Anda di halaman 1dari 142

FRP REINFORCED CONCRETE AND ITS APPLICATION IN BRIDGE SLAB DESIGN

by

YUNYI ZOU

Submitted in partial fulfillment of the requirements For the degree of Doctor of Philosophy

Dissertation Adviser: Dr. Arthur Huckelbridge Supported by Saada Family Fellowship

Department of Civil Engineering CASE WESTERN RESERVE UNIVERSITY

January, 2005

CASE WESTERN RESERVE UNIVERSITY SCHOOL OF GRADUATE STUDIES

We hereby approve the dissertation of

Yunyi Zou ______________________________________________________


candidate for the Ph.D. degree *.

(signed)_______________________________________________ (chair of the committee)

Arthur Huckelbridge

________________________________________________

Clare Rimnac

________________________________________________

Dario Gasparini

________________________________________________

Robert Mullen

________________________________________________

________________________________________________

(date) _______________________

Nov. 11, 2004

*We also certify that written approval has been obtained for any proprietary material contained therein.

Dedication

To my parents Zou JiShen and Chen XiuFang To my wife Yuping

Table of Contents
Table of Contents List of Figures List of Tables Acknowledgements List of Abbreviations 1 3 10 11 12

Abstract

13

Chapter 1

Background and Introduction of the Problem

15

Chapter 2

Experimental Analysis of FRP Reinforced Concrete under Fatigue Load 36

Motivation for the Testing Program Description of Testing Program Experimental Results Qualitative Discussion

36 37 43 55

Chapter 3

Simulation of Crack Growth

78

Estimation of Crack Opening Estimation of Crack Growth Sensitivity Analysis of the Model and Variation of Crack

78 83

Growth Estimation Simulation of Experiment Results

90 98

Chapter 4

Finite Element Modeling and Analysis of a Realistic FRP Reinforced Concrete Slab 104

Analysis of Slab Strips Analysis of Full Bridge Slabs Empirical Design of Bridge Slabs

104 115 127

Chapter 5

Conclusions

130

Chapter 6

Future Research

134

References

136

List of Figures

Figure 2.1 Figure 2.2 Figure 2.3 Figure 2.4 Figure 2.5 Figure 2.6 Figure 2.7 Figure 2.8 Figure 2.9 Figure 2.10 Figure 2.11 Figure 2.12 Figure 2.13 Figure 2.14 Figure 2.15 Figure 2.16 Figure 2.17 Figure 2.18 Figure 2.19

Aslan 100 GFRP by Hughes Brothers Isorod by Pultrall Specimen Section Details and Loading Condition Cyclic Load Test Setup Sketch of Data Acquisition System Specimen C5 x 8.5 H5 Specimen C3 x 8.5 H5 Specimen C4 x 8.5 H5 Specimen C6 x 8.5 H5 Specimen C3 x 8.5 P5 Specimen C4 x 8.5 P5 Specimen C5 x 8.5 P5 Specimen C5 x 8.5 P5 Specimen C5 x 8.5 P5OL Specimen C5 x 8.5 S5 Injecting Dye into Cracks Typical Crack Profiles Definitions of Elastic and Plastic CMOD Elastic CMOD under Ramp Load vs Number of Cycles for Group H Specimens (Pmin=2.2 KN Pmax=15.6 KN)

38 39 40 41 42 44 45 45 46 48 49 50 51 51 52 54 54 56

57

Figure 2.20

Plastic CMOD under Ramp Load vs Number of Cycles for

Group H Specimens (Pmin=2.2 KN Pmax=15.6 KN) Figure 2.21 Elastic CMOD under Ramp Load vs Number of Cycles for Group P Specimens (Pmin=2.2 KN Pmax=15.6 KN) Figure 2.22 Plastic CMOD under Ramp Load vs Number of Cycles for Group P Specimens (Pmin=2.2 KN Pmax=15.6 KN) Figure 2.23 Figure 2.24 Figure 2.25 Figure 2.26 Figure 2.27 Figure 2.28 Hysteresis of Beam C4x8.5H5 under Cyclic Load Hysteresis of Beam C3x8.5H5 under Cyclic Load Hysteresis of Beam C5x8.5H5 under Cyclic Load Hysteresis of Beam C6x8.5H5 under Cyclic Load Hysteresis of Beam C3x8.5P5 under Cyclic Load Hysteresis of Beam C4x8.5P5 under Cyclic Load Static Pre-Cracking Figure 2.29 Hysteresis of Beam C4x8.5P5 under Cyclic Load Fatigue Pre-Cracking Figure 2.30 Figure 2.31 Figure 2.32 Figure 2.33 Figure 2.34 Hysteresis of Beam C5x8.5P5 under Cyclic Load Hysteresis of Beam C6x8.5P5 under Cyclic Load Unit Width Pseudo Energy Loss vs Number of Cycles in Group H Unit Width Pseudo Energy Loss vs Number of Cycles in Group P Effect of 40% Overload on CMOD Overload Pmax=22.3 KN Beam C5x8.5H5 Figure 2.35 Effect of 30% Overload on CMOD Overload Pmax=20 KN Beam C6x8.5H5 Figure 2.36 Effect of 40% Overload on CMOD Overload Pmax=22.3 KN

58

58

59 61 62 62 63 63

64

64 65 65 66 67

68

69

Beam C5x8.5P5 Figure 2.37 Effect of 40% Overload on CMOD Pmax=22.3 KN Beam C6x8.5P5 Figure 2.38 Elastic and Plastic CMOD under Ramp Load vs Number of Cycles for Specimens C5x8.5H5OL, C5x8.5S5 and C5x8.5H5M (Pmin=2.2 KN Pmax=15.6 KN) Figure 2.39 Figure 2.40 Hysteresis of Beam C5x8.5S5 under Cyclic Load Effect of 40% Overload on CMOD Overload Pmax=22.3 KN Beam C5x8.5S5 Figure 2.41 Unit Width Pseudo Energy Loss vs Number of Cycles in Specimen C5x8.5P5OL and C5x8.5S5 Figure 2.42 Specimen C5 x 8.5 H5M

69

70

71 73

73

74 75

Figure 3.1 Figure 3.2

Debonded Length Representation A Typical Finite Element Mesh with Debonded Length Representation of Specimen C4x8.5P4

79

80 82

Figure 3.3 Figure 3.4

Fictitious Material Representation A Typical Finite Element Mesh with Fictitious Material Representation of Specimen C4x8.5P4

82 84 85 88

Figure 3.5 Figure 3.6 Figure 3.7 Figure 3.8

Assumed Stress Distribution at a Cracked Section A Hinge Model Verification of Hinge Model Sensitivity Analysis Results on Parameter C

for Beam C6x8.5H5 (m=3.76) Figure 3.9 Sensitivity Analysis Results on Parameter m for Beam C5x8.5H5 (C=6.76x10.4) Figure 3.10 Sensitivity Analysis Results on Initial Crack Length a0 for Beam C5x8.5H5 (C=6.76x10.4 m=3.76) Figure 3.11 Sensitivity Analysis Results on Initial Crack Spacing L for Beam C5x8.5H5 (C=6.76x10.4 m=3.76) Figure 3.12 Sensitivity Analysis Results on Concrete Elastic Modulus Ec for Beam C5x8.5H5 (C=6.76x10.4 m=3.76) Figure 3.13 Sensitivity Analysis Results on Concrete Elastic Modulus Ef for Beam C5x8.5H5 (C=6.76x10.4 m=3.76) Figure 3.14 Sensitivity Analysis Results on Specimen Width b for Beam C5x8.5H5 (C=6.76x10.4 m=3.76) Figure 3.15 Sensitivity Analysis Results on Specimen Height h for Beam C5x8.5H5 (C=6.76x10.4 m=3.76) Figure 3.16 Crack Length and Crack Opening Increment versus Number of Cycles, Beam C3x8.5H5, C=6.76x10.4, m=3.48 Figure 3.17 Crack Length and Crack Opening Increment versus Number of Cycles, Beam C4x8.5H5, C=6.76x10.4, m=3.57 Figure 3.18 Crack Length and Crack Opening Increment versus Number of Cycles, Beam C5x8.5H5, C=6.76x10.4, m=3.76 Figure 3.19 Crack Length and Crack Opening Increment versus Number of Cycles, Beam C3x8.5P5, C=6.76x10.4, m=3.39

94

94

95

95

96

96

97

97

99

100

100

101

Figure 3.20

Crack Length and Crack Opening Increment versus Number of Cycles, Beam C4x8.5P5, C=6.76x10.4, m=3.55 101

Figure 3.21

Crack Length and Crack Opening Increment versus Number of Cycles, Beam C5x8.5P5, C=6.76x10.4, m=3.74 102

Figure 3.22

Crack Length and Crack Opening Increment versus Number of Cycles, Beam C6x8.5P5, C=6.76x10.4, m=3.88 102 103

Figure 3.23

Size Effect of Beam Width on Paris Equation

Figure 4.1

Slab Strip Model under One Wheel Load (Girder spacing 1.8m, 16M Bar at 100mm, Slab thickness 215mm) 106

Figure 4.2

Transverse Normal Stress Contours Under One Wheel Load of Design Truck (Girder spacing 1.8m, 16M Bar at 100mm) 107

Figure 4.3

Slab Strip Model under One Axle Load of Design Truck (Girder spacing 1.8m, 16M Bar at 100mm, Slab thickness 215mm) 108

Figure 4.4

Transverse Normal Stress Contours Under One Axle Load of Design Truck (Girder spacing 1.8m, 16M Bar at 100mm) 108

Figure 4.5

Slab Debonded Length Representation under One Axle Load (Girder spacing 2.7m, 16M Bar at 100mm, Slab thickness 215mm) 109

Figure 4.6

Transverse Normal Stress Contours Under One Axle Load of Design Truck(Girder spacing 2.7m, 16M Bar at 100mm) 109

Figure 4.7

Slab Strip Model under One Axle Load of Design Truck(Girder spacing 3.6m, 16M Bar at 100mm, Slab thickness 215mm) 110

Figure 4.8

Transverse Normal Stress Contours Under One Axle Load of

Design Truck (Girder spacing 3.6m, 16M Bar at 100mm) Figure 4.9 Transverse Normal Stress Contours Under One Axle Load of Design Truck (Girder spacing 1.8m, 16M Bar at 150mm) Figure 4.10 Figure 4.11 Bar Stress Under Design Truck Load with 16M Bars Slab Model under Load of Design Truck, Lane Load and Self-Weight (Girder spacing 1.8m, Slab thickness 215mm, No Diaphragm) Figure 4.12 Transverse Normal Stress Contours under Loads of Design Truck, Lane Load and Self-Weight. Figure 4.13 Slab Model under Load of Design Truck Only with Girders Fixed Vertically (Girder spacing 1.8m, Slab thickness 215mm) Figure 4.14 Transverse Normal Stress Contours Under Design Truck Only with Girders Fixed Vertically. Figure 4.15 Transverse Normal Stress Contours Under Design Truck Only with Diaphrams. Figure 4.16 Slab Model under Load of Design Truck, Lane Load and Self-Weight (Girder spacing 2.7m, Slab thickness 215mm) Figure 4.17 Transverse Normal Stress Contours Under Loads of Design Truck, Lane Load and Self-Weight. Figure 4.18 Slab Model under Load of Design Load and Self-Weight. (Girder spacing 3.6m, Slab thickness 215mm) Figure 4.19 Transverse Normal Stress Contours Under Design Load and Self-Weight.

110

111 112

120

120

121

121

122

122

123

123

124

Figure 4.20 Figure 4.21

Maximum Crack Opening under Design Load in Model Bridge Maximum Crack Opening Under Design Load and Ohio Legal Load 5C1 in Model Bridge, 1.8 m Girder Spacing

124

125

Figure 4.22

Slab Model under Load of Ohio Legal Truck Load 5C1 and Self-Weight (Girder spacing 1.8m, Slab thickness 215mm) 125

Figure 4.23

Transverse Normal Stress Contours Under Loads of Ohio Legal Truck Load 5C1 and Self-Weight 126

List of Tables

Table 2.1

Specimen Descriptions

40

Table 3.1 Table 3.2 Table 3.3

Calibrated Debonded Length for Group P Specimens Calibrated Ficticious Material Properties Hinge Assumption Verification

81 83 89

10

Acknowledgements

I want to take this opportunity to thank my advisor Dr. Huckelbridge for his guidance. Throughout my research, I have enjoyed our discussions very much. His teaching will benefit me for years to come.

I also feel so fortunate and blessed to have Dr. Saada as my instructor and sponsor. Nothing in this dissertation would be possible without his support. To me, he is a role model for living and working.

11

List of Abbreviations

AASHTO

American Association of State Highway and Transportation Officials

ACI CMOD FE FRP GFRP LEFM LFD LRFD RC

American Concrete Institute Crack Mouth Opening Displacement Finite Element Fiber-Reinforced Polymers Glass FRP Linear Elastic Fracture Mechanics Load Factor Design Load and Resistance Factor Design Reinforced Concrete

12

FRP Reinforced Concrete and its Application in Bridge Slab Design


by Yunyi Zou

ABSTRACT For decades, bridge slabs have been troubled by the corrosion of steel reinforcements. The unique corrosion resistance of FRP (Fiber-Reinforced Polymers) bars makes them a promising alternative to steel bars. Because of the relatively low elastic modulus of FRP reinforcement, the post-cracking serviceability often is the controlling factor in the flexural design of FRP reinforced concrete. Since bridge deck slabs are under repeated traffic loads, it is the post-cracking serviceability under cyclic loads that becomes vital in the design and maintenance decision-making process.

Experiments have been conducted to investigate the post-cracking flexural performance of FRP RC (reinforced concrete) under constant amplitude cyclic loading. Each specimen tested was a beam with a single FRP bar at the bottom. Two different types of FRP bars were used. The crack opening was monitored for specimens of different size. Up to 2 million cycles of cyclic loads have been applied at 100% service load levels. It has been found that there are two stages in the crack growth of FRP reinforced concrete. The first stage is early growth, which is characterized by increasing crack mouth opening displacement (CMOD). The second stage is the stabilization of CMOD and crack length. No fatigue failure was encountered in the testing under service loading and moderate overloads. The effects of moderate overload on observed crack

13

growth were also investigated.

The performances of two different FRP bars were

compared. A model was proposed to predict long term crack growth in FRP R/C under cyclic loading, based on the Paris equation.

Two FE (finite element) crack representations were examined.

One was a

debonded length representation. In this model it was assumed that there was a debonded length around each crack, within which there was no tangential interaction between concrete and reinforcement. Beyond the debonded length, the interface between concrete and reinforcement was tied with no relative movement. The other representation

examined was a fictitious material crack representation. A fictitious material was placed in a triangular crack cross section, with a maximum width of 2.5mm (0.1 in). Then, the modulus of elasticity of the fictitious material was calibrated, based on the observed testing results, after crack growth had stabilized. Both representations have been used to analyze bridge slabs. Finally, an empirical slab design was discussed.

14

Chapter 1

Background and Introduction of the Problem

The advantages of Fiber-Reinforced Polymers (FRP) include a high ratio of strength to mass, excellent fatigue characteristics, excellent corrosion resistance, electromagnetic neutrality, and a low axial coefficient of thermal expansion. Generally speaking, the disadvantages of FRP reinforcement include its higher cost, lower Youngs modulus (except for Carbon FRP), lower failure strain and lack of ductility. The

transverse coefficient of thermal expansion (CTE) is also much larger than the longitudinal CTE. The long-term strength of FRP can be as low as 70% of its short-term strength, and ultra-violet radiation can damage FRP. FRP reinforcement is also not effective for compression reinforcement because of the compression instability of the slender axial fibers. There is a lot of potential to apply FRP in bridge engineering for structural elements in corrosive environments with low ductility demand.

For decades, reinforced concrete slabs have been used as bridge decks both in United States and around the world. The relatively inexpensive concrete and steel

reinforcement have served very well in most respects. In recent years, rehabilitation of national highway bridges has been a priority, due to the aging and deteriorating superstructures. One of the major causes of superstructure deficiency is the corrosion of steel reinforcement. In this case, the excellent corrosion resistance and light weight of FRP make it potentially superior in long term performance to conventional reinforcing steel, and, particularly in the case of Glass FRP (GFRP), potentially competitive economically.

15

Serviceability covers many different aspects of structural performance related to particular applications. The most commonly encountered serviceability requirements in RC structures are maximum deflection and crack opening control. Cracking is a complex phenomenon, particularly in composite materials. For quasi-brittle materials such as concrete, the tensile stress gradually drops to zero after reaching a peak value. There exists an inelastic zone at the tip of the crack, known as the fracture process zone. Within the fracture process zone, the stress decreases as it approaches the crack tip. Shah (1995) summarized the interaction within the fracture process zone as microcracking, crack deflection, aggregate bridging, crack face friction, crack tip blunting by voids, crack branching, and etc. It has been reported that the measured fracture process zone is almost independent of specimen thickness; the crack length generally is deeper on the sides than in the middle. Consequently, it was pointed out that applicability of linear elastic fracture mechanics (LEFM) is limited for plain concrete to large structures, with a relatively small fracture process zone. In the case of smaller scale structures, the aforementioned

complexity in concrete cracks deters the direct application of LEFM.

