal gradient. The data points in Fig. 10 are connected with dotted lines to better present the results, and interpolation of temperature between point was not intended. 500

500 30 min 60 min 90 min 120 min 150 min 180 min 188 min

300

400

Temperature (°C)

Temperature (°C)

400

30 min 60 min 90 min 120 min 150 min 180 min 188 min

300

Slab-A

100

0 0

End of Slab

200

End of Slab

200

Slab-B

100

0 75

150

225

300

0

Distance (mm)

75

150

225

300

Distance (mm)

Fig. 10. Temperature gradient in unexposed zones

3.2 Load and deflection behaviour Bond stresses at the interface between the concrete and the FRP bars transfer tensile force to the FRP reinforcement under flexural stresses. Splitting cracks will develop along the FRP bars if the tensile strength of the concrete is not adequate to resist stresses induced by mechanical interlocking between the two materials while the surface treatment of the FRP reinforcement is still intact [2]. After the slabs were loaded, no longitudinal splitting cracks were observed underneath the slabs before the fire. The only cracks on the slabs were transverse flexural cracks as marked in Fig. 11. These cracks may have expedited the thermal propagation into the concrete thickness; however, the thermocouple readings did not show evidence of increased temperatures at the locations of the cracks.

18

Fig. 11. Marked cracks on the soffit of Slab-A after loading

The specified superimposed load was applied in four steps as recommended by CAN-ULC-S101 [15]. The effects of the four levels of load increments are evident in the strain readings shown in Fig. 12. The sides of the slabs were not accessible to assess the depth of the cracks; however, the cracks were visible from underneath as shown in Fig. 11. Although the slabs were unloaded, permanent deformations and cracks remained after unloading. Non-zero strains in Fig. 12 are the permanent strains that remained in the reinforcing bars and top surface of the slabs after the firsttime loading. 2500

2500

S1 S2 S3 S4 S5 S6

2000

Slab-B

Microstrians

Microstrians

2000

S1 S2 S3 S4 S5 S6

Slab-A

1500

1500

1000

1000

500

500

0

0 0

4

8

12

16

0

20

4

8

12

16

20

Superimposed Load (kN/m)

Superimposed Load (kN/m)

(a)

(b)

Fig. 12. Strains on the reinforcing bars and top surface of concrete before fire (see Fig. 2 to locate the gauges): (a) Slab-A; (b) Slab-B

The loading history of the slabs and the deflection behaviour shows (Fig. 13) a similar deflection for both slabs. The first-time loading load-deflection behaviour of the slabs are shown in dashed line in Fig. 13. In addition, the unloading behaviour is shown in dashed lines, and the second19

time loading of the slabs are shown in solid lines. The permanent deflection of the slabs after first-time loading for both slabs was 16 mm. The second-time loafing of the slabs caused maximum mid-span deflection (30 mm) similar to the first-time loading defection taken into account the permanent deflection. Second-Time Loading (Slab-A) First-Time Loading (Slab-A) Second-Time Loading (Slab-B) First-Time Loading (Slab-B)

35

20

30 15 25 20 10 15 10

Loading Before Fire

5

Unloading

Superimposed Load (kN/m)

Superimposed Moment (kN.m)

40

5 Permanent Deflection

0

0 0

5

10

15

20

25

30

35

Deflection (mm)

Fig. 13. Loading history of the slabs before fire test

The deflections during fire versus time, and furnace temperature-deflection curves of the slabs are plotted in Fig. 14 (a) and (b), respectively. The deflection values at the start of the fire test were set to zero to reflect only the heat-induced deformation resulting from the thermal bowing of concrete. Thermal bowing occurred due to the thermal gradient within the depth of concrete leading to unequal expansion of the soffit with respect to the top surface of the slabs. Although the deflection curve after the fire initially increased at a rapid rate of approximately 1.4 mm/min, the response stabilized after approximately 20 minutes, and the deflection increased gradually with a lower rate. After sustaining the loads for 3 hours, the load was gradually increased from 19.1 kN/m. Upon reaching 23.2 kN/m, the deflection quickly increased in Slab-A indicating failure.

