Вы находитесь на странице: 1из 9

Engineering Structures 40 (2012) 267275

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Nonlinear nite element modeling of cracking at ends of pretensioned bridge girders


Pinar Okumus a,, Michael G. Oliva b, Scot Becker c
a

Department of Civil and Environmental Engineering, University of Wisconsin Madison, 1415 Engineering Drive, 1203 Engineering Hall, Madison, WI 53706, United States Department of Civil and Environmental Engineering, University of Wisconsin Madison, 1415 Engineering Drive, 1214 Engineering Hall, Madison, WI 53706, United States c Wisconsin Department of Transportation, 4802 Sheboygan Ave., Madison, WI 53707, United States
b

a r t i c l e

i n f o

a b s t r a c t
Recent bridge designs have created efcient prestressed concrete girder sections with thin webs, and high levels of prestress. The transfer of the large stresses from strands to concrete causes these slender sections to undergo cracking at the ends of the girders. Due to the large amount of cracking, a nonlinear analysis is necessary to reveal and understand the behavior of the concrete and reinforcement bars at prestress release. Finite element modeling is an excellent tool to perform this task. The accuracy of the analyses, however, depends on the input parameters, some of which are challenging to dene for a nonlinear problem. This paper identies the input parameters and modeling features that have signicant impact on the results of nonlinear nite element analyses for pretressed concrete girder ends. The sensitivity of the results to the tensile properties of concrete, the bond properties between concrete and steel rebars and the form of prestress transfer to the concrete were evaluated. The impact of modeling related properties such as the order of elements on the predicted results was also investigated. Available test data was used to verify the modeling techniques. Once veried, the nite element modeling was extended to girders where signicant cracking is observed. The full eld tensile strain patterns obtained through the veried nite element models are used to explain observed cracking. The effectiveness of the end reinforcement bars, intended to control cracking, was examined. 2012 Elsevier Ltd. All rights reserved.

Article history: Received 23 September 2011 Revised 22 December 2011 Accepted 16 February 2012 Available online 29 March 2012 Keywords: Nonlinear nite element analysis Prestressed girder End cracks Concrete plasticity Prestress transfer

1. Introduction The fundamental concept behind prestressed concrete design is the transfer of compression force from steel prestressing strands to surrounding concrete to achieve a crack free member under service loads. The properties of concrete and strand interface cause this transfer of stress to occur within a relatively short distance from the girder end, with high stress conditions. As a result, an overstressed girder end region may exhibit visible cracks in distinctive patterns. This contradicts the crack free goal of prestressed concrete. Heavily prestressed deep girders designed for long bridge spans appear to experience more severe end cracks than traditional smaller girders. Characteristic cracks observed in very similar patterns in all deep Wisconsin wide anged I girders are shown in Fig. 1 right after prestress release. These cracks might lead to durability problems when they are not enclosed in the end diaphragms and exposed to the environment. If the salt water seeping through the cracks reaches strands, corrosion and loss of bond between strands and concrete may lead to structural capacity losses.

Corresponding author. Tel.: +1 608 262 1262.


E-mail addresses: okumus@wisc.edu (P. Okumus), oliva@engr.wisc.edu (M.G. Oliva), scot.becker@dot.wi.gov (S. Becker). 0141-0296/$ - see front matter 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.engstruct.2012.02.033

Endeavors to explain the stress eld at the girder ends, since the early 1960s, utilized either empirical methods, simplied linear analytical concepts, or strut and tie models. The cracks, to this day, remain an issue despite the girder end design provisions of codes such as the AASHTO LRFD Bridge Design Specications [1]. Prompted by the lack of comprehensive nonlinear analyses of the prestressed girder end regions in the literature, this study employed nonlinear nite element analysis (FEA) to assess the likelihood of cracking and the stress distribution at the girder ends during and shortly after prestress release. Among the most widely known girder end analytical studies is the so called GergelySozen model [2]. Assuming concrete behaves linearly until cracking, the model formulates the area of web reinforcement required to control crack size using free body analysis. Other analytical research [3,4] investigated the strut and tie method for the design of the girder end region, some specically for posttensioned girders [5]. The strut and tie method is most easily applicable after the material becomes inelastic, provides a lower bound estimate of the strength capacity, and there is more than a single truss conguration for a given case. The majority of the previous research work consists of empirical or semi-empirical studies. The AASHTO LRFD Bridge Design Specications, Article 5.10.10.1 [1] seems to have adopted the equation derived from the experimental study conducted by Marshall and Mattock [6]. Tuan et al. [7] and Dunkman et al. [8] monitored