In the case of FRP RC beams under bending, as soon as cracking occurs, there is a surge of forces in the bars. Cracks tend to grow in fatigue load environments. The tensile forces in the bars and resultant compressive force in concrete increase as depth of intact concrete and fracture process zone decrease. The relatively low Youngs modulus of FRP, which is about one fifth of the Youngs modulus of conventional steel reinforcement, will generate larger crack lengths and crack mouth opening displacements (CMOD) compared with conventional steel reinforcement, particularly so if

16

reinforcement ratios are similar in magnitude for both cases. Consequently, the aggregate bridging will be less, crack face friction will be smaller, and the zone of microcracking will also be smaller due to suppression by concrete in compression. This is the essential difference between plain concrete, conventional reinforced concrete (RC) and FRP RC, which tends to make LEFM a good approximation for crack modeling for FRP RC.

P. Gergely and L. Lutz (1968) analyzed test results from various investigators on crack openings in conventional reinforced concrete. A multiple regression analysis was performed on crack openings with respect to different variables. It was found that steel stress magnitude was the most important variable. The concrete cover was an important variable but not the only secondary consideration. Bar size was also found not to be a major variable, and crack opening tended to increase with increasing strain gradient. The significant variables identified were effective area of concrete, the number of bars, concrete cover and stress level. The recommended equation for bottom crack in English units was as follows. w = 0.076 f s 3 d c A c
(1-1)

Where is the ratio of the distance from the neutral axis to extreme tension fiber to the distance from the neutral axis to the center of the tensile reinforcement; dc is the concrete cover to bar center; fs is the tensile stress in steel bars; Ac is the effective tension area of concrete.

In ACI 440.1R-01, the Gergely-Lutz equation has been modified to estimate the crack opening of FRP RC members by simply replacing the steel strain with FRP strain.

17

To account for the difference in bonding between steel and FRP, a corrective coefficient kb is introduced. The final equation of crack opening in millimeter is as follows.
w= 2 .2 k b f f 3 d c A c Ef
(1-2)

Where Ef is the Youngs modulus of FRP bar; is the ratio of the distance from the neutral axis to extreme tension fiber to the distance from the neutral axis to the center of the tensile reinforcement; dc is the concrete cover to bar center; ff is the tensile stress in FRP bars; Ac is the effective tension area of concrete. The coefficient kb is assumed to be one for FRP bars having bond behavior, similar to steel bars. Many researchers have suggested different values for different bar surfaces. ACI 440.1R-01 listed values of kb by Gao et al. to be 0.71, 1.00, 1.83 for three currently popular types of GFRP bars. A value of 1.2 was suggested for deformed FRP bars by the report, in the case of no available experimental data.

Carpinteri et al (1993) used a LEFM to model a simply supported steel RC beam. The total stress intensity factor is the superposition of KI due to the bending moment and to the bar force. An energy concept was used to examine the steel yielding, bar slip and crack growth under different conditions. The total energy was calculated in terms of bending moment and rebar force. The relationship between rebar force and bending moment was derived based on the relationship of energy release rate and stress intensity factors. In the analysis of cyclic loading, three cases were discussed based on comparing the magnitudes of peak moment with plastic flow moment, slippage moment and fracture moment. It was assumed that cracks would propagate if the peak moment exceeded the

18

fracture moment. Apparently, this model is more applicable to the case of low cycle fatigue (relatively high rebar stress levels).

In early 1960s, P.C. Paris (1963) applied fracture mechanics to fatigue problems. The proposed equation is as follows.

da = C (K ) m dN

(1-3)

Where a is the crack length; N is the number of cycles; K is the stress intensity factor difference at maximum and minimum loading; C and m are material parameters. Although Pariss law was developed for steel, researchers have tried to verify if it was also valid for concrete.

Considerable work done has been focused on plain concrete. The report by ACI committee 215 provides general knowledge about fatigue strength of concrete and reinforcement. Fatigue fracture of concrete is characterized by considerably large strains and microcracking. The S-N curve of concrete is approximately linear between 102 and 107 cycles, which indicates that there is no apparent endurance limit for concrete. The fatigue strength for a life of 10 million cycles of load and a probability of failure of 50 percent, regardless of whether the specimen is loaded compression, tension or flexure, is approximately 55 percent of the static ultimate strength.

Perdikaris et al. (1987) conducted experiments on single-edge-notched plain concrete beams under four-point bending. Crack length was also recorded based on the CMOD compliance measurements. It was concluded that the Paris equation results in

19

significant errors of 100% although R2s, which is the fraction of the variance in the data that is explained by a regression, were close to one for different specimens. It was believed that large errors were part of the nature with exponentials.

Baluch et al. (1987) also tried to verify if the Paris equation is valid for concrete. The experiments were three-point bending on single-edge-notched plain concrete beams of 51mm wide x 152mm deep x 1360mm. Similarly, a compliance test was first

performed so that crack length could be obtained. For the same beam specimen under different R (=Kmin/Kmax), it was found that Paris equation is applicable in plain concrete. The material parameter m was found to be 3.12, 3.12 and 3.15 at R=0.1, 0.2, 0.3 respectively. It was subsequently concluded that m was independent of R. The material parameter C was reported to be on the order of 10-24 and 10-25; it appeared from the article that the units of C was mm/[Pa m1/2]m , although the units were not stated explicitly. The authors suggested that C might be related to R. Foremans equation (1967) which includes the effects of R was also explored by the authors. da C (K ) m = dN (1 R)( K c K max )
(1-4)

where Kc is the fracture toughness of the material of interest at the appropriate thickness. Different material parameters C and m were inferred under different R values for the same type of specimen, however. Therefore, it was concluded that Foremans equation was not applicable in plain concrete.

Z.P. Bazant et al (1992) investigated the size effect in fatigue fracture of concrete. The specimens under three-point bending were geometrically similar in length, height

20

and notch length. The thickness was constant for all beams. The results of fatigue tests were presented with the plots of log(a/N) versus log(K/KIf). Different lines were obtained for different beam size, although they were parallel to each other. The authors combined the Paris law with a size effect law, for fracture under monotonic loading; the revised Paris law is a function of a size adjusted stress intensity function.

Due to the nature of cracks in concrete, a method of compliance calibration is normally used in crack length determination for pure concrete. However, it has been questioned that effects of the fracture process zone will stiffen the crack, and true compliance will be lower than the one obtained from a notched specimen. Therefore, the crack length will presumably always be underestimated by compliance calibration methods.

Swartz et al (1984) investigated the validity of the compliance calibration method, utilizing a three point bending test setup. All specimens had small starter notches at midspan and they were precracked to a desired crack length using CMOD as a control. It took a couple of cycles for the specimen to achieve the desired crack length according to the compliance calibration curve. Dye would then be applied at the crack section. The test results showed that the compliance method consistently overestimated the actual crack length. The surface cracks revealed by the dye correlated well with the crack depth predicted by a calibrated compliance. For ratios of crack length to beam height greater than 0.26, the difference of average interior crack length and surface crack depth was about 25mm.

21

Swartz et al (1981) also compared the effects of fatigue pre-cracking and static pre-cracking. For the same notched plain concrete beams, one group was pre-cracked by fatigue after one million cycles and the other group was statically pre-cracked to the same crack depth. Under three-point and four-point bending, it was reported that failure strength and associated maximum stress intensity factor of the statically pre-cracked beams are slightly higher than those of pre-cracked by fatigue. It was then concluded that static pre-cracking was acceptable, even for fatigue testing.

Efforts have been made to predict the growth of cracks due to fatigue loading. Balaguru and Shah (1981) proposed a model to simulate the increase of deflection and crack opening for steel RC. The components included in the model were as follows: (a) the cyclic creep of concrete; (b) the reduction of stiffness due to cracking and bond deterioration; (c) reinforcing steel softening. The experimental data was cited from other articles, which was limited to 100,000 or 50,000 cycles. The rebar stress range was between 69 MPa and 276 MPa (10 ksi and 40 ksi). The maximum rebar stress utilized was almost twice the rebar fatigue stress limit. The crack opening was recorded

photographically. It appeared that there were only five data points recorded within 100,000 cycles. The general trend of the model was that crack opening always increased with the number of cycles applied. However, the motivation of using a stress range twice the rebar fatigue stress limit may be questioned. A limit of 100,000 cycles is generally not enough in the fatigue test of reinforced concrete.

22

The finite element method has been widely used in reinforced concrete analysis. There are two different approaches in crack modeling in finite element analysis. One is smeared crack modeling, which is generally better when overall load/deflection behavior is of primary interest. Initially, the concrete is assumed to be isotropic. The reinforced concrete cracks when the stress reaches an assumed failure surface. Instead of literally representing the crack in the concrete FE mesh, the concrete member remains as a continuum. The constitutive equations are then modified to reflect the cracked state.

The other popular method is discrete cracking modeling, utilized when detailed local behavior is investigated, as done early on by Ngo and Scordelis (1967). Based on local stresses in the finite element mesh, some element nodes are separated to model a discrete crack. Since it is costly and tedious, this method is generally only applicable in certain special circumstances.

Darwin (1993) performed a review of finite element analyses on conventional reinforced concrete. The survey results are summarized as follows. (a) Reinforcement. Reinforcements can be modeled in three methods - (1) distributed reinforcement within elements, (2) discrete bar element between element nodes, and (3) uniaxial element embedded in the element. In all cases, reinforcements and concrete are modeled as separate materials. Perfect bonding is always assumed. Fortunately, load-deflection behavior is not sensitive to the bonding unless the failure mode is bond slip, which is not deemed to be a valid design. (b) Concrete under Tension. Tension stiffening and tension softening have improved the numerical stability of simulation. Tension stiffening was

23

first used to account for the residual tensile strength of concrete between cracks. Tension softening uses the concept of fracture mechanics to achieve similar effects. (c) Concrete in Compression. It was found that the overall performance of a model is more related to the details of crack representation and shear retention after cracking, than the details of different concrete constitutive models in compression. (e) Load Increment. It was advised to take small load increments and assure that convergence is achieved at every step.

Perfect bond models, however, are invalid for the purpose of crack analysis. In the vicinity of a crack, there is inevitable bond-slip between rebar and concrete. Efforts have been made to model bonding. Manufacturers of FRP bars are aware of the necessity to model the bond-slip of their bars. Hughes Brothers, Inc. had sponsored a couple of institutions to investigate the phenomenon. A variety of results were obtained, as

different testing methodologies generated different results. This is an indication of the complexity of the issue.

Larralde et al (1993) tested the bonding of FRP bar and concrete. There was a longitudinal helical wrap around bar surface. Since there was little cracking in the concrete after bond failure, it was believed to be an indication of low local bearing stress between the indentations of the FRP bars and the surrounding concrete.

There is generally no fatigue failure within the FRP bars themselves. Therefore, research activities have been directed to the bonding between FRP and concrete under

24

fatigue loading. C. Shield et al (1997) investigated the thermal and mechanical fatigue effects on the bonding between GFRP bars, steel bars and concrete. The specimens were 300mm x 457mm x 1220mm. At both the top and bottom of a specimen, there was one protruding test bar, with one supplementary bar on each side. The embedment lengths were selected to be about ten diameters or more, in order to ensure sufficient development length. Some specimens were cycled under pullout loads between 18KN and 45 KN for 100,000 cycles. Other specimens were stored in an environmental

chamber for three and a half months while temperatures changed between -20oC and 25oC for 20 cycles. Basically, eccentric pull-out tests were conducted. The slips at the loading end and free end were monitored. All of the specimens failed in bond with concrete splitting around the test bars. It was found that GFRP specimens showed no reduction in bond strength after mechanical fatigue, while there was a 13% reduction with steel bars specimens. Thermal fatigue, however, caused more bond degradation in GFRP specimens than in steel bar specimens. In the test setup, the protruding portion of the test bar was loaded, which made it impossible to apply realistic number of cycles, due to the damage to the bar.

C.E. Bakis et al (1998) investigated the effect of cyclic loading on bonding of Glass FRP (GFRP) bars in concrete. The experiment scheme was the RILEM bond beam. The beam section was 100mmx180mm. Some of the bars tested are no longer manufactured. The CP bars in their tests exhibited behavior very similar to the Aslan 100 bars made by Hughes Brothers Inc. The bar diameters were 10.1mm, 12.7mm and 16mm. An embedment length of five diameters was used. The load amplitude was

25

selected to achieve 90%, 50% and 75% of the ultimate bond strength. The bar slip at the free ends was recorded. In the case of CP bars, the residual slip after the first cycle was a significant portion of slip at the end of the 100,000 cycles, ranging from 75% to 25%, depending on the load magnitude. It was found that the residual bond stiffness was actually higher than the initial bond stiffness. The actuator displacement verses load observations also supported that conclusion. The authors suggested that slipping of bars might aid the apparent interlocking with concrete. It was also recognized by the authors that bond failure should not occur in properly designed members with working stress of up to 20% of ultimate strength. This work was contradictory to the finding by C. Shield et al. (1997) that cyclic loading did not enhance the bonding stiffness. The experiment setups were similar to each other for the two investigations, but the load levels were very different.

Cosenza et al. (1997) discussed the bonding behavior between concrete and FRP bars, and a survey of bond-slip models was presented. FRP bars were categorized into straight bars and deformed bars. Straight bars were smooth, grain-covered or sandblasted prismatic rods. Deformed bars were ribbed, indented, twisted or braided. It was stated that the bond was controlled by the factors including chemical bond, friction due to FRP surface roughness, mechanical interlock of FRP bars against concrete, normal pressure between FRP bars and concrete. An effect of bar size has been observed, with the average bond resistance decreasing as bar size increased. A top bar effect also exists for FRP. Longer embedment length resulted in lower average bond resistance. Among environmental conditions, bond strength was not closely related to temperature changes;

26

chemical conditions, however, such as high alkalinity were shown to be detrimental to bonding. Popular bond-slip models are the Malvars model, BPE model, modified BPE model and CMR model. All of these models use exponential functions to model the first branch of increasing bond stress and slip. The softening branch was modeled linearly, for convenience. The authors stated that modified BPM model presented the best agreement with the available experiment results.

A. Katz (2001) tested five different types of FRP bars. Each FRP bar was embedded in a concrete block and 450,000 cycles of cyclic loads were applied. Between each 150,000 cycles, the specimens were immersed in water of 60oC and 20oC to simulate a deterioration process. At the end of the fatigue tests, a pullout test was conducted for each specimen. Three mechanisms of failure observed were abrasion of rod surface, delamination of outer layer of resin, and abrasion of cement particles entrapped between rod and concrete. It was concluded that helical wrapping of FRP bars did not increase bond resistance under cyclic loading. A sand covered bar surface did improve bonding; such bars were able to maintain maximum loading for a relatively long slip.

Flexural response of FRP reinforced concrete were reported by Benmokrane (1996) and other investigators. The general consensus is that at small load, the crack pattern in FRP concrete is similar to that of steel reinforced concrete. As the load increases, however, there are more cracks with larger crack openings in FRP concrete than in traditional steel reinforced concrete, for comparable reinforcement ratios. This

27

behavior is expected, since FRP has a much lower modulus of elasticity, compared with traditional steel reinforcement. The moment /curvature diagrams of lightly reinforced FRP beams are clearly bi-linear, with the bend point at the crack initiation moment level.

GFRP reinforced concrete beams were analyzed by Vijay and GangaRao (2001); different modes of failure were compared. The compression controlled failure mode presented not only higher flexural strength, but also a more ductile failure than the tension controlled failure mode. This result was consistent with ACI 440.1R-01

suggested design criteria. A parameter DF was defined as the ratio of energy absorption at ultimate strength to that at a limiting curvature value. To satisfy both the serviceability deflection limit of L/180 and crack opening limit of 0.016 inches, the curvature limit was set to be 0.005/d. The parameter DF then became a unified indicator, covering both serviceability and strength. The tensile strength of concrete is typically assumed to be 7.5 f c' , with an assumed elastic modulus of 57000 f c' (using U.S. units with stress units in psi). The tensile strain at cracking is thus assumed to

be cr = 7.5 f c' 57000 f c' = 0.0013 .

The curvature at first cracking cr is

approximately 2 cr / h = 0.0026 / h , for a symmetric section. Vijay and GangaRao thus have used twice the curvature at first cracking as the limiting curvature in their design criterion.

Bridge decks of traditional steel reinforced concrete have been analyzed and tested by many researchers, including Graddy et al.(1995). With the general purpose FE program SAP as their primary analysis tool, they performed a sequence of linear finite

28

element analysis, utilizing a smeared crack representation. A cracking stress of 0.1fc was used for plain concrete, with Kupfer's (1969) criterion. Cracks were only deemed possible in the directions parallel to the transverse and longitudinal reinforcement, i.e., the model failed to simulate nonorthogonal cracks. assigned for each round of analysis. New material parameters were

The element utilized was an eight node

isoparametric solid element, of the same size for the entire model, with edges parallel to the edges of model. Comparing analytical results with available experimental data, the study indicated that load-deflection was accurately simulated in the analysis, while the predicted stresses in the reinforcements were very different from those observed experimentally. The work was limited to the ultimate strength studies of RC slabs, and the serviceability of these slabs was not investigated.

Many researchers including Graddy et al.(1995) and others have noticed the effect of arching action in traditional steel reinforced concrete. Before a concrete slab cracks, the dominant resistance is flexure. After the concrete cracks, a dome architecture exists underneath the concentrated loads, if the cracked concrete is excluded. In-plane, or membrane stress, then becomes more significant. The results of theoretical analysis and experiments have shown that the arching action contributes to the slab strength. Arching action for multiple wheel loads is uncertain, however, especially in the case of FRP reinforced concrete slabs. Arching effect at service load levels for FRP slab has not been investigated.