20

100

100

80

Slab-A Slab-B

Deflection (mm)

Deflection (mm)

80

60

40

20

Slab-A Slab-B 60

40

20

0

0 0

30

60

90

120

150

180

210

0

200

400

Time (min)

600

800

1000

1200

Temperature (°C)

(a)

(b)

Fig. 14. Deflection behaviour of the slabs during fire as a function of: (a) fire exposure time; (b) furnace temperature

Strains during fire under sustained load are shown as a function of temperature on top of the reinforcing bars in Fig. 15(a) and (b) for Slab-A and B, respectively. Plotting the strains versus the temperature at the same locations revealed that the strains began to increase before heat reached the bars. These strains were caused by the rapid thermal bowing of the slabs.

(a)

(b)

Fig. 15. Strains in the reinforcing bars (see Fig. 2 to locate the gauges) vs temperatures in (a) Slab-A and (b) Slab-B;

Fig. 16 shows strains as a function of time of fire exposure in the reinforcing bars at the midspan of both slabs. With the start of the fire, the strains increased until the strain gauges lost their contact with the reinforcing bars. Most of the strain gauges were functional for the first 20 minutes of fire exposure. The strains then decreased gradually due to the weakened contact 21

indicating debonding of the gauges. Consequently, the readings of the strain gauges are useful and valid for only the first 20 minutes of the fire exposure. The readings in the highlighted area in Fig. 16 are useless. The strains at the midspan of Slab-A increased from 0.22% at the beginning of the fire to 0.33% within 20 minutes. The readings of the strain gauges correlate well with the deflection response of the slabs which showed relatively a quick increase in deflections

Strain Gauge Functionality Limit

25 00

30 00

S5-Slab-A S2-Slab-A S5-Slab-B S2-Slab-B

15 00

20 00

Micro Strain

35 00

during the first 20 minutes of the test.

0

5

10

15

20

25

30

Time (min)

Fig. 16. Selected strains in the reinforcing bars vs fire exposure time (see Fig. 2 to locate the gauges)

The fact that the strain readings eventually began to decrease despite increasing curvature of the slabs supports indicates that the gauges failed to function after certain minutes of fire exposure. Three potential reasons could explain the incorrect readings of the strain gauges. Firstly, the operational temperature of the adhesive used to adhere the gauge to the abraded FRP bar surface was reported as 95 °C. This results in a weakened bond between the strain gauge and FRP. Secondly, the functional temperature of the strain gauge was reported as 85 °C, and there was no guarantee for the correct readings of the sensors. Finally, the glass transition temperature of the resins used as the interface between the sand coating and the core of the bars as well as the resin between fibres was 110 °C, and reaching that temperature would affect the bond of the reinforcement with any other materials (i.e. either strain gauges or concrete). 22

Compression strains on the unexposed surface at the midspan and 1.0 m from the north end are shown for Slab-B and only at the midspan for Slab-A in Fig. 17. The compressive strains were slightly higher at the midspan of Slab-B as compared with Slab-A and the strains obtained at the end of the test were as high as 0.16 and 0.18% for Slab-A and Slab-B, respectively. The strain gauges performed well because the maximum temperature at the end of test measured at the top surface of concrete was 102 °C.

Fig. 17. Concrete strains on top surface of Slab-B (see Fig. 2 to locate the gauges)

3.3 Failure modes and visual observations According to ASTM E119 [13], the average temperature increase of the unexposed surface of the slab with respect to the room temperature has to remain below 139 °C. In addition, passage of flame through the slab is not permitted. Neither of these failure criteria associated with thermal behaviour occurred during the fire. The strength of FRP reinforced concrete elements at high temperatures relies primarily on the tensile strength of the reinforcing FRP bars and FRP-concrete bond strength at such extreme conditions. Depending on the fire extent, the affected area as well as reinforcing layout, FRP bar rupture, or pullout failure might occur. Providing adequate anchorage at the ends by either 23