268

P. Okumus et al. / Engineering Structures 40 (2012) 267275

for the extended FEA study on Wisconsin girders with signicant cracking. The full eld strain distribution at the girder end region and girder end zone reinforcement right after the prestress release is presented. The stresses obtained through FEA are used to explain the causes of characteristic cracks. The FEA was performed using Abaqus/CAE. 3. Simulated girders The base line for the verication of the FEA procedures was a Texas 1778 mm deep I girder tested by OCallaghan [9]. This girder was selected for the verication and sensitivity study due to the large amount of test data reported. OCallaghan placed strain gages on selected rebars at 419 mm, 811 mm, 1575 mm from the bottom of the girder. These lines of gages will be referred in this paper as gage line I, II and III respectively. The stresses on rebars at the gage locations were reported. In addition information on the distribution of the bond stresses was also available through gages on selected strands in the end region. This girder had welded deformed rebar cage of 517 MPa average yield strength and additional four rebars of 414 MPa average yield strength bundled at 102 mm spacing closest to the end. The average concrete strength was 46 MPa. The girder had 46 low relaxation 1862 MPa ultimate strength 15.2 mm diameter strands in the bottom ange at an eccentricity of 582 mm, in addition to four straight dummy strands on the top ange to facilitate construction. The FEA study with conrmed modeling techniques and parameters were then extended to resemble Wisconsin girders which exhibit the most signicant cracking. The 1372, 1829 and 2083 mm deep wide ange I girders represented heavily prestressed girders with 40, 48, 46 strands, 8 of each were draped. Strain patterns are demonstrated for a 1372 mm deep girder. The release strength of girder concrete studied was 47 MPa, the rebars had yield strength of 414 MPa. The low relaxation strands were 15.2 mm diameter and had 1862 MPa ultimate strength. The cross section, strand patterns and end zone rebar details of these sections are given by the Wisconsin Department of Transportation, Bureau of Structures [10]. Only a quarter segment of the full girder is modeled utilizing the symmetric nature of the geometry and loading. Fig. 2 shows a typical meshed model with boundary conditions. 4. FEA sensitivity study 4.1. Material nonlinearity 2. The scope The aim of this study was to achieve an accurate nonlinear nite element prediction of strain and stresses in pretensioned concrete girder ends. Due to nonlinearities in a girder end zone, particularly stiffness changes and redistribution of strains after cracking, explaining the strain distribution after the rst cracks is beyond the capabilities of linear analyses or free body diagrams. A nonlinear stress analysis of the crack prone pretensioned girder ends does not appear in the literature. The accuracy of the FEA depends heavily on the input parameters. The tensile properties of concrete after cracking, the concrete steel bond, and the concretestrand bond are the important parameters that dene the cracking behavior. The sensitivity of the FEA results to varying values of these parameters was investigated. The role of the analysis type, and the element type on the efciency and convergence of the models was studied. The results of each model were compared to available test results. The best representations of the input and modeling parameters were used Material nonlinearity, particularly for concrete in tension, was a key part of this cracking focused problem. Linear FEA studies exist in the literature. In order to assess the level of accuracy loss with linear models, a linear FEA was run and the results were compared to the ones with nonlinear properties. From a computational perspective, the FEA with the linear elastic concrete material model was much simpler and cheaper. The two cases behave similarly until the concrete elements reached their theoretical cracking strength. Once the cracking occurs in concrete, the redistribution of stresses is expected to take place and the rebars become engaged. Elastic material models which do not capture the stiffness loss of the concrete elements during cracking are not capable of representing this and subsequent crack growth. Fig. 3 presents the predicted principal tensile strains for a 2083 mm deep Wisconsin girder after complete prestress release with and without concrete material nonlinearity. A qualitative comparison between actual cracking in Fig. 1 and the two models

Fig. 1. Typical characteristic inclined, web, and Y cracks.