29

Canadian investigators have been active in the research of fiber reinforced concrete and steel-free bridge deck system. The transverse reinforcements are mainly external steel straps or FRP bars. B. Bakht et al (2000) reviewed different types of straps. They included fully studded straps, partially studded straps, cruciform straps, FRP bars and diaphragms. Three models of steel free slabs with different straps were tested to failure under monotonic loading. The mode of failure was mostly punching shear failure as expected, but at a much larger load. An additional specimen was tested under 1000 cycles of pulsating load between 0 and 88 KN (20 Kips) prior to the static testing. The results of the latter static testing indicated that the forces in straps increased, due to shakedown in the slab. The authors concluded that actual failure loads of the steel-free deck slabs are more than 10 times larger than the theoretical failure load attributable to bending alone.

Similar research was conducted by Salem et al (2002). A finite element model was developed for a steel free concrete deck. The lateral reinforcement was a cruciform strap. The concrete was fiber reinforced concrete, so as to control cracking due to creep and shrinkage. The results showed that the load at slab failure was only increased by 11% for a two girder model and 15% for a three girder model, when the inertia of girders was increased by 150%. The position and location of the lateral straps were also

analyzed. The ultimate load of the slab was insensitive to the strap position. For practical purposes, it was recommended to weld the straps to the top flange of girders.

30

Yost (2002) tested the performance of concrete slabs reinforced by FRP grids. The product is commercially known as NEFMAC, and is composed of continuous high strength reinforcing fibers, impregnated within a vinyl ester resin. A two dimensional grid sheet was formed with redundant overlaying. The test was conducted under monotonic loading with AASHTO HS25 truck load. The ultimate load was five times as much as the HS25 criterion. The field testing of a bridge slab had shown that the strains and deflections were well within the design limits.

The effects of pulsating and moving loads on traditionally reinforced concrete slabs were studied by Perdikaris et al (1988, 1989). The research covered both the AASHTO orthotropic reinforced slabs and the Ontario isotropic reinforced slab. In the prototype, the three beams were space at 2.13m (7 ft). The orthotropic reinforcement pattern consisted of a top and bottom layer of transverse and longitudinal steel reinforcing bars 19M (#6) spaced at 188mm and 376mm respectively. In the isotropic reinforcement pattern, the spacing in both directions was 437mm. In either case, the spacing was fairly large. The restricting boundary conditions were considered in the research. Models of 1/6.6 and 1/3 scale were tested. The maximum fatigue load was 60% of the static ultimate strength, which was fairly high. The results showed that fatigue life of slabs with isotropic reinforcement is twenty times that with orthotropic reinforcement. The factors of safety at static ultimate failure are 14 and 23, however, for those of isotropic and orthotropic reinforcements, respectively.

31

Bridge slabs are constantly under traffic load. Due to serviceability requirements, such as crack opening and lateral load distribution, ultimate strength is usually not crucial in the slab design. However, the AASHTO design methodology is still presented from the perspective of strength design. The design moment in the load factor design (LFD) methodology was assumed to be (S+2)P/32 per foot of slab width, where S is the effective span length of slab in feet and P is the design wheel load. The formula is in U.S. units. In the AASHTO load and resistance factor design (LRFD), an equivalent width of bridge slab was defined for strength design in AASHTO Table 4.6.2.1.3-1. In the case of a concrete slab over multiple girders, the width is taken as 660+0.55S for positive moment and 1220+0.25S for negative moment, where the girder spacing is S. The methodology is believed to simplify the bridge deck design process.

The Ontario Highway Bridge Design Code has recognized the in-plane or membrane forces in typical bridge slabs. The slab design was reduced to a prescription of isotropic reinforcement. The reinforcement pattern is orthogonal in the slab. A minimum reinforcement ratio of 0.003 is required in both directions, top and bottom. The restrictions of the empirical design are as follows. a.) The span length of a slab is less than 3.6m (12 ft). b.) The ratio of span to thickness does not exceed 15. c.) The slab thickness is 225mm minimum and the spacing of the bars is 300mm maximum. d.) Intermediate diaphragms will not be spaced at more than 8m. requirements are then assumed to be met automatically. The crack control

32

A similar empirical design methodology is available in AASHTO (2000). Reinforcement is required at both directions of each face. The minimum amount of reinforcement is 0.570 mm2/mm of steel for each bottom layer and 0.380 mm2/mm of steel for each top layer. The maximum spacing of reinforcement is 450 mm.

The most common type of bridge is a concrete deck, supported on multiple girders. Except in the case of large horizontal curvature, girders are usually analyzed and designed individually. In other words, 1-D finite element analysis is common practice in the bridge design consulting industry. Therefore, the AASHTO design codes have

traditionally provided lateral distribution factors which account for the maximum possible portion of wheel load (half of axial load) acting on one girder. In the AASHTO LFD design codes, simple formulas of load distribution factors are listed. For girder spacing S less than 3.6m (12 feet), the distribution factor (DF) is S/5.5. In the current LRFD codes, the formulas for DF are as follows. DF = 0.06 + ( S 0.4 S 0.3 K g 0.1 ) ( ) ( ) 14 L 12 Lt s3 S 0.6 S 0.2 K g 0.1 ) ( ) ( ) 9.5 L 12 Lt s3 (one design lane loaded)
(1-5)

DF = 0.075 + (

(two or more design lane loaded)

(1-6)

where S is the girder spacing; L is the bridge span, Kg is longitudinal stiffness parameter and ts is slab thickness. Many researchers have been involved in the evaluation of distribution factors. The majority of the work has been finite element analysis of bridge structures of steel reinforced concrete slabs on multiple girders. Mabsout el al (1997) reviewed finite element analysis of bridges and analyzed a bridge with a span of 17m (56 ft). Concrete slabs may be modeled with shell elements or isoparametric continuum

33

elements. Girders may be modeled as 3-D beam elements with rigid links to the slab. Sometimes, the web may be modeled with shell elements and flanges may be modeled with beam elements, or the entire girder may be modeled with shell elements. It was found that different models produced distribution factors similar to NCHRP 12-26 (1987), but all were less than AASHTO (1996). The analysis results showed that the distribution factor decreased, as the bridge span became larger.

In summary, the advantages of FRP make it a potentially better choice in applications such as bridge deck slabs. The performance of FRP RC under monotonic loading has been understood fairly well. ACI 440.1R-01 proposes to design for a

strength failure mode of concrete crushing, to achieve better ductility. Some other criteria which are serviceability oriented have been reported.

The serviceability of FRP RC, particularly in fatigue environments, deserves to be further investigated before engineers can be expected to be confident with this fairly new material. The predicted maximum crack opening of FRP RC has been converted from conventional RC, although the bond properties are different for steel and FRP. The Paris law appears to be applicable in concrete, with a size effect being detected. FRP itself possesses excellent fatigue properties; the bond durability under cyclic loads, however, has not been thoroughly investigated. There have been varying results, mostly based on pullout tests, on residual bond strength following limited cyclic loads.

34

Verification with a different experimental methodology is needed. The current bridge design code is conservative in terms of load distribution. Although the current design methodology is strength oriented, serviceability is often the critical factor in bridge deck slab design. The finite element method has been successfully used for a long time in bridge structural analysis; analysis results are typically based on uncracked concrete slab properties, which is generally reasonable for steel RC.

Crack growth in FRP reinforced concrete is yet fully understood, particularly in fatigue environments. Investigation of FRP RC fatigue performance is crucial in

applications such as bridge slabs. In this study, experimental results on fatigue testing of FRP RC will be presented. Subsequently, the crack opening displacement and crack growth will be modeled utilizing the finite element method and fatigue/fracture theory, respectively. A finite element model will be developed to simulation the crack opening of the test specimens. A fatigue model will be created to simulate the observed crack growth under cyclic loading. An empirical equation for final crack opening will be proposed. A sensitivity analysis on the crack growth model will also be conducted to evaluate the effects, the uncertainty and the randomness of different parameters. The finite element model will then be extended to the analysis and crack opening estimation of realistic FRP reinforced concrete bridge deck slabs under actual AASHTO wheel loads. Finally, the overall performance of an FRP RC slab on a single span bridge of multiple girders will be analyzed. Under the condition of a cracked slab, the lateral load distribution factor will be discussed. Other implications on the serviceability provisions in ACI 440.1R-01 will be discussed.

35

Chapter 2

Experimental Analysis of FRP Reinforced Concrete under Fatigue Load

Motivation for the Testing Program

Due to its high corrosion resistance, FRP is set to be a promising alternative to steel reinforcement in bridge decks. Typically a major concern in an FRP bridge slab is its serviceability, rather than its strength. Crack opening is one of the important

indicators of serviceability. Crack opening and its growth in FRP RC are related to the fatigue characteristics of FRP bar, concrete, and their interface. The bond-slip and crack growth mechanisms at different rebar spacing have not yet been fully investigated.

The behavior of FRP reinforced concrete under fatigue loading has been investigated thus far by simple pullout tests, or by RILEM beam bond tests, following an interval of cyclic loading. There are two shortcomings with these approaches. One is that the testing condition is not the actual working condition of rebar in an environment such as a bridge deck. A small bond length is normally used in a RILEM beam or a concrete pull-out block. Conclusions drawn under such conditions may not always be applicable to typical in service conditions. The second issue is that such tests are sensitive to specimen imperfections. With portions of bar exposed, the bar is susceptible to local damage due to unintentional stress concentrations and eccentricities which may not be representative of in-service conditions; such variations can be especially critical in fatigue testing.

36

The proposed experiment focused on fatigue-induced crack growth in FRP RC under service-level cyclic loading, in specimens more representative of in-service applications. Specimens were actual beams reinforced with FRP bars. Beams of

different widths were used to simulate bridge slabs of different bar spacing/reinforcement ratios. Traditionally, the minimum thickness of a bridge slab is 215 mm (8.5 inches). Concrete bridge slabs are typically designed with sufficient depth such that no shear reinforcement is needed, and so that the expected load distribution among bridge girders is achieved. Therefore, the performance of FRP reinforced concrete in the flexural response modes is of primary interest to bridge deck designers.

Description of the Testing Program

FRP beams of identical depths and spans, but with four different widths were fabricated. The concrete was composed of type III cement, water, fine aggregate and coarse aggregates with weight proportions of 1.0/ 0.5/ 2.0/ 2.83. The nominal

compressive strength target was 34.5MPa (5000 psi). The compressive strength from a cylinder test was 27.9 MPa (4045 psi). The tensile strength from a split-cylinder test was 4.9 MPa (715 psi).

37

Figure 2.1 Aslan 100 GFRP by Hughes Brothers

The first set of FRP bars tested, which are reported herein, were Aslan 100 GFRP made by Hughes Brothers, Inc. (see Figure 2.1). As shown above, the bars are sand coated with a helical wrap along the length. The reported tensile strength is 655 MPa (95 ksi) for No. 16 (#5) bars. The reported modulus of elasticity is 40.8 GPa (5.92E6 psi). To simulate a typical bridge slab section, beams were all 1830 mm (6 feet) long and 215mm (8.5 inches) thick. The beam widths were 76 mm, 102 mm, 127 mm and 152 mm (3, 4, 5 and 6 inches) which represent typical bar spacing in bridge decks. For identification purposes, they are categorized as group H and they are labeled as C3x8.5H5, C4x8.5H5, C5x8.5H5, C6x8.5H5, respectively. The first letter C stands for the constant amplitude; the beam size in U.S. units follows; H shows the manufacture of the bars as Hughes Brothers, Inc.; the last number is the size, #5, of the FRP bar. Within each beam, there was one No. 16 FRP bar (#5 diameter 5/8 inches) at the bottom of each beam (tensile region) with 25 mm (1 inch) cover to the bar surface.

38

Figure 2.2 Isorod by Pultrall

The second set of FRP bars tested were Isorod GFRP made by Pultrall, ADS Composites Group (see Figure 2.2). The bars are also sand coated, without a helical wrap along the length. The tensile strength is 674 MPa (98.9 ksi) for #5 bars. The modulus of elasticity is 42 GPa (6.1x106 psi). Similarly, test specimens were all 1830 mm (6 feet) long and 215mm (8.5 inches) thick. The beam widths were 76 mm, 102 mm, 127 mm and 152 mm (3, 4, 5 and 6 inches) which represent typical bar spacing in bridge decks. For identification purposes, they are categorized as group P and they are labeled as C3x8.5P5, C4x8.5P5, C5x8.5P5, C6x8.5P5, respectively.

One extra specimen, C5x8.5P5OL, of section 127mmx215mm with bars of Isorod was made to investigate the effect of overload pre-cracking. One more specimen,

C5x8.5H5M, of section 127mmx215mm with bars of Aslan 100 was singled out with cracks adjacent to each other, to investigate the effect of multiple cracks. One specimen, C5x8.5S5, of section 127mmx215mm was made of 16M (#5) steel rebar, for comparison purposes.

39

Specimen C3x8.5H5 C4x8.5H5 C5x8.5H5 C6x8.5H5 C5x8.5H5M C3x8.5P5 C4x8.5P5 C5x8.5P5 C6x8.5P5 C5x8.5P5OL C5x8.5S5

Width (mm) 76 102 127 152 127 76 102 127 152 75 127

Height (mm) 215 215 215 215 215 215 215 215 215 215 215

Reinforcement 16M (Aslan 100) 16M (Aslan 100) 16M (Aslan 100) 16M (Aslan 100) 16M (Aslan 100) 16M (Isorod) 16M (Isorod) 16M (Isorod) 16M (Isorod) 16M (Isorod) Steel

Test Sequence 2 3 1 4 5 6 7 10 9 8 11

Table 2.1 Specimen Descriptions

610mm

610mm

610mm

215mm 25mm FRP

Figure 2.3 Specimen Section Details and Loading Condition

40

Figure 2.4 Cyclic Load Test Setup

The specimens were all under four point bending (see Figure 2.3 and 2.4). The beam was loaded symmetrically with two loads at the third points. The cracks within the pure bending region were monitored. The maximum cyclic service load was determined based on the creep rupture stress limit of 0.20ffu for FRP bars, in accordance with ACI 440.1R-01, resulting in a cyclic rebar stress level of 645 MPa (~20 ksi). The minimum and maximum loads were 2225 N (500 lb) and 15600 N (3500 lb) respectively. The resulting moments are greater than the theoretical cracking moments. Based on nominal kb value of 1.2, the predicted crack openings are 0.68 mm, 0.75 mm, 0.80 mm and 0.84 mm for beam widths of 76 mm, 102 mm, 127 mm and 152 mm, respectively.

According to ACI 440.1R-01, the performance of FRP is dependent on the testing frequency. Endurance limit was found to be inversely proportional to loading frequency

41

in carbon FRP. Higher cyclic loading frequencies in the 0.5 to 8 Hz range corresponded to higher bar temperatures due to sliding friction. For a bridge slab under traffic load, the stress of a rebar reaches maximum when a truck axle load is applied on the top of the slab at the same location. For a truck with axle spacing of 3.6m (12 feet) at 65 miles per hour, the frequency of passing axles may be as high as 7.94. However, for a bridge of 10,000 ADTT (average daily truck traffic), the truck load is applied at a frequency of 0.23 Hz. So, the overall frequency is 1.8 Hz, which is the product of 7.94 and 0.23. Therefore, the frequency at which load was cycled was at 2 Hz in the tests. The percentage of overload was decided based on traditional AASHTO load factor design. The overload was defined to have the value of factor at 1 instead of 1.3 in the factored load. Therefore, for specimen C5x8.5H5 and C6x8.5H5, the effect of a modest (30% to 40%) overload was also investigated.

Specimen

Knife Edge Grouted to Specimen Crack Opening Displacement Gage MTS System

Figure 2.5 Sketch of Data Acquisition System

42

Static pre-cracking was used. The loading was stopped as soon as cracks became visible for all specimens, except in the case of the overload pre-cracking investigation. MTS clip-on crack opening displacement gages 632.02B-20 and 632.02C-20 were then installed on the cracks which had been initiated as shown above. The maximum arm displacements of the instruments are +2.540 mm to -1.270 mm (+0.1000 in to -0.050 in) and +3 mm to -1 mm (+0.118 in to -0.039 in)), respectively. Two DCDTs were also fastened on each side of the specimen in the mid-span to measure the relative beam deflection, within the pure bending region, for average curvature estimation. All eleven specimens were tested under the same initial cyclic load amplitude. The crack mouth opening displacement (CMOD) was recorded under a ramp load and the first 20 cycles of cyclic load at the beginning of each test interval, in order to track the evolution of crack development with increasing load cycle counts.

Experimental Results

(1) Group H - Aslan 100 GFRP Rebar by Hughes Brothers, Inc.

The first specimen tested was C5x8.5H5. Two cracks appeared within the pure bending region after static pre-cracking and two more cracks were observed immediately after the test started. The approximate crack spacing was 190mm (7.5). After the first test interval of 5,000 cycles, the crack lengths became visually constant. After more cycles were applied, there was no sign of distress with the specimen, and all cracks were stable. All crack tips stopped at approximately 38mm (1.5 inches) below the top of beam, which was near the theoretical neutral axis. The specimen did not appear to have

43

any distress at the end of testing of one million cycles. The crack length was virtually the same. There was no concrete spalling near the rebar at the bottom of specimen. It was also found that there was no scaling in the specimen concrete surfaces were sound with no loss of surface mortar and aggregates. To investigate the effect of overload, Pmax was increased to 22,300 N (5.0 kips), corresponding to a rebar stress level of 25 ksi. After 10,000 cycles of this overload, the specimen was still in good condition.