protecting some length of the FRP reinforcement from fire or using bent reinforcement could enhance the fire resistance and change the early bond failure to tensile rupture of bars [4]. In FRP reinforced concrete with low cover, the effects of the thermal expansion of FRP materials are more evident, and the bursting stresses caused by the notably different thermal expansion coefficient of FRP and concrete will create micro cracks [20]. In addition, moisture in the concrete will create steam, thereby inducing internal pressure that can cause spalling. Nonetheless, in Slab-A and Slab-B, no signs of spalling were observed because of normal strength concrete with 10 months of drying before testing. Both slabs carried the superimposed load for more than three hours. At 184 minutes, the load was increased gradually causing Slab-A to fail. The holding frame was removed from the top of the furnace as a safety precaution at the NRC Fire Laboratory. Fig. 18(a) shows the slabs immediately after the fire test as flames were still evident underneath the slabs. The flames were from the burning of GFRP spacer bars that had been used to secure the bottom reinforcements layer in place and ensure the 60 mm of concrete cover. Fig. 18(b) shows the excessive deflection of Slab-A. Because the complex loading system of the floor furnace, the hooks of the slabs had to be cut so that the congested loading pads could be accommodated on the slabs. Consequently, there was less control over the specimens to stop them from falling into the furnace and the potential damage after failure. When the load was increased after three hours, Slab-A started to deform rapidly, and a gap appeared between Slab-A and the adjacent filling block. The gap allowed flames to pass through increasing the risk of possible damage to the overhead loading system. When the load reached 23.2 kN/m, Slab-A failed and Slab-B was unloaded.

24

(a)

(b)

(c)

Fig. 18. (a) Slabs in the frame moved away from the furnace; (b) failure of Slab-A; (c) Slab-B after fire test

As described in Section 2.4, the load applied on the slabs during fire was well beyond the expected load on the slabs in a real fire incident. Despite this, the slabs endured more than three hours of fire exposure under a sustained load causing a moment of 45 kN.m and only failed once the moment reached 52.3 kN.m. Post-fire investigations of the failed slab were performed by cutting both ends of the failed Slab-A using a concrete saw to reveal the end conditions of the reinforcement (Fig. 19). A rectangular prism, 120 × 65 ×1200 mm, was cut to reach the reinforcing bars in the concrete.

(a)

(b)

(c)

Fig. 19. Investigation of the failed slab after fire test: (a) Unexposed zone; (b) Cutting the concrete; (c) Pulled out reinforcing bars at one end

As it was evident from the material tests conducted on Rebar-A bars at high temperatures, the bars were able to carry a sustained tensile stress as high as 375 MPa at 545 °C [12]. The 25

maximum stress in the bottom reinforcement of the slabs was calculated as 158 MPa at room temperature. The maximum measured temperature at the top and bottom of the reinforcement in the exposed zone was less than 300 and 450 °C. The failure mode as shown in Fig. 20(a) was initiated by the pullout of the reinforcing bars at the ends of the slab. The observed failures were of a similar pattern in all bars of the lower reinforcing layer with various amounts of slip, which resulted from FRP-concrete bond deterioration as temperature continuously increased. According to the pullout test results [12], Rebar-A with 68 mm embedment length could still hold 18 kN at 135 °C. Considering 175 mm of each reinforcing bar was located in the unexposed zone, at an average temperature of 135 °C along the reinforcing bar in the unexposed zone, the bars would pullout under 46 kN. FRP-concrete bond failure was expected to occur given that the maximum calculated tensile force in the reinforcing bars due to the applied load was 32 kN before the start of fire. Temperatures varied non-uniformly along the reinforcing bars in the unexposed zone ranging from 100 to 250 °C at 75 and 150 mm from the ends of the slabs, respectively. In Slab-A failure appeared due to progressive loss of bond, which initiated slip of all of the individual reinforcing bars leading to large deflections. Loss of bond at the end zone of Slab-A resulted in a slip of approximately 30 mm leading to substantial deflection of the slabs and subsequent rupture of the reinforcing bars at midspan. If the rupture of the reinforcing bars had happened first, the reinforcing bars at the ends would not have pulled out since the rupture of the reinforcing bars would have released the force in the bonded anchor region.