strains on the vertical end zone stirrups. These strain values gave rise to practical recommendations on the end zone reinforcement efciency. However, strain data collected by strain gages is bound to be discrete and provides limited information, particularly when cracking occurs away from a gage location. FEA can provide the strain eld for the entire continuum in any direction; can help explain what triggers cracking and can also be used for parametric studies with minimal cost. The experimental study conducted by OCallaghan [9] measured numerous reinforcement bar strains at three different vertical levels along the girder length in the end region on Texas girders. The variation of the stresses the rebars carry along the girder length was presented. Kannel et al. [11] utilized FEA to simulate the restraint provided by the uncut strands, the effects of the order in which the strands are cut, and debonding some strands at the girder end. The main shortcoming of this model was the use of linear elastic material properties for concrete. Breen et al. [5] simulated a posttensioned girder with two dimensional FEA, where the stress transfer mechanism is concentrated at the girder end and much simpler compared to the gradual stress transfer in pretensioned girders.

P. Okumus et al. / Engineering Structures 40 (2012) 267275

269

Fig. 2. The boundary conditions and the mesh.

Fig. 3. Principal tensile strain comparison of a linear and a nonlinear FEA model.

in Fig. 3 shows the alignment of cracks with the high strain areas in the nonlinear model. The linear model causes the concrete elements maintain their initial stiffness, with maximum tensile strain and stresses up to 314 microstrains and 1545 psi at the web. The cracking strain and strength for this concrete was calculated to be 124 microstrains and 600 psi. The unrealistic excessive strain and stress values prove that linear models are not suitable to simulate prestressed girder ends. For the nonlinear model, the magnitudes of maximum tensile stress did not exceed the cracking stress. Much higher tensile strains were obtained, magnitudes reaching 0.38 mm when converted to crack opening. The highest strains

were in the inclined cracking region, matching the location where the largest cracks are observed in the eld. It was not possible to obtain rebar stresses within an acceptable error range with the linear model. The maximum rebar stress with a linear model was predicted to be 51.7 MPa for the web bar closest to the end. On the other hand, the rebar stress for the same bar was 193.1 MPa with a nonlinear model. In order to decrease the computational cost, nonlinear material properties were only assigned in the end region. The length of the region with nonlinear material properties was at least as long as the girder depth from the girder end. Nonlinear regions shorter than this resulted in signicant errors in the results.

270

P. Okumus et al. / Engineering Structures 40 (2012) 267275

Fig. 4. Constitutive model for concrete in compression (left) and tension (right).

4.2. Concrete material model


160

The material model used was the concrete damaged plasticity model from the Abaqus/CAE material library. The constitutive model of concrete was based on FIB Model Code 2010 [12], and AASHTO LRFD Bridge Design Specications [1] as shown in Fig. 4. The initial modulus of elasticity is assumed to be the same for tension and compression and calculated per AASHTO LRFD Bridge Design Specications Section 5.2.4.2.The cracking strength of concrete is calculated using the AASHTO LRFD Bridge Design Specications Section C5.4.2.7. This was judged to be a less conservative but more accurate representation than the cracking strength given by the FIB Model Code 2010, Volume 1, Section 5.1.5 based on the FEA results. For the nonlinear range, the constitutive model for concrete in tension and compression were based on the mathematical models given by FIB Model Code 2010, Volume 1, Section 5.1.8. The concrete tension behavior is composed of two discrete parts: the pre cracking and post cracking stages. The post cracking stage can be dened in terms of strains, crack opening or the fracture energy. Mesh sensitivity is a potential issue due to narrower crack widths with ner meshes [13]. A fracture energy based concept was used to dene the stressdisplacement (crack opening) behavior as opposed to a stressstrain relation in order to overcome this problem.

120

rebar (MPa) 80
40

Test by O'Callaghan [9] FEA, GF = 147 N/m FEA, GF = 115 N/m FEA, GF = 2x147 N/m

0 0 500 1000 1500

Distance from the girder end (mm)


Fig. 5. The resultant rebar stresses along the girder length with varying levels of GF.

conditions and the size of the specimen [12]. In the absence of tests, the fracture energy in N/m is given to be estimated by Eq. (1), where fcm is the mean compressive strength of concrete in MPa.
0:18 GF 73 fcm