Figure 2.6 Specimen C5 x 8.5 H5

The second specimen tested was C3x8.5H5. Three cracks appeared at static precracking and three more were observed at 20,000 cycles. The crack spacing was between 130mm (4.5 inches) to 165mm (6.5 inches). The tips of the cracks stopped at

approximately 50mm (2 inches) below the top of beam. Due to the larger bearing stress at both supports, the concrete at the bearing locations started crumbling near the end of 2 million cycles of testing. There was no sign of concrete distress elsewhere in the

specimen. No overload was applied due to the degraded condition of the concrete in the vicinity of the bearings.

44

Figure 2.7 Specimen C3 x 8.5 H5

The behavior of specimen C4x8.5H5 was similar. Three initial cracks were generated at static pre-cracking. Two more cracks appeared at 6000 cycles. The average crack spacing was between 130mm ( 4.5 inches) and 180mm ( 7 inches). The tips of the cracks stopped at approximately 45mm ( 1.75 inches) below the top of beam, up to 1.8 million cycles. To investigate the effect of overload, Pmax was again increased to 22,300 N (5.0 kips) for 15,000 cycles. No addition distress was found in the specimen.

Figure 2.8 Specimen C4 x 8.5 H5

45

The behavior of specimen C6x8.5H5 was somewhat different. Only one crack was generated at static pre-cracking. Extra load was added after the appearance of the first crack but no additional cracks appeared. During the subsequent fatigue testing, no new cracks appeared up to 140,000 cycles, at which point the CMOD gage debonded. (The single crack had ceased to grow in length, however, prior to 10,000 cycles.)

The Pmax was raised at that point to 20000 N (4.5 kips) to explore the effect of overload. A new crack appeared 700 cycles later. The newly formed crack was

instrumented, and Pmax was again lowered to its initial value of 15,600 N ( 3.5 kips). After an additional 35,000 cycles of fatigue load at Pmax of 15,600 N, the primary crack did not show any sign of further growth induced by the 700 cycles of overload. Therefore, Pmax was raised back to 20,000 N and 40,000 additional cycles were applied, with both the primary crack and secondary crack remaining stable.

Figure 2.9 Specimen C6 x 8.5 H5

46

To further investigate the overload effect, Pmax was finally increased to 22,300 N (5.0 kips). A third crack was found around 400 cycles; a total of 40,000 cycles were applied at this load level, with the second and third cracks monitored. All cracks became stable and no addition signs of distress were noted.

(2) Group P - Isorod GFRP made by Pultrall, ADS Composites Group

The first specimen tested was C3x8.5P5. Three cracks appeared at static precracking within the pure bending region and one outside the pure bending region. The average spacing was 150mm (6 inches). After the first run of 3,000 cycles, the crack lengths became visually constant. After more cycles were applied, there was no sign of distress such as spalling and scaling with the specimen, and all cracks were stable. No new cracks were found in the specimen. The controlling system crashed at a load cycle count of 30,000. All crack tips stopped at approximately 45mm (1.75 inches) below the top of beam. The specimen did not appear to have any distress at the end of 270,000 testing cycles. To investigate the effect of overload, Pmax of 20,000 N (4.5 kips) was applied. After 10,000 cycles of overload, the specimen was still in good condition.

47

Figure 2.10 Specimen C3 x 8.5 P5

The second specimen was C4x8.5P5. Within the pure bending region, only one crack appeared at static pre-cracking and one more was observed at 400 cycles. A clip gage was immediately installed for the new crack. An additional crack appeared at load cycle 1000. The average spacing was 200mm (8.5 inches). The tips of the initial cracks stopped at approximately 38mm (1.5 inches) below the top of beam after the application of 900,000 load cycles. Excessive overload was tested at Pmax of 29,000 N (6.5k), which resulted in 276 MPa (40 ksi) of rebar stress. This stress level was equivalent to the data cited by Balaguru and Shah (1981) in their model to simulate the increase of deflection and crack opening for steel RC. The general trend of their model was that crack opening always increased with the number of cycles applied. After 200 cycles of overload, existing cracks started branching and a new crack appeared. After 3000 cycles of

overload, the concrete cover started falling off, as debonding became more pronounced; it was determined that the specimen had reached fatigue failure at that point.

48

Figure 2.11 Specimen C4 x 8.5 P5

Specimen C5x8.5P5 behaved somewhat differently. One initial crack of 130mm long was generated at static pre-cracking. One new crack appeared at 110 cycles. At around 900 cycles, two new cracks appeared, with one crack of initial surface length 120mm (4.75 in), between the first two cracks at the midspan region. The average crack spacing was 115mm (4.5 inches) within the pure bending region. The tip of the newer crack at midspan was dormant for about 100,000 cycles, and then began growing. (Unfortunately, no more gages were available to acquire the crack opening evolution of this crack.) The tips of all cracks stopped at approximately 50mm (2 inches) below the top of the beam at 1.25 million cycles. To investigate the effect of overload, Pmax of 22,300 N (5.0 kips) was applied for 10,000 cycles. By the end of the test, the concrete at the left bearing started crumbling.

49

Figure 2.12 Specimen C5 x 8.5 P5

The specimen C6x8.5P5 behaved similarly. Only one crack was generated at static pre-cracking, as expected. No extra load was initially added, to avoid any plastic hardening of the concrete-rebar interface bonding. During the subsequent fatigue testing, no new cracks appeared up to 1,300,000 cycles. The Pmax of the cyclic load was then raised to 22300 N (5.0 kips) to explore the effect of overload. One new crack appeared within 400 cycles of overload. After 50,000 cycles of overload, however, there was no indication of severe distress. Subsequently, Pmax was raised to 29,000 N (6.5 kips). The two existing cracks then started branching. After 155,000 cycles of this overload were applied, the specimen was still in good shape. All cracks became stable and no addition signs of distress were found.

50

Figure 2.13 Specimen C6 x 8.5 P5

(3) Overload Pre-cracking

In all tests to this point, cracks had been generated with minimum possible static loading, which is equivalent to fatigue pre-cracking. Experiments were also conducted to investigate the case of overload pre-cracking. Additional static overload was applied after cracks had appeared, followed by cyclic load at service level. For specimen

C5x8.5P5OL, there was no further growth of crack length during the course of fatigue testing. For specimen C6x8.5H5, crack lengths did continue developing during fatigue testing..

Figure 2.14 Specimen C5 x 8.5 P5OL

51

(4) Conventional steel RC

A similar test was conducted for a specimen made with conventional steel reinforcing. Static pre-cracking was used, and five cracks appeared, with two very close to each other. The initial crack length was between 100mm (4 inches) and 120mm (4.75 inches). The crack spacing was ranging between 140mm (5.5 inches) to 180mm (7 inches). As cyclic load testing started, there was no visible growth of the cracks. No new crack was generated during the test. At the end of 1,000,000 cycles, there was no sign of distress within the specimen.

To further investigate the overload effect, Pmax was first increased to 22,300 N (5.0 kips). The specimen was still in good shape after 150,000 cycles. Then, Pmax was then increased to 29,000 N ( 6.5 kips) which represented 200% of working stress; the specimen appeared to be intact after 30,000 cycles of this load level.

Figure 2.15 Specimen C5 x 8.5 S5

52

For all specimens, the attempts to monitor average curvature through the measurement of relative displacements within the test section, failed to produce consistently usable results, particularly for large cycle counts. First, the magnitude of deflection at the mid-span, relative to the line of the two 1/3 span loading points, was very small in magnitude (only on the order of a few thousandths of an inch), resulting in a low resolution for the measured DCDT data to begin with. It was also inevitable for specimens to shift positions over time under the dynamic load, even though a minimum non-zero load was maintained, and to exhibit some secondary torsional movement, due primarily to minor imperfections in the specimen and supports, all of which contributed to measurement difficulties. It was decided finally to utilize only the more reliable crack gauge data in the subsequent analyses.

(5) Crack Profile Characterization

The crack profile may be investigated in the methods of laser holographic interferometry, acoustic emission and dye penetration. For some specimens in group P, the crack length profile was investigated using dye penetration. A notch was made at the top of a crack for a specimen. The specimen was then loaded in the three point bending mode, so as to open the crack. As the cracks opened up, rubber sheets were clamped to each side of the specimen around the crack (see Figure 2.14). Black ink was injected into the notch, penetrating the crack until reaching the tip the crack. After about two hours, the reinforcement was cut off and the crack examined; the images of cross sections of C4x8.5P5, C5x8.5P5 and C6x8.5P5 are illustrated in Figure 2.15.

53

Figure 2.16 Injectiing Dye into Cracks

Figure 2.17 Typical Crack Profiles

54

Quantitative Discussion

The balanced reinforcement ratio is 0.0048, for an FRP tensile strength of 655MPa (95 ksi) and a concrete strength of 27.6 MPa (4000 psi). The reinforcement ratios tested were 0.013, 0.010, 0.008 and 0.007 for specimens C3x8.5H5, C4x8.5H5, C5x8.5H5 and C6x8.5H5, respectively. Specimen C6x8.5H5, which displayed a

somewhat different behavior than the other specimens, had the lowest reinforcement ratio, although it was still slightly over-reinforced. As mentioned earlier, the predicted service load crack openings, based on ACI 440.1R-01 criteria, were between 0.68 mm and 0.84 mm for all four specimens, at the suggested nominal kb value of 1.2. The experimental results show that the service load crack openings, measured immediately after static pre-cracking, were 0.15 mm, 0.16 mm and 0.17 mm for group H specimens C3x8.5H5, C4x8.5H5 and C5x8.5H5, respectively. These experimental observations were only about 25% of the predicted value. The opening of the single crack in specimen C6x8.5H5 was 0.26mm, which was still less than 30% of the predicted value. In group P, the service load crack openings, measured immediately after static pre-cracking, were 0.16 mm, 0.17 mm, 0.19 and 0.22 mm for group P specimens C3x8.5P5, C4x8.5P5, C5x8.5P5 and C6x8.5P5, respectively. Based on these limited tests, it appears that the modified Gergely-Lutz equation may be overly conservative in predicting actual static service load crack openings, at least for the bars tested in this investigation. According to the limited test results, a kb value of 0.4 may be more realistic for initial static crack opening prediction. Another finding was that there was hardly any difference between group H and group P. The reason is that initial static CMOD at working stress level is

55

more related to the modulus of elasticity of FRP bars than the surface bonding. The elastic properties of two groups of FRP bars were approximately the same.

The growth of crack opening versus number of cycles may be represented by a sum of an elastic CMOD and a plastic CMOD. The elastic CMOD is calculated as the difference of CMOD at maximum and minimum load, which disappears after unloading. The residual CMOD at minimum load is the plastic CMOD, which does not disappear after the removal of loading (see Figure 2.18), and tends to show a greater increase with the number of applied load cycles than elastic CMOD does.

Load

CMOD Plastic CMOD Elastic CMOD

Figure 2.18: Definitions of Elastic and Plastic CMOD

Figures 2.19, 2.20, 2.21 and 2.22 display the evolution of elastic CMOD and plastic CMOD for each specimen in group H and P, with increasing load cycle counts,

56

respectively. As can be seen in Figure 2.19 and 2.20, elastic CMOD tended to grow slowly at first, but then stabilize to a nearly constant value after a few thousand load cycles. Plastic CMOD tended to grow slowly throughout cyclic loading, but at a

decreasing rate, with increasing load cycles counts. By the end of the tests of one million cycles, the plastic CMODs were about one quarter of the elastic CMODs. Based on the experimental results, it would appear to be conservative to use one and half of the initial CMOD as an estimate of maximum total crack opening under cyclic loads.

0.400
C3 x8 .5H5

C4 x8 .5H5

C5x8 .5H5

0.300
Elastic CMOD (mm)

C6 x8 .5H5

0.200

0.100

0.000
1 10 100 1000 10000 100000 1000000 10000000 Number of Cycles N

Figure 2.19

Elastic CMOD under Ramp Load vs Number of Cycles for Group H Specimens (Pmin=2.2 KN Pmax=15.6 KN)

57

0.2

C3 x8 .5H5

C4 x8 .5H5

C5x8 .5H5

C6 x8 .5H5

0.15
CMOD (mm)

0.1

0.05

0 1 10 100 10000 100000 100000 1E+07 0 Numbe r of Cycle s N 1000

Figure 2.20

Plastic CMOD under Ramp Load vs Number of Cycles for Group H Specimens (Pmin=2.2 KN Pmax=15.6 KN)

0.300

C 3x8.5P 5

C 4x8.5P 5

C 5x8.5P 5

0.250
Elastic CMOD (mm)

C 6x8.5P 5

0.200

0.150

0.100 1.E+00

1.E+01

1.E+02

1.E+03

1.E+04

1.E+05

1.E+06

Number of Cycles N

Figure 2.21

Elastic CMOD under Ramp Load vs Number of Cycles for Group P Specimens (Pmin=2.2 KN Pmax=15.6 KN)

58

0.2
C3 x8 .5P5
C4 x8 .5P5

0.15
CMOD (mm)

C 5x8.5P 5
C6 x8 .5P5

0.1

0.05

0 1 10 100 1000 10000 100000 1000000


Numbe r of C ycle s N

Figure 2.22

Plastic CMOD under Ramp Load vs Number of Cycles for Group P Specimens (Pmin=2.2 KN Pmax=15.6 KN)

Based on the above definitions and experiment results, the crack growth of FRP RC is herein further divided into two stages. The first stage is crack development, during which the length and opening of a crack both increase with the number of cycles under the applied cyclic loading. With lower reinforcement ratios (wider bar spacings), the crack development stage appears to continue to a higher cycle count. Both elastic and plastic CMOD tend to grow with the number of cycles, but the elastic CMOD grows much more markedly with increasing load cycles, during this period of crack development. Specimens C3x8.5H5, C3x8.5P5, C4x8.5H5, C4x8.5P5, C5x8.5H5 and C5x8.5P5 approached the end of crack development between 3000 and 6000 cycles. It took specimens C6x8.5H5 and C6x8.5P5 about 10,000 cycles to exhibit fully developed cracks.

59

Once the peak elastic CMOD is reached, crack growth reaches the second stage, or crack stabilization. The second stage is characterized by nearly constant crack length, nearly constant elastic CMOD, and a slow accumulation of plastic CMOD. The fact that there was only one crack in the specimen C6x8.5H5 suggests that there is more potential for cracks exhibiting larger CMOD for beams of lower reinforcement ratio (wider bar spacings). The general trend of plastic CMOD, which was calculated as the residual CMOD at zero loading, is that it increases with the number of cycles applied, although at a decreasing rate. Due to the finite resolution of the measurement technique, plastic CMOD was subject to a greater relative measurement error.

Figures 2.23 through 2.26 show the evolutionary history of the total vertical load versus CMOD hysteresis of group H under cyclic loading. The area encompassed by the hysteresis loop for any particular load cycle is related to the energy loss by the specimen during that load cycle. If the area becomes larger, it implies that more damage was induced within that load cycle and vice versa. The slope of the hysteresis loop is related to the integrity of the bonding mechanism between concrete and the FRP bar. For beam C3x8.5H5 (see Figure 2.24), the slope of the hysteresis loops decreased slightly as the number of cycles increased. For beam C4x8.5H5 (see Figure 2.23), all hysteresis loops were nearly parallel to each other (constant stiffness). The area contained within the hysteresis loops decreased slightly with increasing load cycles. For beam C5x8.5H5 (see Figure 2.25), the hysteresis slopes again decreased slightly as the number of cycles increased. The area of hysteresis also decreases slightly with increasing load cycles for beam C5x8.5H5. For beam C6x8.5H5 (see Figure 2.26), a larger CMOD was recorded

60

for the single crack. The hysteresis slopes again decreased slightly with increasing load cycles.

The characteristics of hysteresis of group P in Figure 2.27, 2.28, 2.29, 2.30 and 2.31 were similar to those of group H except for specimen C5x8.5P5. This difference is believed to be attributable to secondary cracks, which is discussed later. There are two sets of hysteresis for the two cracks in specimen C5x8.5P5. One crack was generated by static precracking and the other was initiated during cyclic loading. The characteristics of the two cracks were almost identical, showing no difference between cracks initiated by static precracking and by fatigue precracking.