26

(a)

(b)

(c)

(d)

Fig. 20. Failure mode observations: (a) pulled out reinforcing bars at left end; (b) tensile rupture of all reinforcing bars at midspan; (c) charred and torn reinforcing bars; (d) undamaged reinforcing bars at the right end

4

Numerical modelling

A numerical estimation is shown to provide additional insight to supplement standard fire tests. Numerical modelling can quickly evaluate the heat distributions and the temperatures of FRP and concrete providing an approximation of the fire resistance of the structural members. The thermal evaluations of FRP reinforced concrete sections will facilitate prediction of the FRP performance in different locations of the member corresponding to either fire-exposed or unexposed zones. The fire test results provide a basis for verification of this numerical modelling. Thermal properties of the concrete are significant factors in the model because they determine the heat propagation within concrete. As the temperature increases, thermal properties such as specific heat, conductivity, and expansion change with temperature. Thermal heat transfer in the materials is carried out in three ways which are conduction, convection, and radiation. Through conduction property of a solid material heat transfers from the fire-exposed surface of the solid to the cool region through the thickness of the material. Before heat can transfer within the thickness of a sloid, it needs to reach the surface of the exposed element. In a standard fire test, the gas temperature inside of the furnace follows the

27

time-dependent fire curves and this creates a transient-temperature condition. There are three approaches in defining the heat on the exposed surface of the element: 1)If the time-temperature record of the exposed surface of the element is known, the heat transfer analysis can be performed without involving convection and radiation of heat in the furnace. 2)If the fire load is known, heat flux can be defined to be applied to the element. 3)Generalized heat convection transfer parameters as well as the emissivity values of the concrete for a standard fire test in the European Code [21] can be used to analysis heat transfer from hot gas in a standard fire test furnace by the means of convection and radiation. In the current analysis, the third method was employed. Heat Transfer analysis feature in ABAQUS [22] is capable of performing convective and radiative heat transfer analysis. Convection heat transfer coefficient for the exposed surface was defined as 25 and 9 W/m2 K for the unexposed top surface according to the recommendation of Eurocode for a standard furnace [21]. With accurately computing the temperature of the exposed surface throughout the fire exposure time, the accuracy of the model relies on the appropriate material characteristics that vary widely with the type and quantity of aggregate used in the concrete mix [23]. Temperature dependent specific heat capacity and thermal conductivities of concrete must be defined properly into the finite element model. Then, the temperature profile within the concrete was computed through nonlinear heat conduction analysis. The thermal properties of concrete are defined in several sources such as Lie [19] and the European Code [21]. These two sources define thermal conductivity and specific heat in a fairly similar way except that the specific heat defined by Lie

28

spikes between 400 to 600 °C because of the assumed presence of quartz. Moisture content and aggregate type affect the specific heat of concrete, but the effects of moisture become insignificant when temperatures are greater than 200 °C. Although the specimens were cured for 10 months before fire test, they still had water contents of approximately 5%. During fire and upon reaching 100 °C, water begins to vaporize. Then, all the supplied heat to the specimens is assumed to evaporate the moisture until the specimen has fully dried. The effect of water content in temperature distribution within the specimens is taken into account by incorporating the sudden increases in specific heat properties of concrete. In the current modelling, thermal conductivity and specific heat were taken from European Code [21] for a normal weight carbonate concrete. Two assumptions were made to simplify the numerical model: 1) the modelling presented here excludes the effects of moist migration towards the cool regions of the heated concrete, and 2) the presence of internal reinforcing bars was considered insignificant in affecting the heat transfer through the specimens. To model the heat propagation into the anchor zone, the model had to be able to calculate thermal conduction in two directions. This was critical because the thermal field in the anchor zone had a significant influence of the failure behaviour of the specimens. The bidirectional heat propagation in Fig. 21 shows the contour of the thermal distribution near the support after three hours of fire exposure from underneath. In the modelling, the support regions (200 mm at both ends of the slabs) were not directly exposed to fire. In addition, the soffit of Slab-A is shown in Fig. 21 and the extent of damage to the fireexposed zone can be visually compared to the adjacent protected zone.