4.3. Rebarconcrete bond properties Rebars are modeled as linear elastic elements, since end zone rebars are not expected to yield during strand detensioning. The interaction of the reinforcement bars and concrete was modeled through tension softening of the concrete implicitly. Tension stiffening, the term used by Abaqus, is introduced to concrete to represent the added ductility that would be provided by the rebars after cracking. The rebar elements were embedded in concrete. In Abaqus, this means that the response of the concrete elements is used to constrain the translational degrees of freedom of the rebar nodes. Once concrete elements reach their cracking limit, their stress carrying capacity drops with increasing deformation, transferring their force to the steel rebars. Reinforcement bar elements distribute the failure and softening to address any mesh sensitivity problems with nonlinear analyses. Tension softening is directly related to the fracture energy, GF as shown in Fig. 4. FIB Model Code 2010 Section 5.1.5.2 states the fracture energy should be determined by related tests. The fracture energy depends on the water cement ratio, the maximum aggregate size, and the age of concrete and is affected by the curing

This equation calculates the fracture energy for the base girder tested by OCallaghan [9] as 145 N/m. The earlier version of the Model Code, CEB-FIB Model Code 1990 [14] estimates the fracture energy for the same strength concrete to be between 70 N/m and 115 N/m depending on the maximum aggregate size. Due to the number of parameters that affect the fracture energy measurements and the lack of uniaxial tension test data, the sensitivity of the rebar stresses to varying values of fracture energy needed to be investigated. The results of FEA with varying levels of fracture energy were compared. GF = 115 N/m given as the upper bound by the Model Code 1990 [14], GF = 145 N/m as given by the Model Code 2010 [12], and GF = 2 145 N/m or double the FIB 2010 Model Code value. The results as compared to the measured rebar stresses at gage line I of the base girder are shown in Fig. 5. It is seen that the fracture energy of 145 N/m as prescribed by the FIB Model Code 2010 is a good approximation that created only 7% error in the peak rebar stresses. Using the material model provided by the CEB-FIB Model Code 1990 introduced a 15% error in the peak values. Using two times the fracture energy given by the Model Code resulted in 13% error in the peak values, however converged much faster. It was observed that FEA with larger fracture energy levels converged signicantly faster. Fracture energy values below 100 N/m did not return converging results and therefore were judged not to be practical.

P. Okumus et al. / Engineering Structures 40 (2012) 267275

271

4.4. Prestress transfer The cracking in Wisconsin girders typically occurs during or right after the prestress release. Therefore the main loading considered for the FEA was prestressing forces on the girder. Some minor widening of cracks is observed weeks after the girders are cast; however the scope of this study only covered the immediate cracking upon release. Temperature loads were not included in the models as the temperature changes during prestress release are minimal. Some amplication of the stresses due to the sudden cut of the strands is expected. However, studying the dynamic effects of the strand cutting process or the response of concrete under sudden loads was not within the scope of this project. Modeling the exact condition between strand and concrete as controlled by the adhesion, friction due to the twisted strand conguration and Poisson radial expansion of the strand signicantly increases the analysis cost. Modeling the entire girder end region would not have been possible with such an approach. Rather, the prestress load was applied directly on the concrete as surface loads around the diameter of the strands along the prestress transfer length as shown in Fig. 2. The loading from every two strands was applied to the girder on a separate step to identify the stresses individual strands created. The research results on the strand transfer length and bond stress distribution in the literature are rather scattered as stated by Buckner [15], and Tabatabai and Dickson [16]. Therefore the three different transfer length and bond stress distributions were studied to pick the best representation for the problem. The rst model used the transfer length calculated by 60 times the strand diameter with bond stresses assumed uniform per AASHTO LRFD Bridge Design Specications, Section C.5.11.4.2 [1]. The transfer length for the 15.2 mm in strands was calculated as 914 mm. The second model assumed a linear bond stress distribution over 914 mm long transfer length. The third model used the bond stresses and transfer length as measured through the tests by OCallaghan [9] via strain gages on strands. The results from three models are compared to the test results at gage line I in Fig. 6. As expected, the FEA results aligned with the test results when the transfer length as measured through the tests was used. The linearly varying bond stresses over a 914 mm transfer length gave results similar to the test results with a total average error of 15%. The model with uniform strandconcrete bond stress over the 914 mm as given in AASHTO LRFD [1] resulted in poor prediction of rebar stresses. The distribution of the bond stresses plays an important role in the distribution of the girder end stresses.