16
1

14 12
Load (KN)

2000 2 3 0 ,0 0 0 1,8 4 1,50 0

10 8 6 4 2 0 0 0.05 0.1 0.15 0.2 0.25


Elastic CMO D (mm)

Figure 2.23 Hysteresis of Beam C4x8.5H5 Under Cyclic Load

61

16
1

14 12
Load (KN)

2 2 0 ,0 0 0 9 8 8 ,0 0 0 2 ,0 0 0 ,0 0 0

10 8 6 4 2 0 0 0.05 0.1
Elastic CMO D (mm)

0.15

0.2

Figure 2.24 Hysteresis of Beam C3x8.5H5 Under Cyclic Load

14
1

12 10
Load (KN)

10 ,0 0 0 2 8 0 ,0 0 0 1,0 0 0 ,0 0 0

8 6 4 2 0 0 0.05 0.1 0.15 0.2 0.25


CMO D (mm)

Figure 2.25 Hysteresis of Beam C5x8.5H5 Under Cyclic Load

62

16 14 12
Load (KN)
500
40,000
140,000

10 8 6 4 2 0 0 0.1 0.2 0.3 0.4 0.5


Elastic CMO D (mm)

Figure 2.26 Hysteresis of Beam C6x8.5H5 Under Cyclic Load

14 12 10
Load (KN)

1 2000 20,000

8 6 4 2 0 0 0.05 0.1 0.15 0.2 0.25


Elastic CMO D (mm)

Figure 2.27 Hysteresis of Beam C3x8.5P5 Under Cyclic Load

63

14 12 10
Load (KN)

2000

310,000

900,000

8 6 4 2 0 0 0.05 0.1 0.15 0.2 0.25


Elastic CMO D (mm)

Figure 2.28

Hysteresis of Beam C4x8.5P5 Under Cyclic Load Static Pre-Cracking

14
1

12 10
Load (KN)

1600 310,000 900,000

8 6 4 2 0 0 0.05 0.1 0.15 0.2 0.25 0.3


Elastic CMO D (mm)

Figure 2.29

Hysteresis of Beam C4x8.5P5 Under Cyclic Load Fatigue Pre-Cracking

64

14 12 10
Load (KN)
1
10,000
1,050,000

8 6 4 2 0 0

0.05

0.1

0.15

0.2

0.25

CMO D (mm)

Figure 2.30

Hysteresis of Beam C5x8.5P5 Under Cyclic Load

14 12
1

10
Load (KN)

10,000 600,000

8 6 4 2 0 0

1,000,000

0.05

0.1

0.15

0.2

0.25

CMO D (mm)

Figure 2.31

Hysteresis of Beam C6x8.5P5 Under Cyclic Load

65

Since plane section is an assumption at a cracked section, the product of bending moment and the opening angle of the crack is not the true energy at the section. To better quantify the evolution of hysteretic behavior, the area enclosed within hysteresis loops of bending moment versus rotation is defined as pseudo energy loss. The pseudo energy loss is one measure of energy loss or damage per cycle and it was tracked for each specimen as the number of load cycles increased. The pseudo energy loss is due to many factors including cracking, micro-cracking, friction, damping, etc. The pseudo energy loss per crack, at unit width, was then calculated by dividing by the width of the specimen and plotted as a function of load cycle count (see Figure 2.32 and 2.33). From Figure 2.32 and 2.33, it was apparent that the general trend of pseudo energy loss/cycle/beam width was downward with increasing load cycles.

2.00E-05
C3 x8 .5H5

C4 x8 .5H5

1.60E-05
Energy Loss(N-m/mm)

C5x8 .5H5

C6 x8 .5H5

1.20E-05

8.00E-06

4.00E-06

0.00E+00 1 10 100 1000 10000 100000 1000000 1E+07 Number of Cycles

Figure 2.32

Unit Width Pseudo Energy Loss vs Number of Cycles in Group H

66

2.50E-05 2.00E-05
C3 x8 .5P5
C4 x8 .5P5

Energy Loss(N-m/mm)

1.50E-05 1.00E-05 5.00E-06 0.00E+00 1 10 100 1000 10000

C5x8 .5P5
C6 x8 .5P5

100000 1000000

Numbe r of Cycle s

Figure 2.33

Unit Width Pseudo Energy Loss vs Number of Cycles in Group P

These tests seem to indicate that fatigue damage to FRP RC is related to bar spacing/reinforcement ratio. The actual normalized areas contained within the hysteresis loops of specimens C5x8.5H5 and C6x8.5H5 were much smaller than those of specimens C3x8.5H5 and C4x8.5H5. Similarly in group P, as the width (bar spacing) of the specimen increased, generally speaking, the pseudo energy loss per cycle decreased. (The higher energy loss per cycle in specimen C5x8.5P5 was apparently due to the small initial crack length.) This observation would seem to imply that the energy required for crack growth becomes less and crack growth becomes more brittle as beam width (bar spacing) increases, which is analogous to the fracture behavior of metals. The dominant damage manifestation was the increase of plastic CMOD with increasing load cycles.

67

The effect of overload cycles was also investigated in FRP RC. Due to the fact that crack development typically occurs primarily within the initial 10,000 load cycles, overloads were only applied in the second stage of crack evolution (characterized by stabilized cracks) in group H and P. For specimen C5x8.5H5, a 40% overload was applied after one million cycles of service level fatigue loading. The hysteretic behavior showed very little changes after 30,000 cycles of this 40% overload (see Figure 2.34). For specimen C6x8.5H5, a 30% overload was applied after 180,000 cycles of service level fatigue loading. The hysteretic behavior of this specimen also showed very little change after 30,000 cycles of this 30% overload (see Figure 2.35). Similar results were obtained in group P, as shown in Figure 2.36 and 2.37. It would appear that relatively modest overloads, up to 40% over service load levels, have a minimal impact on subsequent service level crack opening behavior.
20
1 millio n

aft er 10 0 0 OL

16

aft er 10 ,0 0 0 OL

Load (KN)

12

0 0.1 0.15
CMO D (mm)

0.2

0.25

Figure 2.34

Effect of 40%Overload on CMOD Overload Pmax=22.3 KN Beam C5x8.5H5

68

25
b efo re o verlo ad at 18 0 k cycles

aft er 3 0 ,0 0 0 cycles o f o verlo ad

20
Load (KN)

15 10 5 0 0 0.1 0.2
CMO D (mm)

0.3

0.4

Figure 2.35

Effect of 30% Overload on CMOD Overload Pmax=20 KN Beam C6x8.5H5

20
1.25 million after 3000 OL

15
Load (KN)

after 10,000 OL

10

0 0 0.05 0.1 0.15


CMO D (mm)

0.2

0.25

0.3

Figure 2.36

Effect of 40%Overload on CMOD Overload Pmax=22.3 KN Beam C5x8.5P5

69

20
1 millio n

after 50 0 0 OL1

after 50 ,0 0 0 OL1

15
Load (KN)

10

0 0
Figure 2.37

0.05

0.1

0.15

0.2

0.25

CMO D (mm)

Effect of 40%Overload on CMOD Pmax=22.3 KN Beam C6x8.5P5

The effect of precracking by static overload is unique.

The elastic CMOD

actually decreased as more cyclic loads were applied at service level, as shown in Figure 2.38. At the same time, no additional plastic CMOD was accumulated until about 10,000 cycles, for the Isorod rebars. The explanation is as follows: during static overload crack initiations, the bonding between concrete and rebar experiences inelastic hardening. In the subsequent loading cycles, there was fatigue hardening, but no additional plastic CMOD was generated, at the working stress level. Only after 10,000 cycles of loading, the hardening effect was offset by the accumulated fatigue damage, and CMOD started growing again.

70

Comparing group H, group P and the static overload crack initiation specimen, the elastic CMOD of overload pre-cracked specimen C5x8.5P5OL was actually smaller that that of C5x8.5H5 and C5x8.5P5, which clearly indicated that there was inelastic hardening during overload. However, the plastic CMOD due to the static overload hardening was not recorded, since crack measuring instrumentation had not yet been installed. The actual total CMOD of specimen C5x8.5P5OL therefore should be larger than those of specimens C5x8.5H5 and C5x8.5P5.

0.35 0.30 0.25 CMOD (mm) 0.20 0.15 0.10 0.05 0.00 1 -0.05 Numbe r of Cycle s 10 100 1000 10000 100000 1000000 Elastic CMOD C5x8.5P5OL Plastic CMOD C5x8.5P5OL Elastic CMOD C5x8.5S5 Plastic CMOD C5x8.5S5 Elastic CMOD C5x8.5H5M

Figure 2.38

Elastic and Plastic CMOD under Ramp Load vs Number of Cycles for Specimens C5x8.5H5OL, C5x8.5S5 and C5x8.5H5M (Pmin=2.2 KN Pmax=15.6 KN)

The conventional steel reinforced concrete specimen behaved differently from the FRP reinforced concrete specimens. Compared with FRP bars, steel reinforcement has a much larger modulus of elasticity, therefore the location of the neutral axis was

71

considerably lower, or closer to rebar at the bottom. The crack lengths observed were from 100mm to 120mm (4 in to 4.75 in), which were lower than those observed in FRP RC, with similar reinforcement ratios. During the service level fatigue testing, the crack length was visually constant, although no overload was applied. Figure 2.39 shows the evolutionary history of the elastic and plastic CMODs versus the number of cycles. The crack development stage was less noticeable than that of FRP RC, due to the small magnitude of growth in crack opening, and lasted for fewer cycles (only about 500 cycles). In the crack stabilization stage, there was minor fluctuation of CMOD observed. The plastic CMOD did not become significant until after 200,000 response cycles. This phenomenon is clearly related to the bonding mechanism between conventional deformed rebar and concrete. The dominant bonding mechanism for steel rebars is bearing against the rolled on lugs, instead of friction as is the case for FRP rebars. Bond slip for conventional steel rebars is therefore generally less than that for FRP rebars.

Figure 2.39 shows the hysteresis evolutionary history of the CMOD versus number of cycle under cyclic load. The hysteresis at different testing stages was

approximately parallel to each other under working stress. No rotation was visible with the hysteresis, which indicated that there was no degradation of the bonding. The area encompassed by the hysteresis loop is also much smaller comparing to that of FRP RC. As more cycles were applied, the general trend was that the energy consumed decreased. After one million cycles under working stress, an overload of 40% above working stress was applied. It was only under overload that hysteresis started to rotate and bond

72

degradation began as shown in Figure 2.40. At the same time, unlike FRP RC, there was only a small amount of plastic CMOD induced by overload.

14
1

12 10
Load (KN)

10,000
280,000
1,000,000

8 6 4 2 0 0 0.01 0.02 0.03


CMO D (mm)

0.04

0.05

0.06

Figure 2.39 Hysteresis of Beam C5x8.5S5 Under Cyclic Load

20
1 millio n

aft er 10 0 0 OL

16
Load (KN)

aft er 150 ,0 0 0 OL

12 8 4 0 0.02 0.04 0.06


C MO D (mm)

0.08

0.1

Figure 2.40

Effect of 40%Overload on CMOD Overload Pmax=22300 N Beam C5x8.5S5

73

The area enclosed within hysteresis loops was again calculated for specimen C5x8.5S5 and C5x8.5P5OL as the number of load cycles increased, to monitor the pseudo energy loss per cycle. The pseudo energy loss per crack at unit width for specimen C5x8.5P5OL and C5x8.5S5 was plotted as a function of load cycle count in Figure 2.41. In addition to the same general trend of energy loss/cycle, the pseudo energy loss per crack, at unit width, for specimen C5x8.5S5 is also smaller than that of C5x8.5P5 and C5x8.5H5.

7.00E-06 6.00E-06 Energy Loss(N-m/mm) 5.00E-06 4.00E-06 3.00E-06 2.00E-06 1.00E-06 0.00E+00 1 10 100 1000 10000 100000 1000000
C5x8.5S 5 C5x8.5P 5OL

Numbe r of Cycle s

Figure 2.41

Unit Width Energy Loss vs Number of Cycles in Specimen C5x8.5P5OL and C5x8.5S5

Thus far, the discussions have been limited to an individual crack in a specimen. Due to the random nature of fatigue and cracking, sometimes, there are cracks in close proximity to each other.

74

Figure 2.42 Specimen C5 x 8.5 H5M

In specimen C5x8.5H5M shown above in Figure 2.42, there was one crack right next to a monitored crack, as shown in the photo. Unfortunately, as the monitored ramp loading was applied at every 10,000 cycles, the crack growth at the beginning of the test was not captured. In addition, due to operation problems, no plastic CMOD was

acquired. The plots of elastic CMOD versus cycle count, however, indicated that elastic CMOD started declining with number of cycles after 70,000 cycles, until one million cycles had elapsed.

In specimen C5x8.5P5, there was a crack at midspan, which was 115mm (4.5 in) from the two monitored cracks. The crack spacing of 115mm was low compared with other specimens of same size. It appeared that the crack started growing as expected, as more cycles were applied. But, after 10,000 cycles, the elastic CMOD started to

decrease. At around one million cycles, it began to stabilize. Comparing the elastic and plastic CMODs, they were very close to each other. In other words, the elastic CMOD became less, at the expense of a larger plastic CMOD..

75

The shear stress distribution along a bar can also be utilized as a vehicle to further explain this multiple-crack phenomenon. The rebar/concrete shear stress is at a

maximum a short distance from the crack surface, and then decreases with distance from the crack surface. When a secondary crack appears close to a primary crack, there is additional shear stress introduced onto the bar surface. Consequently, the primary crack appears stiffer and its elastic CMOD becomes smaller. As more cycles are applied, the secondary crack will propagate along with the primary one and this stiffening effect may become even stronger. The reported observation by other investigators that cyclic load improves the bonding stiffness between concrete and FRP bars may be attributable to initiation of secondary flexural cracks, limiting the opening requirement of monitored cracks.

In summary, the experiments covered one steel rebar and two different types of FRP bars, with four beam widths utilized to simulate different bar spacing in bridge deck slabs. The cracks were initiated by static pre-cracking. The CMODs were recorded when specimens were under cyclic load at working stress level. Effects of overloads at pre-cracking, and, on fully developed cracks were also investigated. For FRP RC, the results indicate that there are two stages of crack growth. One is crack development, which is characterized by growing crack opening and length. The second stage is crack stabilization at which the length of a crack is approximately constant, with slower growth in crack opening. This characterization was further verified through monitoring of the evolution of hysteresis plots, thus quantifying the evolution of energy loss per

76

cycle. The plastic CMOD appeared to grow slightly with the number of cycles, but at a decreasing rate.

Under overloads of 30% or 40% over working stress, the specimens showed no sign of distress or significant alteration of crack opening behavior. For the steel RC, the crack opening and length are smaller than those of FRP RC with similar reinforcement ratios. The crack development stage for conventional steel R/C was observed to be shorter and less apparent when compared with FRP RC. The initiation of secondary cracks close to primary cracks will tend to decrease the observed opening of the primary cracks prior to convergence. The reported improvement of bonding stiffness due to cyclic loads by others may be attributable to this phenomenon.

The profiles of crack length for C4x8.5P5, C5x8.5P5 and C6x8.5P5 indicated that the crack tips stopped at the same normal distance length from the top of the beam, although the surfaces were a little uneven. It suggests that crack length is nearly uniform across the width in the case of cyclic loading. The normal crack length will be a better parameter for the irregular surface of cracks.

77

Chapter 3

Simulation of Crack Growth

The experimental results have illustrated the performance of a number of FRP RC beam specimens. The general trend in the results of different specimens suggests the

possibility of developing a model to predict the evolution of cracks in FRP RC in fatigue environments. Due to the fact that crack growth is composed of development and

stabilization, the simulation is divided in two steps. First, initial crack mouth opening distance will be estimated.

Although ACI 440 has modified the Gergely-Lutz equation, it is difficult to estimate crack opening for complex structures and loading such as bridge deck slabs. A finite element method will be calibrated to estimate the initial CMOD at static pre-cracking of the specimens which were tested. Once valid finite element model parameters are established, it will be used to analyze more complex structures with cracks. Secondly, a fatigue model will be used to simulate the crack development. The objective is to be able to make use of the existing experiment results, and predict the performance of other structures.

Estimation of Crack Opening

It has long been known that reinforced concrete is difficult to model. To estimate the opening of a crack in a reinforced concrete beam, a discrete crack model will be used. Reinforcing bars may be modeled as truss element, which is the rebar element in ABAQUS program, which was utilized for this investigation. The drawback of this approach is that the

78

model is then limited to perfect bonding between concrete and rebar, which is not the case in reality. When a rebar is represented discretely embedded within continuum elements, it becomes possible to introduce the bond-slip effect into the model.

Initially, a substantial effort was made to simulate the mechanical behavior of the contact between the concrete and rebar. No success was achieved, however, in

establishing convergence in this effort using the finite element program ABAQUS. The apparent reason for this difficulty was that the overall behavior of the finite element (FE) model is very sensitive to the local bond-slip model. The model is not necessarily unique, however, and its apparent nature is dependent somewhat on the testing methodology used to observe it. Finding a feasible bond-slip model was considered beyond the scope of this study.

Fully Bonded

Debonded

Figure 3.1 Debonded Length Representation

Two simplified finite element modeling approaches have been proposed, however, for this investigation. They are the debonded length representation and the fictitious material representation. In the first case, shown above in Figure 3.1, a crack is modeled as a precracked surface. The bar has perfect bonding with concrete beyond a certain debonded

79

length from the crack surfaces. Within the debonded length, it is assumed that there is no tangential interaction between concrete and rebar. The actual bond stress is zero at the crack surfaces. It reaches its maximum value, however, relatively close to the crack surface. At distances further away from the crack surface, the bond stress decreases. Based on this representation, the stress within reinforcement can be calculated. The disadvantage of this model is that the interaction of two crack surfaces is neglected. The debonded length will be calibrated based on the CMOD at the beginning of the experiments for different specimens. A typical finite element model with debonded length representation is shown below.

Figure 3.2

A Typical Finite Element Mesh with Debonded Length Representation of Specimen C4x8.5P4

A debonded length was calibrated for each specimen based on the experimental results. Different specimens will result in different values of debonded length as shown in
80

the table below. It appears that a debonded length of 50mm per crack is conservative for all different specimens. Interestingly, after adding the debonded lengths of each crack within the pure bending region, the general trend was that the debonded length becomes smaller as beam width increases. It may be understood that the bonding is improved between FRP bar and concrete as beam width increases, and the debonded length subsequently decreases.