29

Fig. 21. Protected and exposed zones in the model and the slab after fire test near the support

The model was verified by comparing the numerical results with the experimental measurements at several locations along the slabs depth. As shown in Fig. 22, the model estimated the temperatures of the slabs within a very close range of experimental values for Slab-A at all depths. The FE temperature predictions for Slab-B at depth of 20 mm from the heated slab bottom surface is off by approximately 100 °C until the results converge at 120 minutes. At 40 mm and 60 mm, the model’s prediction was within a reasonable agreement with the measurements from the experiments. Given the accurate predictions of the model for Slab-A, the variation between FE model and the thermal measurements at the depth of 20 mm of Slab-B can be attributed in part to the potential dislocation of the sensors located in the concrete prior to casting (see section 2.3)

30

1200

800

Soffit FE-20 mm FE-40 mm FE-60 mm Exp-20 mm Exp-40 mm Exp-60 mm

1000

Temperature (°C)

1000

Temperature (°C)

1200

Soffit FE-20 mm FE-40 mm FE-60 mm Exp-20 mm Exp-40 mm Exp-60 mm

600 400 200

800 600 400 200

0

0 0

30

60

90

120

150

180

210

Time (min)

0

30

60

90

120

150

180

210

Time (min)

(a)

(b)

Fig. 22. FE model thermal predictions within the thickness of (a) Slab-A; (b)Slab-B

As explained in section 3.3, bond degradation due to the heat propagation toward the protected cool zone was the primary reason of the failure. The numerical model’s capability to predict the non-uniform lengthwise thermal distribution is a valuable feature to predict the failure of FRP reinforced slabs. The likely displacement of the thermocouples during the concrete casting is one reason for some differences between the readings in the identical positions of Slab-A and Slab-B. Moreover, slight variances in the length of protected zone can significantly influence the thermal distribution in the protected zone. In a simple approach, heat transfer via convection and radiation was assigned only to the fire-exposed zone. Temperatures throughout the fire time in the anchor zone at the positions of 75 and 150 mm from the ends of the slabs at the depth of 60 mm are shown in Fig. 23. The FE model has clearly underestimated the temperatures in the anchor zones as compared to the experimental readings from both slabs.

31

300

300

Temperature (°C)

250 200

Exp-R3 Exp-R7 FE-150 mm Exp-R1 Exp-R9 FE-75 mm

Slab-A 250

Temperature (°C)

Exp-R3 Exp-R7 FE-150 mm Exp-R1 Exp-R9 FE-75 mm

150 100 50

200

Slab-B

150 100 50

0

0 0

30

60

90

120

150

180

210

Time (min)

0

30

60

90

120

150

180

210

Time (min)

Fig. 23. Test and the model temperatures in the unexposed zone

Authors have investigated the effect of insulating layers in the anchor zone and the temperature readings on the surface of the protected zone. In contrast to the concept of insulating, temperatures were fairly high in the protected zone particularly in vicinity of fire-exposed zone. The intensity of heat quickly lessens as moving away from the fire-exposed zone. Under these circumstances, modification in the finite element was required to account for the effects of heat in the protected zone. In an iterative approach, the modified heat transfer through convection process was utilized in the protected zones. The concrete surface emissivity and convection factors proposed by Eurocode 1992-1-2 [21] were implemented into the finite element model. The factors were chosen for the case of a standard fire test where the bottom surface of a concrete element, e.g. slab, is exposed to hot air rising from a fire in the furnace. Since the unexposed zones were protected from direct fire exposure and hot air, the convection factor must be different from the values recommended for the directly exposed surfaces of slab on a standard fire test furnace. Several values were tried, and the temperature prediction in the unexposed zone

32

was investigated until the predicted temperatures were close to the measured values. Heat transfer through radiation in the unexposed zones were excluded from the analysis. When the convection factor was taken as 15 W/m2 in the unexposed zone, the closest predictions were achieved. It should be mentioned that the test setup parameters such as the width and thickness of the insulating layers and the effectiveness of the layers in the conducted full-scale fire tests is not a universal practice nor completely resembles the heat transfer conditions at the ends of a slab or beam in a real fire incident. The modified heat transfer analysis here showed that the realistic parameters must be used for every distinct situation. The FE model temperature predictions with the modified heat transfer in the protected zone (Fig. 24) show an improved accuracy was achieved. At 150 mm away from the end of Slab-A, the model estimated the temperature within the range of the measured values from thermocouples R3 and R7 at the same locations at two ends. For Slab-B, the model at both locations (75 and 150 mm) is in agreement with the experimental measurements.