4.5. Element type The sensitivity of the results with respect to the element type used in the nonlinear girder end region was investigated. The region was meshed with rst and second order elements of tetrahedral shapes. The results for the 1st order tetrahedral elements (C3D4) and 2nd order tetrahedral elements (C3D10) were compared to the test results provided by OCallaghan [9] at gage line I in Fig. 7. The numbers in parenthesis indicate the name of the elements in ABAQUS. The hexahedral elements were not used as they are prone to shear locking. It is also more difcult to mesh arbitrary volumes with hexahedral elements. The results of the analysis with 1st order tetrahedral elements had the best correlation with the test results. The Abaqus [13] suggests that 2nd order elements generally give higher accuracy per degree of freedom for the solution for elasticity problems. However, for plasticity applications, it advises that the rst order elements are in general preferred for accuracy. The results of this study, verify this recommendation. The analysis time for the analysis with 2nd order elements was in an order of magnitude longer compared to that with 1st order elements. The use of 2nd order elements was found infeasible. The steel rebars were modeled using one dimensional truss elements, the order of which was compatible with the surrounding three dimensional elements representing concrete (T3D2 for rst order analysis and T3D3 for second order analysis). The steel prestressing strands were not included in the model as mentioned earlier in the previous section. The smallest dimension of the elements used was 38.1 mm or less based on a balance between relative error and computation cost. The element size was gradually increased away from the girder end. 4.6. Additional verication Based on the sensitivity studies described above, and using what proved to be the best modeling parameters, the correlation of the analysis results for gage line II and gage line III with the test data reported by OCallaghan is as shown in Fig. 8. The single high rebar stress spike from OCallaghans data at one gage location on gage line II is judged to be either an error in gage data or due to a gage location at a concrete crack. At other locations the FEA was found to provide an accurate representation of the response. 5. Results extended to Wisconsin girders Once the modeling techniques were veried using available test data on a Texas girder, and the most suitable input parameters and modeling techniques were selected; the FEA was extended to represent the standard Wisconsin girders. The full eld strains on concrete and stresses in rebars are presented here for the Wisconsin 1372 mm deep girder. Wisconsin girders of three different depths have the same bottom and top ange, are often heavily prestressed and experience very similar end cracking patterns. Therefore, the strain patterns shown here are representative of all Wisconsin I girders. 5.1. Reactions in concrete Of particular importance to this research were the principal components of strains. The principal tensile strains and their directions determine where and in which direction the concrete cracking would occur. After the prestressing force was transferred to the concrete, the magnitudes of the principal tensile strains in the concrete are as shown in Fig. 9. Positive strains indicate tension.

160

Test by O'Callaghan [9] FEA, uniform bond FEA, linearly varying bond FEA, bond as measured [9]

120

rebar (MPa)

80

40

0 0 500 1000 1500

Distance from the girder end (mm)


Fig. 6. The resultant rebar stresses along the girder length with varying prestress transfer rates.

272

P. Okumus et al. / Engineering Structures 40 (2012) 267275

160

120

Test by O'Callaghan [9] FEA, 1st order tetra FEA, 2nd order tetra

rebar (MPa) 80
40

0 0 500 1000 1500

Distance from the girder end (mm)


Fig. 7. The resultant rebar stresses along the girder length with varying element order.

100 Test by O'Callaghan [9] FEA 60

100 Test by O'Callaghan [9]

FEA
60

rebar (MPa)
20

rebar (MPa)
20

-20 0 500 1000 1500

-20 0 500 1000 1500

Distance from the girder end (mm)

Distance from the girder end (mm)

Fig. 8. FEA results compared to the test results by OCallaghan [9] for gage line II (left) and III (right).

Fig. 9. Principal tensile strain contours.

P. Okumus et al. / Engineering Structures 40 (2012) 267275

273

Fig. 10. Principal strain directions in elevation on Section AA.

The principal strain direction plots as given in Fig. 10 for compression and tension can be used as a supplement to the contour plots to explain cracking. The direction of the principal tensile strains, indicated by the short lines in the gures, determines the direction of crack opening. The compression principal strains are concentrated around the strands where compression struts spread out into the girder as prestressing is transferred along the girder length. The four regions of the girder where principal tensile strains were signicantly higher than average strains are marked in Figs. 9 and 10. Region I: The highest concrete tension strains occur at region I for girders with draped strands. Region I is close to the draped strands where the inclined cracks are often visible as indicated in Fig. 1. The FEA shows that the principal tensile strain direction is perpendicular to inclined cracking even without the draped strands. Comparison of FEA models with and without draped strands, however, showed that draped strands play a signicant role in