Specimen

Beam Width (mm) 75 102 127 152

Measured Crack Opening (mm) 0.22 0.19 0.17 0.16

C3x8.5P5 C4x8.5P5 C5x8.5P5 C6x8.5P5

Computed Crack Opening (mm) 0.23 0.21 0.17 0.15

Debonded Length/Crack (mm) 50mm 38mm 50mm 25mm

Total Debonded Length (mm) 200mm 75mm 100mm 50mm

Table 3.1 Calibrated Debonded Length for Group P Specimens

In the fictitious material model, a crack region is simulated as a triangular area filled with a fictitious material (Figure 3.3). The bar has been smeared into the concrete section. The base dimension of the triangular area of fictitious material has been arbitrarily selected as 2.5mm (0.1 in). The height is the true height of a crack. A small base dimension, however, will make this representation insensitive to crack length. The justification of the model is as follows. The section modulus of uncracked FRP section is less dependent on the bar due to the low reinforcement ratio. The stress within the reinforcement bridging the crack surfaces can be taken into account by the tensile stress within the fictitious material. Another benefit of the fictitious material is that it can include the interaction of two crack surfaces. As mentioned earlier, there is a fracture process zone near the tip of a crack. Tensile stress within the fictitious material will be applied on the crack surfaces, to account

81

for the interaction between crack surfaces. The disadvantage of the model is that an estimate of stress in the reinforcement is not available.

Fictitious Material

Figure 3.3 Fictitious Material Representation

Figure 3.4

A Typical Finite Element Mesh with Fictitious Material Representation of Specimen C4x8.5P4

82

The Youngs modulus, Efic, of the fictitious material was calibrated for each specimen based upon the assumed triangular dimensions and upon the experimental results after the CMOD reached its maximum value. Different specimens resulted in different Efic values as shown in the Table 3.2, below. It is another possible indication of a size effect. Also, different FRP bar types will generate different Efic due to differing bond-slip properties.

Specimen C3x8.5P5 C4x8.5P5 C5x8.5P5 C6x8.5P5

Beam Width (mm) 75 102 127 152

Measured Final Crack Opening (mm) 0.24 0.20 0.20 0.20

Computed Crack Opening (mm) 0.24 0.20 0.20 0.20

Efic

27.6 MPa (4000 psi) 24.1 MPa (3500 psi) 13.6 MPa (2500 psi) 12.4 MPa (1800 psi)

Table 3.2 Calibrated Ficticious Material Properties

Estimation of Crack Growth

The methodology used to simulate crack growth is based on the Paris equation and beam theory. Two further assumptions are made in this simulation. The first assumption concerns the stress distribution at the crack. Normally, tensile stress in concrete is excluded in the analysis, due to the fact that tensile strength is generally only about five to ten percent of the compressive strength. However, the tensile stress is crucial in the evolution of cracks, since it is the stress distribution near the crack tip that dictates the stability of a crack.

83

Concrete has been categorized as a quasi-brittle material for its fracture process zone at the crack tip. In the case of monotonic loading up to failure, a fracture process zone may be modeled as additional distributed tensile stresses between crack surfaces, with the maximum as the assumed tensile strength, at the crack tip.

In the case of cyclic loading, the fracture process zone may behave differently. At the beginning of cyclic loading, the interlocks, including aggregate bridging and crack surface friction within cracks, have to be overcome. The more cyclic loads are applied, the fewer the interlocks become. Consequently, under repeated loading, several components in the fracture process zone, such as aggregate bridging and crack face friction, will diminish due to the brittleness of concrete. Therefore, in this model, the fracture process zone is ignored and the tensile stress beyond a crack is included (Figure 3.5).

fc

a hb M

ft ac

Crack tip ff, Af

c1

Figure 3.5 Assumed Stress Distribution at a Cracked Section

84

In the diagram above, ac stands for the crack length; hb stands for the beam height; fc and ft are the compressive and tensile stresses in concrete; ff and Af are the stress and area of the FRP reinforcement.

Assuming that Youngs moduli at compression and tension are the same for concrete, the following equations are obtained, based on compatibility and equilibrium conditions:
fc a = f t hb a ac
fc a Ec = f f hb a c1 E f
(3-1)

(3-2)

1 1 f c awb = f f Af + f t (hb a ac ) wb 2 2
1 2 1 M = A f f f (hb c1 a ) + f t wb (hb a ac ) (hb ac ) 3 2 3

(3-3)

(3-4)

Substituting equation (3-1) and (3-2) into (3-3), the depth of compressive concrete is obtained:
wb (hb ac ) 2 + 2 a= Af E f Ec (hb c1 )
(3-5)

2wb (hb ac ) + 2

Af E f Ec

Finally, substitute the above equation into equation (3-4), the result is as follows.

1 h ac Ec (h a ac ) 2 wb M = Pf (hb c1 a) + b 3 3 A f E f h a c1 where Pf is the force within FRP bar.

(3-6)

85

The other behavioral assumption included is a hinge model, shown below in Figure 3.6. It has been observed in the tests of FRP RC beams that crack initiation is sudden, and that the initial crack depth is more than half of the beam depth, under either monotonic or cyclic loading. Consequently, the moment of inertia of a cracked section is much less than that of the uncracked section. In other words, most of the deformation is concentrated at the cracked sections. It is therefore not unreasonable to assume that sections in between cracks are undeformed, and that the opening of cracks in FRP RC account for all the deformation of the beam. This assumption will later be verified, based on finite element analysis. Utilizing this behavioral assumption, the following equation is obtained.

L ac

wc

Figure 3.6

A Hinge Model

wc L + wc = = ac R

(3-7)

From beam theory, -EIy=M and y=1/R. So the following relation is obtained.

86

ac =

EI wc M L + wc

(3-8)

Taking the derivatives on both sides of the equation gives the following results. da c = EI L dwc M ( L + wc ) 2
(3-9)

Substituting into the Paris equation, the following equation is obtained, after rearrangement.
dwc M ( L + wc ) 2 M CK m = LCK m dN EI L EI
(3-10)

Note that wc in the above equation is the elastic portion of the crack opening; L is the spacing of cracks; E and I are the beam modulus of elasticity and moment of inertia respectively; and
N is the number of cycles. K is the range of the difference of stress intensity factors for M

and Pf which are calculated based on a standard handbook (Tada et al 1985).

The stress intensity factor for pure bending is as follows.


KI = 6 M ac h
2 b

[1.12 1.39(

ac a a a ) + 7.32( c ) 2 13.1( c ) 3 + 14.0( c ) 4 ] hb hb hb hb

(3-11)

In the case of a concentrated load on the crack surface, the stress intensity factor is listed as follows:
G( c ac , ) ac hb
3

KI =

2P ac

a c (1 c ) 2 1 ( ) 2 hb ac

(3-12)

87

G(

a c a c a c a c ac , ) = g1 ( c ) + g 2 ( c ) + g 3 ( c )( ) 2 + g 4 ( c )( )3 hb ac hb ac hb ac hb ac hb ac a a a a ) = 0.46 + 3.06( c ) + 0.84(1 c )5 + 0.66( c ) 2 (1 c ) 2 hb hb hb hb hb ac a ) = 3.52( c ) 2 hb hb ac a a a a 3 a a a ) = 6.17 28.22( c ) + 34.54( c ) 2 14.39( c )3 (1 c ) 2 5.88(1 c )5 2.64( c ) 2 (1 c ) 2 hb hb hb hb hb hb hb hb ac a a a a 3 a a a ) = 6.63 + 25.16( c ) 31.04( c ) 2 + 14.41( c )3 + 2(1 c ) 2 + 5.04(1 c )5 + 1.98( c ) 2 (1 c ) 2 hb hb hb hb hb hb hb hb

g1 (

g2 ( g3 ( g4 (

where ac is the crack length; hb is the beam height; c is the distance from the load to the crack edge; P is the concentrated load, M is the bending moment.

/2 D/2

D/2

Figure 3.7

Verification of Hinge Model

Finite element models for the calibration of test specimens were also used to verify the hinge assumption. The following sketch is pure bending region of the experiment setup;
D was 610mm (24 in). The angle has the following expression.

88

2 R

(3-13)

Also, we have the following equation. sin

D 2

(3-14)

Due to the small magnitude of , the following equations are obtained. sin = 4 D
(3-15)

The following results are reached by equating equation (3-13) and (3-15). 1 8 = R D2 Equation (3-7) becomes the following.
wc 8 8L L = 2 ( L + wc ) 2 = (U.S. units) ac D 72 D
(3-17) (3-16)

The results of the finite element analysis are listed below in Table 3.3; with relative differences all less than 10%, the hinge assumption is justified.

Specimen

wc ac

L 72

Difference

C4x8.5H5 C4x8.5P5 C5x8.5H5 C5x8.5P5 C6x8.5P5

0.00654 0.01302 6 = 0.00119 = 0.00109 5.5 72 0.00696 0.01121 7.5 = 0.00127 = 0.0011 5.5 72 0.006407 0.006973 12 = 0.00128 = 0.0011 5 72
Table 3.3 Hinge Assumption Verification

9% 8% 9%

89

Sensitivity Analysis of the Model and Variation of Crack Growth Estimation

The unknown Paris Equation parameters in the crack growth simulation are C and m, which will be determined based on experimental results. Another difficulty will be the determination of initial crack length, since crack length may not be uniform across the width direction. In the case of multiple cracks, with possibly different initial crack lengths, it is impossible to use a compliance calibration method. For thin specimen such as 76 mm and 102 mm, the true crack length should be very close to the surface length. Since the initiation of cracks may be triggered by the presence of a random flaw within a structure, the exact spacing of cracks is random. The modulus of elasticity for concrete is also variable,

depending on the ingredients and curing process. The property of an FRP bar is also subject to uncertainty in the manufacturing process and materials themselves. To address these variables with uncertainties, a sensitivity analysis was conducted to evaluate the response of the model to these variables, using specimen C5x8.5H5 as a prototype.

First, the parameters C and m in the Paris equation were investigated. The parameter
m was set to be 3.76, which was similar to the value reported by other researchers (Baluch et al 1987). Three different C values were used; namely, 6.76x10-4, 2.25x10-4, 7.51x10-5

mm/cycle/(MPa m1/2)m , corresponding to 2x10-16, 6.6x10-17 and 3.3x10-17 in/cycle/(psi in1/2)m in U.S. units. The results were shown in Figure 3.8. With three C values at the ratio of

1:1/3:1/9, the opening increment only changed by 0.005mm and the final crack length differed by 5mm. What is more important is that the trajectories of crack opening increment

90

were similar to each other. Obviously, the model is insensitive to C. C was therefore fixed at 6.76x10-4 mm/cycle/(MPa m1/2)m or 2x10-16 in/cycle/(psi in1/2)m .

Three different m values were subsequently specified at 3.66, 3.76 and 3.86. The results showed that the curve for crack opening growth became flat at m of 3.86, and that each was completely different from the others. The crack opening increment, however, did not change more than 0.003mm. Considering the difference of crack opening increments for different values of m values was only 3%, it was concluded that the model of crack opening growth is very sensitive to m although the final crack opening increment is insensitive to m. The plots of this data are shown in Figure 3.9.

The second set of tested parameters was initial crack length and crack spacing with C set at 6.76x10-4 and m at 3.76. Due to the aforementioned difficulties, the measured surface crack was always an estimate. The discrepancy might be as much as 25mm, as suggested by Swartz et al (1984). Trajectories of crack length and opening growth were therefore

generated for an array of initial crack lengths. The difference between the maximum and minimum initial crack length was 38mm (1.5 in). It was found that crack opening growth was sensitive to the initial crack length although the final crack opening increment is insensitive to initial crack length. The final crack opening increment might be 0.01mm less, if the initial crack length was overestimated by 25mm. All trajectories of crack opening growth were similar to each other and the final crack length was not affected.

91

Three different values of crack spacing were examined, with a difference of about 40mm. The curve for crack length growth was not related to different crack spacing. The crack opening increment was approximately 0.05mm more when crack spacing was 40mm larger. This explains the fact that cracks within the pure bending region may have different crack opening growth curve, yet all cracks stop at the same length. The plots are shown in Figure 3.10 and 3.11.

The next two tested parameters were elastic modulus of concrete Ec , and elastic modulus of FRP bars Ef , with fixed C of 6.76x10-4 and m of 3.76. Values of Ec were set at 27.6MPa, 34.5MPa and 41.4MPa. Since manufactures of FRP usually gave the guaranteed
Ef, the values of Ef , 1.15Ef and 1.3Ef were examined. As Ec increased from 27.6MPa to

41.4MPa, the final crack length changed by 20mm although the trajectories were still similar to each other. The crack opening growth, however, was insensitive to concrete modulus of elasticity. With respect to a 15% increase of Ef, the crack length would be 5mm shorter and the crack opening increment would be 0.002mm less, approximately. The plots are shown in Figure 3.12 and 3.13.

Finally, the sensitivity of the specimen size was investigated.

Since forms of

specimens might deform during concrete placement, final specimen size can be different from the nominal size. The width of specimen C5x8.5H5 was given an error of 6mm (0.25 in). The trajectories of crack length growth for three widths of 121mm, 127mm and 133mm almost coincided with each other. The error for crack opening increment was about

0.001mm less, with 6mm increase in beam width and vice versa. The height of specimen

92

C5x8.5H5 was also given an error of 6mm (0.25 in). The trajectory of crack length growth was about 5mm more with 6mm more beam height and vice versa. The error of crack opening increment was about 0.001mm less, with 6mm less beam height, and vice versa. The results of recorded crack opening growth will therefore not be substantially affected by the possible variation of beam size from the nominal size. The plots are shown in Figure 3.14 and 3.15.

In summary, the prediction of growth of crack length and opening is subject to variability inherent with parameters such as modulus of elasticity, crack length, specimen size and C and m in the Paris equation. This model is most sensitive, however, to the parameter m in the Paris equation, due to the nature of the exponential function. The parameter C is the least sensitive parameter in the model. Initial crack length should be measured accurately due to moderate sensitivity. For other variables, such as moduli of elasticity for concrete and FRP, and specimen size, their sensitivities are low and the impacts on the model are of the same order as the resolution of data acquisition, as long as the variability is within a reasonable range.

93

Figure 3.8

Sensitivity Analysis Results on Parameter C for Beam C6x8.5H5 (m=3.76)

Figure 3.9

Sensitivity Analysis Results on Parameter m for Beam C5x8.5H5 (C=6.76x10-4)

94

Figure 3.10

Sensitivity Analysis Results on Initial Crack Length a0 for Beam C5x8.5H5 (C=6.76x10-4 m=3.76)

Figure 3.11

Sensitivity Analysis Results on Initial Crack Spacing L for Beam C5x8.5H5 (C=6.76x10-4 m=3.76)

95

Figure 3.12

Sensitivity Analysis Results on Concrete Elastic Modulus Ec for Beam C5x8.5H5 (C=6.76x10-4 m=3.76)

Figure 3.13

Sensitivity Analysis Results on Concrete Elastic Modulus Ef for Beam C5x8.5H5 (C=6.76x10-4 m=3.76)

96

Figure 3.14

Sensitivity Analysis Results on Specimen Width b for Beam C5x8.5H5 (C=6.76x10-4 m=3.76)

Figure 3.15

Sensitivity Analysis Results on Specimen Height h for Beam C5x8.5H5 (C=6.76x10-4 m=3.76)

97

Simulation of Experiment Results

Based on the sensitivity analysis, it is understood that parameter m is the most crucial one in the model. The model is less sensitive to parameter C and initial crack length. As the goal was to simulate the experiments, all three parameters were first determined so that the model best fits the experimental data, using a brute force approach. The objective function was the minimum of the sum of the square of the differences between the model and the experimental data. For both thin and thick specimens, the result of crack length was always the surface crack length observed. It was then determined that surface crack length could be used in the simulation of crack growth. For parameters C and m in Paris equation, the model is obviously more sensitive to m than C , since m is the exponential term. The value of C is in the order of 10-16 in U.S. units. To simplify the model, a fixed value of C was set to be 6.76x10-4 mm/cycle/(MPa m1/2)m ( 2x10-16 in/cycle/(psi in1/2)m ). Converting to unit of mm/cycle/(Pa m1/2)m, C is in the order of 10-25 which agrees with the results reported by Baluch (1987).

For specimens of group H, all specimens except C6x8.5H5 were simulated due to the fact that there was excessive load at pre-cracking with C6x8.5H5. The results are listed from Figure 3.16 to 3.18. The same analysis was performed for all specimens of group P. Similar results are shown Figure 3.19 to 3.22. A summary was shown in Figure 3.23, which illustrates a size effect, in that m decreases as the beam width decreases. In other words, the beam becomes less brittle as the width becomes smaller.

98

This phenomenon is analogous to that of metallic materials. As the width of a metallic specimen is small, the size of the plastic zone is large and the specimen is in a plane stress state. When the specimen width increases, the radius of plastic zone starts decreasing and the state of stress approaches plane strain. In the case of FRP concrete, however, a fracture process zone exists near the tip of the crack, and, the depth of the fracture process zone is not related the state of stress. This size effect is therefore not caused by the fracture process zone.

Figure 3.16

Crack Length and Crack Opening Increment versus Number of Cycles Beam C3x8.5H5, C=6.76x10-4, m=3.48

99

Figure 3.17

Crack Length and Crack Opening Increment versus Number of Cycles Beam C4x8.5H5, C=6.76x10-4, m=3.57

Figure 3.18

Crack Length and Crack Opening Increment versus Number of Cycles Beam C5x8.5H5, C=6.76x10-4, m=3.76
100

Figure 3.19

Crack Length and Crack Opening Increment versus Number of Cycles Beam C3x8.5P5, C=6.76x10-4, m=3.39

Figure 3.20

Crack Length and Crack Opening Increment versus Number of Cycles Beam C4x8.5P5, C=6.76x10-4, m=3.55

101

Figure 3.21

Crack Length and Crack Opening Increment versus Number of Cycles Beam C5x8.5P5, C=6.76x10-4, m=3.74

Figure 3.22

Crack Length and Crack Opening Increment versus Number of Cycles Beam C6x8.5P5, C=6.76x10-4, m=3.88

102

4
Group H Group P

3.8
m Value

3.6

3.4

3.2 200

150

100

50

Beam Width (mm)

Figure 3.23 Size Effect of Beam Width on Paris Equation

103

Chapter 4

Finite Element Modeling and Analysis of a Realistic FRP Reinforced Concrete Slab

The experiment results and corresponding analytical models have concluded that FRP has significant potential as a possible reinforcement for bridge slabs. The objective of the finite element analysis is to investigate and predict the performance of FRP reinforced concrete slabs under realistic conditions, by applying the aforementioned behavioral parameters with finite element models. First, the debonded length

representation will be used to analyze concrete slab strips, and resulting FRP stresses will be computed. The arching effect will also be examined. Secondly, due to its relative simplicity, the fictitious material representation will be used to investigate the performance of an entire bridge deck.