300

300

Temperature (°C)

250 200

Slab-A

Exp-R3 Exp-R7 FE-150 mm Exp-R1 Exp-R9 FE-75 mm

250

Temperature (°C)

Exp-R3 Exp-R7 FE-150 mm Exp-R1 Exp-R9 FE-75 mm

150 100 50

200

Slab-B

150 100 50

0

0 0

30

60

90

120

150

180

210

Time (min)

0

30

60

90

120

Time (min)

Fig. 24. Test and the model temperatures in the unexposed zone

33

150

180

210

5

Discussion

A full-scale fire test on two GFRP reinforced concrete slabs under flexural loading was performed and the results were very promising for the fire safety of FRP reinforced concrete slabs. The tests investigated the fire safety of the current practice of GFRP reinforced concrete elements with relatively thick concrete cover which has been prescribed conservatively by codes to ensure fire safety [8]. Better understanding of the response of GFRP reinforced slabs in fire was achieved by an insightful investigation into the failure pattern. The most influential factor in the fire resistance of GFRP reinforced members was found to be FRP-concrete bond strength at elevated temperatures. The unexposed length of the members (i.e., the cool anchor zone) is a significant factor in the fire resistance of FRP reinforced concrete elements. For these tests, the anchor zone was very short (only 200 mm) and the slabs were simply-supported without any restraint. Thus, the conditions represented one of the most severe situations in terms of the fire performance of FRP reinforced concrete slabs. The maximum strains in the bars immediately before the fire were measured to be between 2100 and 2300 microstrain (larger than the serviceability strain limit of 2000 microstrain recommended by CSA S806 [8]). Also, all bars extended continuously into the supports. As such, the results may not be conservative for situations with higher strains in the bars or cases where a high proportion of the bars are either cut-off or lap-spliced in the fire-exposed zone. Structural failure of the slab was caused by bond failure of the GFRP reinforcing bars in the unexposed region over the supports. From material tests on such bars [12], the bond strength reduces considerably at temperatures in the region of the glass transition temperature and thus structural failure is expected when the temperature in unexposed region exceeds the glass transition temperature. As such, accurate modelling of temperatures in the unexposed region has 34

direct correlation with potential structural failure. On the other hand, if the bars are sufficiently anchored into unexposed regions where the temperatures remain below the glass transition temperature, then structural failure may occur by tension failure when the temperatures in the bars exceed 400 °C [11]. In the current fire tests, the maximum temperatures in the bars were below 400 °C and thus bond was the governing failure mode. FRP reinforced concrete slabs must be designed considering potential cool anchor zones which would be available for the reinforcing bars to retain the minimum required bond to concrete during a fire incident. Although better fire endurance can be associated with thicker concrete cover, this reduces the effectiveness of the section, and given the lower serviceability performance (deflection) of GFRP reinforced members compared to steel reinforced concrete members, thicker cover is not favorable to the use of GFRP. The results suggest that safe designs could be achieved using thinner concrete covers for the fire design of FRP reinforced concrete members, which is the subject of additional tests to be conducted. Concrete cover could also be reduced by increasing the length of the unexposed zone (i.e., the cooler anchor zone). Splicing of bottom FRP reinforcing bars is not recommended along the spans of elements susceptible to fire; splices should be positioned over supports and unexposed zones wherever possible. 6