amplifying these strains. The radial tensile strains perpendicular to the compression strut around the draped strands align with the principal tensile strains created by the straight strands. Therefore, the draped strands trigger the inclined crack formation. Region II: The second largest strains are located in this web region. The locations and directions of these high strains are representative of the horizontal web cracking shown in Fig. 1. These cracks develop across the tension tie that ties the two compression struts formed along the draped and straight strands. These cracks are attributed to the eccentricity of the strands in the direction of the girder depth. The draped and straight strands create moments in the web in opposite directions. These two reaction moments put the web into tension. The girder web then experiences cracking under heavy loads. Region III: Region III is located in the bottom ange where Y shaped cracks are observed. These cracks form a separation line between two sides of the bottom ange and web. They are close to the strands and may form paths for corrosion agents to reach the

Fig. 11. Principal tensile strain directions on the cross section of the bottom ange.

274

P. Okumus et al. / Engineering Structures 40 (2012) 267275

strands, raising the risk of structural capacity loss. The principal tensile strain directions as obtained through FEA, shown in Fig. 11, agree with a Y shaped crack. The vertical crack seems to be a result of the eccentricity of the strands in the bottom ange across the width of the girder. Conceptually similar to web cracking, strands on each side of the bottom ange create moments in opposite directions on vertical sections. The Y shape forms when this vertical crack meets the lower horizontal web cracks in a transition zone between the web and the bottom ange. Region IV: Region IV marks high strains in the bottom ange area peaking at a distance equal to the transfer length from the girder end. Fig. 11 shows the principal tensile strains on a cut from the bottom ange at the transfer length. These strains, sometimes referred as the bursting stresses, form perpendicular to the compression force applied by the bottom ange strands, and suggest outwards radial pressure. Although the FEA indicated that these strains exceed the elastic limit, no cracking in this region was visible on girders inspected at plants. The lack of visible cracking may be due to the fact that the predicted directions of the tensile strains do not align, making formation of a single crack unlikely. The bottom ange of the real girders might be experiencing numerous interior micro cracks that are not visible during inspection. 5.2. Reactions in reinforcement bars Determining the stresses in the rebars and in the surrounding concrete is essential in evaluating the efciency of the bars in resisting crack opening. The axial stresses in the web and bottom ange rebars in the end zone of the girder are shown by the line color in Fig. 12 on a quarter size model. The vertical bars closest to the very end of the girder, crossing the web cracks, exhibit the largest stresses. On the contrary, the stresses in the bottom ange connement stirrup start lower at the girder end and peak closer to the end of the transfer length. The web reinforcement bars lay parallel to principal tensile strains in the web cracking region as shown in Fig. 10, and therefore they will be the most effective in restraining these cracks. The stresses, obtained through FEA, in vertical web reinforcement bars for heavily prestressed girders do not exceed 106.2 MPa,

182.7 MPa, and 199.3 MPa for 1372 mm, 1829 mm, 2083 mm deep heavily prestressed Wisconsin girders. The web reinforcement bars are only loaded up to 2550% of their yielding capacities. 6. Conclusions This paper describes the modeling procedures and input parameters that lead to accurate results for nonlinear FEA results in pretensioned girder ends with cracks during prestress release. The strain and stress elds of concrete and rebars were presented to explain the reasons behind cracking. Linear models, even though they are computationally low cost options, do not consider the stiffness loss of concrete upon cracking and largely under calculate the concrete strains. They fail to resemble the stress transfer from concrete and rebars due to cracking and therefore underestimate the rebar stresses. Underestimating the fracture energy as an input introduces convergence problems. A 1020% error in the rebar stresses is expected when the fracture energy, as a concrete tensile property, is mispredicted. The fracture energy calculated using the equation in the FIB Model Code 2010 lead to accurate results. Representing the concretestrand bond as uniform, linearly distributed, or as measured by tests has a direct impact on the predicted stress results for prestressed girder ends. The model with uniform concretestrand bond stresses, per AASHTO LRFD Bridge Design Specications, predicted the peak rebar stresses lower than the test values. Using linearly varying bond stresses along the strand transfer length is judged to be an acceptable substitution for the complex modeling of actual friction and bond effects for the test girder where the bond distribution measured was similar to linear varying. The computational cost for use of second order concrete elements was very high and did not improve the predicted stresses. Therefore, these elements are not recommended for similar nonlinear plastic problems. The locations and directions of the plastic strains predicted by the FEA models coincide with the locations and crack opening directions of the cracks observed on the girders. The inclined cracks closer to the top ange are attributed to the contribution of strains created by draped strands to the strains in this area caused by the

Fig. 12. Longitudinal stresses (MPa) in the reinforcement bars.