Analysis of Slab Strips

Since a large number of elements are generated in the debonded length representation, it is not feasible to use it for an entire bridge deck. From the AASHTO design guide, some empirical slab strip widths have been listed in AASHTO Table 4.6.2.1.3-1. The values of strip width are based on experience. For a cast-in-place concrete slab, the width of primary strip is 660+0.55S for positive moments, where S is the girder spacing in millimeters. To account for the effect of continuity, and to simulate a worst case scenario, slabs of two spans over three girders will be analyzed. There are diaphragms between the girders with an X configuration, plus top and bottom chord

104

bars. Design loads are AASHTO design truck loads and slab self weight. The wheel load is 71.3 KN (16 kips) for an HS20 design truck load, with a lateral spacing of 1.8 m. A distributed lane load is not required in the slab strip analysis per AASHTO. For the purpose of simplicity, the following assumptions are made. 1) The values of girder spacing are 1.8 m (6 feet), 2.7 m (9 feet) and 3.6 m (12 feet), which represent the majority of bridges in service. The corresponding strip widths are 1.7 m (66 in), 2.2 m (86 in) and 2.7 m (105 in). 2) There is only one crack at the midpoint of each span,

since a bridge slab is under a concentrated wheel load. 3) One wheel of the design truck will be applied above the crack to represent the critical condition. 4) The slab is 215 mm thick. 5) The steel girder is a typical W36x135 section.

Both the slab and rebar are modeled with 20-node quadratic brick element in the FE program ABAQUS. Debond is designated for a distance of 25 mm on each side of the crack. The rest of the interface between bar and concrete is defined to be fully attached. The FRP bar spacing used in the strip model is 100mm, due to the fact that normal reinforcing steel bar spacing is approximately 125mm in bridge slabs. The initial crack length was set to be 150mm (6 in) which was the initial crack length of specimen C4x8.5P5. The distribution of the wheel load is approximated according to equation 3.6.1.2.5-1 in AASHTO, i.e., the intensity of the wheel pressure is always constant at 0.86 MPa (125 psi), and the width of the loading area is always 0.5m (20 in). The boundary conditions of the model are as follows. All exterior girders were constrained in vertical and girder axial (longitudinal) directions at the bottom of the girder webs. Only one exterior girder was constrained in transverse direction.

105

First, the case of a girder spacing of 1.8m (6 ft) was analyzed. The diaphragm was composed of two cross bars and a bottom horizontal bar with a section area of 1600mm2 (2.5 in2). The cross section area of the top bar of the diaphragm was set to be zero. To investigate the arching effect, one wheel load was applied first at the midspan of the slab as shown in Figure 4.1. The resulting stress contour is plotted in Figure 4.2. It is obvious that the only area of compressive stress was around the loading at midspan. Arching effect was not noticeable for the load case of one wheel load.

Figure 4.1

Slab Strip Model under One Wheel Load (Girder spacing 1.8m, 16M Bar at 100mm, Slab thickness 215mm)

106

Figure 4.2

Transverse Normal Stress Contours Under One Wheel Load of Design Truck (Girder spacing 1.8m, 16M Bar at 100mm)

Next, two wheel loads were applied to the slab strip as shown in Figure 4.3. From the stress contour plot, it can be seen that a significant area of compressive stress appeared both at the slab top near edge and at the bottom of the slab in the middle. The tensile stress also decreased significantly, implying there is significant arching effect in this case of an axle load.

107

Figure 4.3

Slab Strip Model under One Axle Load of Design Truck (Girder spacing 1.8m, 16M Bar at 100mm, Slab thickness 215mm)

Figure 4.4

Transverse Normal Stress Contours Under One Axle Load of Design Truck (Girder spacing 1.8m, 16M Bar at 100mm)

Next, the effects of girder spacing were examined. Two cases of girder spacing, 2.7m and 3.6m, with 16M bar spaced at 100mm were analyzed. To investigate the effects of rebar spacing, the case of 150mm bar spacing at 3.6m girder spacing was also analyzed. The results are illustrated below.

108

Figure 4.5

Slab Debonded Length Representation under One Axle Load (Girder spacing 2.7m, 16M Bar at 100mm, Slab thickness 215mm)

Figure 4.6

Transverse Normal Stress Contours Under One Axle Load of Design Truck(Girder spacing 2.7m, 16M Bar at 100mm)

109

Figure 4.7

Slab Strip Model under One Axle Load of Design Truck (Girder spacing 3.6m, 16M Bar at 100mm, Slab thickness 215mm)

Figure 4.8

Transverse Normal Stress Contours Under One Axle Load of Design Truck (Girder spacing 3.6m, 16M Bar at 100mm)

110

Figure 4.9

Transverse Normal Stress Contours Under One Axle Load of Design Truck (Girder spacing 1.8m, 16M Bar at 150mm)

From the contour plots from larger girder spacing, the arching effect has been illustrated by a dome of concrete under compression, regardless of the girder spacing. The results imply that a composite bridge slab is different from a conventional multiple span continuous slab. The lateral stiffness of the girder and diaphragms supplies

substantial lateral constraint to the bridge slab. The difference in rebar spacing did not appear to have a very significant effect on the concrete compressive zone. Much larger ultimate strength will be achieved when arching action is considered than would be the case when considering the theoretical bending strength of the slab alone.

Next, the bar stress in the slab strip was studied. In Figure 4.10, the bar stresses were plotted versus the distance from the center of the assumed wheel load. The rebar stress obviously decreased at greater distances from the load center. The magnitude of

111

stress at the assumed 3.6m (12ft) girder spacing and the 100mm rebar spacing was about 71.1 MPa (10.3 ksi). The maximum rebar stress at 3.6m (12ft) girder spacing and 150mm rebar spacing was about 89.0MPa (13.0 ksi), which was well less than the tensile strength of the bars. Considering the fact that slabs are under constant traffic load, it was verified that fatigue and serviceability under repeated loads are the critical factors in slab design, rather than the static ultimate strength.

There would be no fatigue failure expected with FRP bars under the calculated stress levels. The compressive stress levels in concrete itself was only about 10% of the ultimate strength, which should produce a fatigue life of over 10 million cycles. The remaining issues, therefore, are durability and serviceability, including crack opening and slab deflection.

120
3.6m Girder Spa. 100mm Bar Spa. 2.7m Girder Spa. 100mm Bar Spa.

100

1.8m Girder Spa. 100mm bar Spa. 3.6m Girder Spa. 150mm Bar Spa.

Bar Stress (MPa)

80

60

40

20

0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Distance from Wheel Load (m)

Figure 4.10 Bar Stress Under Design Truck Load with 16M Bars

112

To analyze the crack opening of the slab strip, diaphragms were placed at each end of the slab with two diagonal bars plus a bottom bar. At the girder spacing of 1.8m (6 ft) with 16M bars spaced at 100mm, the maximum total crack opening under load center was 0.055mm (0.0022 in), under the self-weight and the axle load of a design truck. The maximum crack opening due to the axle load alone was 0.051mm (0.00202 in). The maximum deflection under the axle load was 0.21mm (0.0083 in), which should be acceptable, as it represents a deflection/span ratio of less than 1/8000. Compared with the maximum 0.5mm suggested in ACI 440, the initial crack opening was very conservative for the 1.8m girder spacing.

To further decrease the crack opening of an FRP reinforced slab, a top diaphragm bar, having a cross section area of 1600 mm2 (2.5 in2), which directly connected the top of the webs of adjacent girders, was added at each diaphragm location. The maximum crack opening due to the design axle load decreased to 0.051mm (0.0020 in). The maximum deflection under the axle load was 0.19mm (0.0073 in). To further investigate the top diaphragm bar effect, the cross section area of that bar was increased to 6450 mm2 (10 in2). The maximum crack opening due to axle load decreased only slightly more, to 0.048mm (0.0019 in). The maximum deflection under the axle load was 0.18mm (0.0072 in). Finally, the entire diaphragm was removed. The maximum total crack opening under the load center only increased to 0.058mm (0.0023 in), under the self-weight and axle load of the design truck. The maximum deflection under axle load only increased to 0.22mm (0.0085 in), which obviously implied that restraint from the diaphragms played a very minor role in controlling crack openings and slab deflections.

113

A similar analysis was conducted for the case of a girder spacing 2.7m (9 ft) with 16M bars spaced at 100mm. With diaphragms of two cross bars and a bottom bar, the maximum total crack opening under the load center was 0.089 mm (0.0035 in) under the self-weight and axle load of the design truck. The maximum crack opening due to the axle load only was 0.080mm (0.0032 in). The maximum slab deflection under the axle load was 0.53mm (0.021 in), which, again, should be acceptable. When a top bar was added to the diaphragm, with the same cross section area 1600 mm2 (2.5 in2), the maximum crack opening due to the axle load only was 0.077mm (0.0031 in). The maximum deflection under the axle load was virtually the same. Even after the cross section area of the top was increased to 6450 mm2 (10 in2), the maximum crack opening due to axle load only decreased to 0.077mm (0.00303 in). Since the changes in crack opening and slab deflection were so small, it was concluded that serviceability of concrete slabs in this strip model are insensitive to the addition of top bars to the diaphragms, or even to the presence of lateral diaphragms at all.

Analysis was also performed for the case of a girder spacing 3.6m (12 ft) with 16M bars spaced at 100mm. With diaphragms of two cross bars and a bottom bar, the maximum total crack opening at the load center was 0.13 mm (0.0050 in) under the selfweight and axle load of the design truck, which was still within the crack opening limits suggested by ACI 440. The maximum crack opening due to axle load only was 0.108mm (0.0042 in). The maximum deflection under axle load was 1.4mm (0.055 in). The spacing for 16M bar was then increased to 150mm. Diaphragms were removed for the

114

sake of simplicity, and due to their ineffectiveness. The maximum total crack opening at the load center was 0.16 mm (0.0063 in), under the self-weight and axle load of the design truck, which is also within the suggested design limits in ACI 440.

The results from the slab strip analyses with debonded length representation indicate that a 16M FRP bar, at 150mm spacing, is sufficient flexural reinforcement for concrete slabs with girder spacing up to 3.6m (12 ft). The predicted FRP rebar stress is quite low, and the predicted crack opening is only 33% of the limit suggested by ACI 440, namely an opening of less than .5mm (.02 in). A rebar spacing of 100mm obviously produces even lower predicted crack openings (25% of the ACI 440 limit) and smaller slab deflections (deflection/span ratio below 1/2000).

Analysis of Full Bridge Slabs

The slab strip model is quite often utilized in slab design, due to its simplicity and, presumably, its expected conservatism. The strip width is, however, somewhat arbitrary. Since the ultimate goal of a design is to ensure that the long-term serviceability of bridge slabs is satisfactory, it is then imperative to more fully investigate a complete bridge slab, under the design truck load and lane loads.

In the State of Ohio, there are four legal loads. The heaviest legal truck load is a 5C1, which is composed of one axle load of 53.38 KN (12 kips) followed by four axle loads of 75.62 KN (17 kips). The axle spacings are 3.6m, 1.2m, 9.4m and 1.2m,

115

producing a total length of 15.5m. It is considerably different from the AASHTO design truck or AASHTO design lane load. It is therefore meaningful to compare the

performances of bridge slabs both under the AASHTO design load and a maximum Ohio legal load, such as 5C1.

The model bridge was single span of 18.3m (60 ft) long. Similar to slab strip model, the slab was supported by three girders. The slab thickness was 215mm (8.5 in). Cross bar diaphragms were first not included in the model, but were later added to evaluate their effectiveness. There was again one initial crack assumed midway between girders. One wheel load was located on top of each crack location. Due to the large size of the structure, it would not be realistic to model it using the debonded length representation, which would generate too many elements, especially including discrete rebars. With the fictitious material representation, which was calibrated for final CMOD from the experimental program, it was possible to model the entire bridge and evaluate final elastic crack opening and deflection. Since the experiments were conducted with bars under full service load 140MPa (20 ksi), and bar stresses in slabs were generally much smaller, the calibrated fictitious material representation will tend to provide an upper bound estimate of the expected crack opening.

The design load in the analyses consisted of the AASHTO design truck and the design lane load. The design truck is composed of three axles. The front axle load is 35.6 KN (8 kips). The second and third axles are 142.5 KN (32 kips). The spacing between the first axle and the second axle is 4.27m (14 ft). The spacing between the

116

second axle and the third axle varies between 4.27m (14 ft) and 9.15m (30 ft). In the case of single span bridge, minimum spacing of 4.27m (14 ft) would be the critical condition. The lateral spacing between wheels of each axle is 1.8m (6 ft). The lane load is a uniformly distributed load with the intensity of 9.35 KN/m (0.64 Kips/ft). The transverse distribution width may be assumed to be 3.05m (10 ft). The second axle of the design truck was placed at midspan of the bridge as shown in the model in Figure 4.11. The slab and the fictitious material were modeled with the 20-node quadratic brick element in the FE program ABAQUS.

In Figure 4.11 and 4.12, the model and stress contours are shown for the case of a 1.8m (6 ft) girder spacing, with no diaphragms. Compared with Figure 4.4, it can be seen that the actual zone of concrete under compression is larger than the width of the assumed slab strip. It is also apparent that there was only one large dome of concrete under compression, spanning the two girder spacings. The maximum predicted crack opening under the design load was 0.081mm, with 16M FRP bar at 100mm spacing, in the bridge model.

For comparison purposes, a slab strip model under the same condition was created, using the fictitious material model. The maximum predicted crack opening under the design load was 0.053mm, actually less than that predicted by the full deck model. Since the slab strip is always a narrow strip, relative to the actual imposed stress field, girders underneath the slab strip represent a comparatively large stiffness in the model. To verify the hypothesis that the girder stiffness is the cause of the discrepancy in

117

crack opening prediction, all three girders were fixed vertically for the entire length of the bridge. From the model and results shown in Figure 4.13 and 4.14, the concrete

compressive stress zone had decreased significantly. The single dome was split into two again. The maximum predicted crack opening was 0.041mm, which was smaller than that of the slab strip model, as expected.

Finally, diaphragms were added to the bridge model at the spacing of 4.6m (15 ft). The contours of transverse normal stress were plotted in Figure 4.15. The arching effect was indicated by two compression domes around the axle. The maximum predicted crack opening was 0.058mm. The crack opening predictions from the slab strip model and bridge model with diaphragms are similar. Without diaphragms, the excessive deflections of the interior girder tend to produce in one large dome of concrete under compression, and excessive crack opening predictions.

The crack opening predictions of the bridge model at different rebar spacings were calculated for girder spacings of 1.8m, 2.7m and 3.6m. The models and stress contours are shown in Figures 4.15 through 4.19. As the girder spacing changed from 1.8m (6 ft) to 3.6m (12 ft), there was always one compression dome, as an indication of the arching effect.

The results of crack opening prediction versus reinforcement spacing are plotted in Figure 4.20. The solid lines represented the crack opening predictions of bridges with diaphragms at 4.6m spacing, while the hidden lines of the same color represent the results

118

of the same model without diaphragms. It can be seen that the crack opening increased by 30 to 50% after diaphragms were removed. Although the primary purpose of

diaphragms is to provide lateral stability for the girders, the diaphragms also contribute to the load distribution within the bridge superstructure and crack control within the deck slab.

Even at a girder spacing 3.6m (12 ft), the maximum final crack opening was 0.21 mm, for 16M FRP rebars at 150mm spacing. Based on the results of the fatigue

experiments conducted at service load levels, the final plastic crack opening contribution is typically less than 25% of the elastic contribution, and may be estimated conservatively as half of the elastic crack opening. Even including an impact factor of 1.3, the

conservative estimate of final total crack opening is only 0.41mm, or only 82% of the suggested ACI 440 limit. Using maximum crack opening as the criterion, the spacing 150mm of 16M FRP bar would appear to be satisfactory up to a 3.6m (12 ft) girder spacing.

The serviceability of the bridge model under a realistic load Ohio legal load 5C1 was also investigated. The model is shown in Figure 4.22 for the bridge model with 1.8m girder spacing. The concrete compressive zone was similar to that of AASHTO design load. Nevertheless, the maximum crack opening under the Ohio legal load 5C1 was about 2/3 of the crack opening of AASHTO design load at different reinforcement spacing, as shown in Figure 4.21.

119

Figure 4.11

Slab Model under Load of Design Truck, Lane Load and Self-Weight. (Girder spacing 1.8m, Slab thickness 215mm, No Diaphragm)

Figure 4.12

Transverse Normal Stress Contours Under Loads of Design Truck, Lane Load and Self-Weight.

120

Figure 4.13

Slab Model under Load of Design Truck Only with Girders Fixed Vertically (Girder spacing 1.8m, Slab thickness 215mm)

Figure 4.14

Transverse Normal Stress Contours Under Design Truck Only with Girders Fixed Vertically.

121

Figure 4.15

Transverse Normal Stress Contours Under Design Truck Only with Diaphrams.