Conclusions

A full-scale fire test on two GFRP reinforced concrete slabs with 60 mm of clear concrete cover under flexural loading demonstrated fire endurance of over 3 hours when exposed to the standard ASTM E119 fire curve. The load during the fire was chosen to be greater than the full service load expected on the slabs based on serviceability design criteria (deflection and crack control) that govern the design of FRP reinforced concrete slabs. The three hours of fire resistance under a severe loading condition of 55% of the ultimate moment resistance of the slab showed that fire 35

resistance estimations by CSA S806 are conservative, and that FRP reinforcing bars can be used with lower concrete cover. Although material tests on the bond strength of the reinforcing bars have shown that most of the bond strength is lost at temperatures in the range of the glass transition temperature of the bars, an anchorage length of 200 mm in the unexposed zones was sufficient to anchor the bars. Bond failure occurred since the slabs were simply supported without axial restraint with only 200 mm at each of the slabs not exposed to the fire. None of the reinforcing bars ruptured although the post-fire investigation showed that the resin of the bars was fully burned. Numerical modelling showed satisfactory prediction of the thermal behaviour of the concrete section, and could be used to determine the strength loss of the FRP materials in various locations of the member. Acknowledgments The authors would like to thank the Natural Sciences and Engineering Research Council of Canada (NSERC), Pultrall Inc., and BP Composites for financial and material support. The authors also appreciate the insightful advice of Dr. Benmokrane and cooperation of H. Mohammad (Sherbrooke University). The authors are also grateful to the technical staff of the Fire Research Laboratory at NRC. References [1] Abbasi A, Hogg PJ. Fire testing of concrete beams with fibre reinforced plastic rebar. Composites Part A: Applied Science and Manufacturing 2006;37(8):1142-50. [2] Rafi M, Nadjai A, Ali F. Fire resistance of carbon FRP reinforced-concrete beams. Magazine of Concrete Research 2007;59(4):245-55. [3] Weber A. Fire-resistance tests on composite rebars. In: Proceedings of International Conference on FRP Composites in Civil Engineering (CICE2008). Zurich, Switzerland, 2008. [4] Nigro E, Bilotta A, Cefarelli G, Manfredi G, Cosenza E. Performance under fire situations of concrete members reinforced with FRP rods: bond models and design nomograms. Journal of Composites for Construction 2011;16(4):395-406. 36

[5] Hajiloo H, Green MF, Bénichou N, Sultan M. Fire Performance of FRP reinforced concrete slabs. In: Proceedings of 7th International Conference on Advanced Composite Materials in Bridges and Structures. Vancouver, Canada, 2016. [6] Lie T. Calculation of the fire resistance of composite concrete floor and roof slabs. Fire technology 1978;14(1):28-45. [7] Kodur VK, Bisby LA, Foo SH-C. Thermal behavior of fire-exposed concrete slabs reinforced with fiber-reinforced polymer bars. ACI structural journal 2005;102(6). [8] CSA (Canadian Standards Association). Design and construction of building components with fiber-reinforced polymers. CSA-S806, Mississauga, ON, Canada: 2012. [9] Adelzadeh M, Hajiloo H, Green MF. Numerical Study of FRP Reinforced Concrete Slabs at Elevated Temperature. Polymers 2014;6(2):408-22. [10] Bisby LA, Kodur VK. Evaluating the fire endurance of concrete slabs reinforced with FRP bars: Considerations for a holistic approach. Composites Part B: Engineering 2007;38(5):547-58. [11] Hajiloo H, Gales J, Noël M, Green MF. Material Characteristics of Glass Fibre Reinforced Polymer (GFRP) Bars at High Temperature. In: Proceedings of PROTECT Fifth International Workshop on Performance, Protection & Strengthening of Structures under Extreme Loading. East Lansing, MI, 2015. p. 94-104. [12] Hajiloo H, Green MF. Bond Strength of Glass Fibre Reinforced Polymer Bars in Concrete at High Temperature. In: Proceedings of Conference of the Canadian Society for Civil Engineering. Regina, SK, 2015. [13] ASTM. Standard Test Methods for Fire Tests of Building Construction and Materials. ASTM E119, West Conshohocken, PA: 2015. [14] Hertz KD. Limits of spalling of fire-exposed concrete. Fire safety journal 2003;38(2):10316. [15] Underwriters’ Laboratories of Canada. CAN/ULC-S101-07-Standard Methods of Fire Endurance Tests of Building Construction and Materials. 2007. [16] ACI (American Concrete Institute). Guide for the Design and Construction of Concrete Reinforced with FRP Bars. ACI 440-1R, Detroit, Michigan: 2006. [17] ISIS Canada. Design Manual 3. Reinforcing concrete structures with fiber reinforced polymers. Winnipeg, MB, Canada: 2007. [18] Harmathy TZ. Fire safety design and concrete: John Wiley & Sons, New York; 1993. [19] Lie TT. Fire and buildings: Applied Science Publishers Ltd; 1972. [20] Galati N, Nanni A, Dharani LR, Focacci F, Aiello MA. Thermal effects on bond between FRP rebars and concrete. Composites Part A: Applied Science and Manufacturing 2006;37(8):1223-30. [21] CEN (European committee for standardization). Eurocode 2: Design of concrete structures Part 1-2: General rules- Structural fire design. EN 1992-1-2, Brussels, Belgium: 2004. [22] Hibbitt, Karlsson, Sorensen. ABAQUS/Standard user’s manual version 6.14: Hibbitt, Karlsson, & Sorensen, Inc; 2014. [23] Lie TT. ASCE Manuals and Reports on Engineering Practice No. 78, Structural Fire Protection. New York, NY: American Society of Civil Engineers; 1992.