P. Okumus et al. / Engineering Structures 40 (2012) 267275

275

straight strands. The web cracks were attributed to the eccentricity of the bottom and draped strands over the girder depth. The Y cracks in the bottom ange are similarly a result of the transverse eccentricity of the bottom ange strands on each side of the ange over the width of the girder. Finally, the FEA models captured high strains growing along the bottom strands as the compression strut spreads out. The highest rebar stresses were observed in the web, and are less than half of the yielding strength. Based on comparisons with data measured in girder tests in Texas, the FEA modeling techniques described here give a good representation of the complex stress and strain patterns at the ends of girders where prestress transfer effects cause nonlinear behavior with concrete cracking. These modeling procedures can be effectively used in the future to investigate methods for eliminating or controlling the amount and size of concrete cracks in prestressed girders during the prestress transfer process. Acknowledgements This work was funded by the Wisconsin Highway Research Program of the Wisconsin Department of Transportation. The views presented are those of the authors and not the funding agency. References
[1] American Association of State Highway and Transportation Ofcials. AASHTO LRFD Bridge Design Specications. Customary US Units, with 2008 interim revisions; 2008. [2] Gergely P, Sozen MA. Design of anchorage-zone reinforcement in prestressed concrete beams. PCI J 1967;12:6375.

[3] Crispino ED, Cousins TE, Roberts-Wollmann CL. anchorage zone design for pretensioned precast bulb-T bridge girders in Virginia. Charlottesville (VA): Virginia Transportation Research Council; 2009 June. Report No: FHWA/VTRC 09-CR15. Contract No: 81176. Sponsored by the Virginia Department of Transportation, Federal Highway Administration. [4] Castrodale RW, Lui A, White CD. Simplied analysis of web splitting in pretensioned concrete girders. In: Proceedings of the PCI/FHWA/NCBC concrete bridge conference, Nashville (TN), 2002 October 69. [5] Breen JE, Brdet O, Roberts C, Sanders D, Wollmann G, Falconer B. Anchorage zone reinforcement for post-tensioned concrete girders. Washington (DC): National Cooperative Highway Research Program, Transportation Research Board; 1994. NCHRP Report No: 356. [6] Marshall WT, Mattock AH. Control of horizontal cracking in the ends of pretensioned prestressed concrete girders. PCI J 1962;7:5674. [7] Tuan CY, Yehia SA, Jongpitaksseel N, Tadros MK. End zone reinforcement for pretensioned concrete girders. PCI J 2004;49:6882. [8] Dunkman DA, Hovell CG, Moore AM, Avendano A, Bayrak O, Jirsa JO. Bursting and spalling in pretensioned concrete beams. In: Proceedings of the 2010 international FIB congress and PCI annual convention/exhibition, Washington, DC, 2010 May 29June 2. [9] OCallaghan MR. Tensile stresses in the end regions of pretensioned I-beams at release. Dissertation. Austin (TX): University of Texas, Austin; 2007. [10] Standard Detail Drawings. Bureau of Structures, Wisconsin Department of Transportation. <http://on.dot.wi.gov/dtid_bos/extranet/structures/LRFD/ standards.htm> [updated 31.10.11; cited 31.10.11]. [11] Kannel J, French C, Stolarski H. Release methodology of strands to reduce end cracking in pretensioned concrete girders. PCI J 1997;42:4254. [12] Fdration Internationale du Bton. Fib Bulletins 5556: Model Code 2010 First complete draft. Fdration Internationale du Bton; 2010. [13] Dassault Systmes Simulia Corporation. Abaqus theory manual. In: Abaqus 6.9 documentation. RI (USA): Dassault Systmes; 2009. [14] Comit Euro-International du Bton. CEB-FIP Model Code 1990: design code: London: T. Telford; 1993. [15] Buckner CD. A review of strand development length for pretensioned concrete members. PCI J 1995;40:8499. [16] Tabatabai H, Dickson TJ. The history of the prestressing strand development length equation. PCI J 1993;38:6475.

Вам также может понравиться