Figure 4.16

Slab Model under Load of Design Truck, Lane Load and Self-Weight. (Girder spacing 2.7m, Slab thickness 215mm)

122

Figure 4.17

Transverse Normal Stress Contours Under Loads of Design Truck, Lane Load and Self-Weight.

Figure 4.18

Slab Model under Load of Design Load and Self-Weight. (Girder spacing 3.6m, Slab thickness 215mm)

123

Figure 4.19

Transverse Normal Stress Contours Under Design Load and Self-Weight.

0.3 1.8m Girder Spa.w / Disph. 2.7m Girder Spa.w / Diaph. 3.6m Girder Spa.w / Diaph. 1.8m Girder Spa.No Diaph. 0.2 2.7m Girder Spa.No Diaph. 3.6m Girder Spa.No Diaph.

CMOD (mm)
0.1 0 50

70

90

110

130

150

170

Reinforcement Spacing (mm)

Figure 4.20

Maximum Crack Opening under Design Load in Model Bridge

124

0.1

Design Load Ohio Legal Load 5C1


0.08

CMOD (mm)

0.06

0.04

0.02

0 50 70 90 110 130 150 170

Reinforcement Spacing (mm)

Figure 4.21

Maximum Crack Opening Under Design Load and Ohio Legal Load 5C1 in Model Bridge; 1.8 m Girder Spacing

Figure 4.22

Slab Model under Load of Ohio Legal Truck Load 5C1 and SelfWeight. (Girder spacing 1.8m, Slab thickness 215mm)

125

Figure 4.23

Transverse Normal Stress Contours Under Loads of Ohio Legal Truck Load 5C1 and Self-Weight.

To compute the distribution factors, one girder of composite section is first analyzed under one lane of design load. Then, design loads are first applied to a bridge, and the moment and shear of the critical girder are calculated. The ratios of the moment and shear of single girder and multiple girders will be the distribution factor. For the critical girder, there shall be a line of wheels at its centerline which constitutes majority of the distributed load. In the current AASHTO LRFD design codes, the distribution factors are related to the span of bridge, slab thickness, girder spacing and longitudinal stiffness of the bridge. A cracked section has a smaller stiffness than the uncracked portion the slab in the transverse direction. When a crack is in the vicinity of a wheel load, between girders, more load will be carried by the immediately adjacent girders, and less load will be distributed across the remaining width of the bridge. As the majority of

126

the load is from the wheel on the girder line, the effect of a crack in a slab on the other wheel should be insignificant.

Empirical Design of Bridge Slabs

The analysis discussed has been limited to the serviceability of a deck slab at the maximum positive moment point in the transverse direction. At negative moment deck slab sections at girder lines, there is usually a haunch of 50mm or more. The top flange is also typically encased in the concrete. As a result, the generally smaller negative

moments in the transverse direction, at girder lines, are resisted by larger concrete sections. In the longitudinal slab direction, the bending moments are so small that they are usually ignored, although temperature and shrinkage effects, as well as composite behavior assumptions typically require longitudinal reinforcement in bridge decks, as discussed below.

Aside from transverse stress analysis of wheel or lane load effects, the effects of temperature and shrinkage needs to be included for the longitudinal direction by supplying minimum reinforcement parallel to the traffic direction of the bridge. ACI 440 defines the temperature and shrinkage reinforcement requirement as follows.

= 0.0018

60,000 Es f fu E f

(4-1)

where ffu is the design tensile strength, Ef and Es are the elastic moduli of FRP and steel, respectively. The minimum primary reinforcement ratio was 0.006 for FRP bars in this

127

study. For 215mm thick bridge deck slabs, the temperature and shrinkage reinforcement should typically be 16M top and bottom at 300mm maximum spacing. The analysis results have indicated that 16M at 150mm is satisfactory for maximum positive moment in the transverse direction. According to the results of Perdikaris et al (1988, 1989), the fatigue life of slabs with isotropic reinforcement is twenty times that with orthotropic reinforcement. Therefore, it is proposed that the reinforcement should be a minimum of 16M spaced at 150mm, top and bottom in both directions, for girder spacing up to 3.6m (12 ft). Although this design seems to be simple, it is based on meeting serviceability (crack opening) requirements, and it does provide adequate strength, particularly taking into account the arching effect of typical bridge deck slabs.

In summary, it has been demonstrated that the design criteria for bridge deck slabs should be serviceability, instead of ultimate strength, due to the generally low stress level of design traffic loads. Arching effect does play an important role in the performance of bridge slabs. The additional transverse constraint supplied by the girders and diaphragms enhances the strength and serviceability of bridge deck slabs, and enhances as well the secondary load path represented by arching action.

Empirical designs are often adopted in bridge deck slab design. Conventional strength design ignores fatigue effects on serviceability, and the arching effect on strength. The analysis results of this study indicate that reinforcement of 16M (#5) at 150mm (6 in) spacing is satisfactory for girder spans up to 3.6m 912 ft). FRP bridge deck slabs have considerable potential for improving corrosion resistance in bridge decks.

128

The somewhat larger predicted crack openings associated with FRP RC bridge decks should be considered admissible by bridge owners and inspectors, given the enhanced corrosion resistance.

129

Chapter 5

Conclusions

Based on the limited number of fatigue tests conducted, FRP is a promising alternative to traditional steel reinforcement in bridge deck slabs. The CMOD in an FRP RC member can be considered to consist of elastic and plastic portions. The elastic CMOD disappears after unloading; the plastic CMOD remains after unloading. As more load cycles are applied, the elastic CMOD, under constant load amplitude, experiences growth to stabilization. The permanent CMOD at zero load, which is the plastic portion, generally increases with the number of load cycles.

The CMOD convergence of FRP RC differentiates it from conventional steel RC. At the end of 2,000,000 cycles of full service load testing, the permanent CMOD is less than half of the elastic CMOD. No fatigue failures were observed up to two million cycles. The cracks generated either by static pre-cracking or during cyclic loading are single cracks with no branches. The specimens with more than balanced reinforcement ratio presented multiple cracks in static pre-cracking whereas only one crack appeared for the specimen with a near balanced reinforcement ratio. In addition, the crack opening formula in ACI 440 appears to overestimate the crack opening, even for the case of widely spaced cracks.

Current results from this investigation cannot confirm the results of other researchers, that bonding between concrete and FRP actually can be improved under cyclic loading. The general trend of elastic CMOD is that it grows with fluctuation until

130

convergence. The CMOD reduction encountered in the tests was possibly attributable to the effects of additional cracks, which originated during cyclic testing.

A model which is based on the Paris equation was proposed to predict the evolution of crack growth. Tensile stress at the crack tip was included in the section stress distribution. The stress intensity factor was assumed to be the sum of those for pure bending and for the concentrated force within the FRP bar. The model has been proved to predict crack growth in FRP RC reasonably well. A size effect was observed. As the thickness of the specimen decreases, the exponential parameter in the Paris equation decreased. The sensitivity analysis illustrated that the model was most sensitive to the exponential parameter.

To extend the findings to typical bridge deck slab design, two simplified finite element crack representations were developed to simulate the test specimen behavior, and then to analyze realistic bridge deck slabs. The first representation is a debonded length representation. A debonded length of 25mm (1 in) between the concrete and the FRP rebar, on each side of a crack, was found to be a conservative estimate, for several different reinforcement spacing. No tangential relative displacement was allowed

between rebar and concrete beyond the assumed debonded length.

The second finite element representation is a fictitious material representation. A low modulus fictitious material was placed within a prescribed crack region to account for the bonding stress between concrete and bar near a crack. The modulus of elasticity

131

of the fictitious material was calibrated based on the testing results at the crack stabilization stage. Different reinforcement spacing generated somewhat different moduli of elasticity for the fictitious material.

The debonded length FE representation was used to analyze an AASHTO slab strip model, due to the fact that a very large number of elements were required. The results indicated that the stress in the FRP reinforcement was relatively low. The

assumption that bond fatigue strength is the critical issue in FRP bridge deck slab design, instead of ultimate strength, was thus verified.

The fictitious material FE representation was used to analyze a complete bridge slab. The results indicated that the crack opening computed from a slab strip utilizing the fictitious material model was similar to the result from a complete bridge slab model, when diaphragms were included in the model. Top chord bars in the diaphragms were shown to be ineffective in reducing the predicted crack openings. The diaphragms, however, are instrumental in the serviceability of bridge slabs. By reducing the

differential deflections between girders, arching action within the deck slab between adjacent girders is enhanced. In the case of weak or absent diaphragms, excessive deflections of interior girders will generate one larger compression dome covering more than two girders, with much less effective arching action. Consequently, larger crack opening will appear.

132

The analysis results indicated that slab strip model and complete bridge slab model are similar in the estimation of maximum crack opening. Serviceability is deemed to be critical in ridge slab design instead of ultimate strength. A conservative empirical design method is quite often appropriate in the bridge deck slab design, since arching effect has typically been ignored. Reinforcement of 16M spaced at 150mm top and bottom in both directions appears to be satisfactory for girder spacing up to 3.6m. Both the serviceability (crack opening) requirements and the strength are met as illustrated in the analysis. reinforcement. The current strength based bridge slab design results in excessive

133

Chapter 6

Future Research

The experiments conducted in this research were limited to loads of constant amplitude at a fixed location. In reality, a bridge slab is under moving traffic loads of variable amplitudes. The crack growth under loads of variable amplitude requires

experimental investigation. The variable amplitudes may be quantified using root mean cube effective stress amplitudes. The Paris equation may have to be revised. Size effects under variable amplitude should be evaluated.

The experimental work has been limited to beams with FRP reinforcement. The experiments should be extended to actual FRP reinforced concrete slabs, supported by multiple girders, under moving loads. The complete empirical slab design algorithm deserves to be verified.

Although FRP concrete slab is corrosion resistant, there are many other environmental factors involved during the life span. They include seasonal temperature variation, water invasion, alkaline or acidic solutions and saline solutions. The durability of FRP reinforced slabs under these conditions is instrumental to its applicability in bridge deck infrastructures. Development of accelerated testing techniques for durability would be a breakthrough for FRP technology.

The variability of cracking and crack growth is evident in the experiments. Stochastic models may be developed to describe the variations. The portions of stress

134

intensity factors in the Paris equation may be randomized by adding noise.

The

variability of crack growth may be determined statistically. The distribution of crack opening growth may be predicted by the model. The model may be also extended to random loadings. In this study, the normal crack length has been used. The actually crack profile may be described by a fractal geometry, in which case a crack profile may be simulated.

In offshore structures, corrosion resistance is also one of the major concerns. Low weight and excellent fatigue performance are also the advantages in this environment. In a floating offshore platform, such as a tension leg platform (TLP), the topside load has to be supported, which includes the weight of the accommodation module and production facilities. Any reduction of the topside weight by using FRP composites will reduce the cost of supporting structure. TLP tendons may also be composed of FRP. The dynamic characteristics of FRP tendons under wave action should be studied.

FRP rods can be a direct substitute of steel reinforcement. Being composite also makes it possible to have different forms. Structure types may include different

configurations such as a sandwich type. The most promising structure modules should possess excellent load distribution, light weight and ease in construction.

135

Bibliography

1.

AASHTO LRFD bridge construction specifications, American Association of State Highway and Transportation Officials, 2004

2.

Standard Specifications for Highway Bridges, American Association of State Highway and Transportation Officials, 2000

3.

Fiber Reinforced Plastic Reinforcement, American Concrete Institute, Compilation 33, 1996

4.

440.1R-01: Guide for the Design and Construction of Concrete Reinforced with
FRP Bars, Committee 440, American Concrete Institute, 2001

5.

Considerations of Design of Concrete Structures Subjected to Fatigue Loading, ACI Committee 215, American Concrete Institute, 1992

6.

B. Benmokrane, O. Chaallal, R. Masmoudi, Flexural Response of Concrete


Beams Reinforced with FRP Reinforcing Bars, ACI Structural Journal, Vol.91,

No.2, 1995 7. M.H. Baluch, A.B. Qureshy, A.K. Azad Fatigue Crack Propagation in Plain
Concrete, Fracture of Concrete and Rock, S.P. Shah, S.E. Swartz., SEM-RILEM

International Conference, June 17-19, 1987, Houston, Texas, USA, pp. 80-87 8. C. Bakis, S. Al-Dulaijan, S. Nanni, A. Boothby, M. Al-Zahrani Effects of Cyclic
Loading on Bond Behavior of GFRP Rods Embedded in Concrete Beams,

Journal of Composite Technology & Research, JCTRER, Vol. 20, No. 1, January 1998, pp 29-37.

136

9.

B. Bakht, C. Lam Behavior of Transverse Confining Systems for Steel-Free


Deck Slabs, Journal of Bridge Engineering, Vol.5, No.2, May, 2000, pp 139-147.

10.

P. Balaguru, S.P. Shah A Method of Predicting Crack Widths and Deflections for Fatigue Loading, Fatigue of Concrete Structures, Publication SP-75, American Concrete Institute, Detroit, Editor S.P. Shah, 1982

11.

E. Cosenza, G. Manfredi, R. Realfonzo Behavior and Modeling of Bond of FRP


Rebars to Concrete, Journal of Composites for Construction, Vol.1, No.2, May

1997. pp 40-51. 12. T. Gergely, L. Lutz, Maximum Crack Width in Reinforced Concrete Flexural Members, Causes, Mechanism and Control of Cracking in Concrete, SP-20, R.E. Philleo, American Concrete Institute, pp.87-117. 13. John C. Graddy, Ned H. Burns, Richard E. Klingner, Factors Affecting the
Design Thickness of Bridge Slabs, FHWA/TX-95+1305-3F, Center for

Transportation Research, The University of Texas at Austin. 14. A. Katz Bond to Concrete of FRP Rebars after Cyclic Loading, Journal of Composites for Construction, Vol.4, No.3, August 2000, pp137-144. 15. H. Kupfer, H.K. Hilsdorf, H. Rusch Behavior of Concrete under Biaxial Stress, ACI Journal, Vol.66, August 1969, pp 659-673. 16. J. Larralde, R. Silva-Rodriguez Bond and Slip of FRP Rebars in Concrete, Journal of Materials in Civil Engineering, Vol.5, No.1, February, 1993. 17. M. Mabsout, K. Tarhini, G. Frederick, C. Tayar Finite Element Analysis of Steel
Girder Highway Bridges Journal of Bridge Engineering, Vol.2, No.3, August,

1997, pp 83-87.

137

18.

R. Nutt, T. Zokaie, R.A. Schamber Distribution of Wheel Loads on highway Bridges, NCHRP Project No. 12-26, National Cooperative Highway Research Program, Transportation Research Board, National research Council, Washington D.C., 1987

19.

P.C. Perdikaris, S. Beim, RC Bridge Decks under Pulsating and Moving Load, ASCE, Journal of Structural Engineering, Vol. 114, No.3, March 1988, pp. 591607.

20.

P.C. Perdikaris, S. Beim, S.N. Bousias Slab Continuity Effect on Ultimate and
Fatigue Strength of Reinforced Concrete Bridge Deck Models, ACI Structural

Journal, Vol. 86, No.4, July-August 1989, pp. 483-491. 21. P.C. Perdikaris, A.M. Calomino Kinetics of Crack Growth in Plain Concrete. Fracture of Concrete and Rock, S.P. Shah, S.E. Swartz., SEM-RILEM International Conference, June 17-19, 1987, Houston, Texas, USA, pp. 64-69 22. S.P. Shah, S.E. Swartz, C. Ouyang Fracture Mechanics of Concrete:
Applications of Fracture Mechanics to Concrete, Rock and Other Quasi-Brittle Materials, John Wiley & Sons, Inc., 1995

23.

C. Shield, C. French, A. Retika Thermal and Mechanical Fatigue Effects on


GFRP Rebar-Concrete Bond, Non-Metallic(FRP) reinforcement for Concrete

Structures, Proceedings of the Third International Symposium, Vol. 2, Oct. 1997. pp 381-388. 24. S.E. Swartz, C.M. Huang, K.K. Hu Crack Growth and Fracture in Plain
Concrete Static Versus Fatigue Loading, Fatigue of Concrete Structures,

Publication SP-75, American Concrete Institute, Detroit, Editor S.P. Shah, 1982

138

25.

S.E. Swartz, C.G. Go Validity of Compliance Calibration to Cracked Concrete


Beams in Bending, Experimental Mechanics, Vol.24, No.2, June 1984, pp129-

134. 26. S.E. Swartz, K.K. Hu, M. Fartash, C. Huang Stress-Intensity Factor for Plain
Concrete in Bending-Prenotched versus Precracked Beams , Experimental

Mechanics, Vol.22, No.11, November 1982, pp.412-417. 27. J. Yost Structural Reinforcement of Bridge Decks Using Rigid FRP Grids, Proceedings from the Sixth International Conference on Short and Medium Span bridges, Vancouver, Canada, edited by P.H. Brett, N. Banthia and P.G. Buckland. Montreal, Canada, 2002. 28. P.C. Paris, F. Erdogan A Critical Analysis of Crack Propagation Laws, Transactions of ASME, Journal of Basic Engineering, Vol.85, pp.528-534, 1963 29. H. Tada, P. Paris, G. Irwin The Stress Analysis of Cracks Handbook, Second Edition, Paris Productions Inc., 1985 30. P.V. Vijay, Hota V.S. GangaRao, Bending Behavior and Deformability of
Glass-Fiber-Reinforced Polymer Reinforced Concrete Members, ACI Structural

Journal, Vol.98, No.6, 2001

139

Anda mungkin juga menyukai