37

Figure captions Fig. 1. Reinforcement details and dimensions of the slab in longitudinal and transverse sections .............................. 5 Fig. 2. Thermocouple, strain gauges, and labelling system ....................................................................................... 8 Fig. 3. Instrumenting and casting of the slabs: (a) TCs placements at the end unexposed zone of Slab-A; (b) TCs installed on reinforcing bars; (c) Fully instrumented Slab-B .................................................................................... 9 Fig. 4. Floor furnace test setup .............................................................................................................................. 11 Fig. 5. (a) NRC floor furnace facility; (b) Several jacks to apply uniform load on each slab; (c) slabs on furnace.... 12 Fig. 6. Thermal readings within the thickness ........................................................................................................ 14 Fig. 7. Thermal readings on the main middle rebar ................................................................................................ 15 Fig. 8. Thermal readings in the midspan on three main reinforcing bars ................................................................. 16 Fig. 9- Temperatures at the anchorage zone on the bottom of the bars.................................................................... 17 Fig. 10. Temperature gradient in unexposed zones ................................................................................................ 18 Fig. 11. Marked cracks on the soffit of Slab-A after loading .................................................................................. 19 Fig. 12. Strains on the reinforcing bars and top surface of concrete before fire (see Fig. 2 to locate the gauges): (a) Slab-A; (b) Slab-B................................................................................................................................................ 19 Fig. 13. Loading history of the slabs before fire test .............................................................................................. 20 Fig. 14. Deflection behaviour of the slabs during fire as a function of: (a) fire exposure time; (b) furnace temperature ............................................................................................................................................................................ 21 Fig. 15. Strains in the reinforcing bars (see Fig. 2 to locate the gauges) vs temperatures in (a) Slab-A and (b) Slab-B; ............................................................................................................................................................................ 21 Fig. 16. Selected strains in the reinforcing bars vs fire exposure time (see Fig. 2 to locate the gauges).................... 22 Fig. 17. Concrete strains on top surface of Slab-B (see Fig. 2 to locate the gauges) ................................................ 23 Fig. 18. (a) Slabs in the frame moved away from the furnace; (b) failure of Slab-A; (c) Slab-B after fire test .......... 25 Fig. 19. Investigation of the failed slab after fire test: (a) Unexposed zone; (b) Cutting the concrete; (c) Pulled out reinforcing bars at one end.................................................................................................................................... 25 Fig. 20. Failure mode observations: (a) pulled out reinforcing bars at left end; (b) tensile rupture of all reinforcing bars at midspan; (c) charred and torn reinforcing bars; (d) undamaged reinforcing bars at the right end .................. 27 Fig. 21. Protected and exposed zones in the model and the slab after fire test near the support ............................... 30 Fig. 22. FE model thermal predictions within the thickness of (a) Slab-A; (b)Slab-B ............................................. 31 Fig. 23. Test and the model temperatures in the unexposed zone............................................................................ 32 Fig. 24. Test and the model temperatures in the unexposed zone............................................................................ 33

38