Experimental and Analytical Forensic Investigation of Bridge Timber Piles under Eccentric Loads

Transcription

1 98 Experimental and Analytical Forensic Investigation of Bridge Timber Piles under Eccentric Loads Daniel J. Borello 1, Bassem Andrawes 2, Jerome F. Hajjar 3, Scott M. Olson 4 1 Newmark Civil Engineering Laboratory, Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, B112 Newmark Civil Engineering Lab, 205 North Matthews Ave, Urbana, IL, USA; PH (847) ; dborello@illinois.edu 2 Newmark Civil Engineering Laboratory, Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, 3122 Newmark Civil Engineering Lab, 205 North Matthews Ave, Urbana, IL, USA; PH (217) ; andrawes@illinois.edu 3 Newmark Civil Engineering Laboratory, Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, 2129b Newmark Civil Engineering Lab, 205 North Matthews Ave, Urbana, IL, USA; PH (217) ; jfhajjar@illinois.edu 4 Newmark Civil Engineering Laboratory, Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, 2230d Newmark Civil Engineering Lab, 205 North Matthews Ave, Urbana, IL, USA; PH (217) ; olsons@illinois.edu ABSTRACT This study examines the structural behavior of bridge timber piles under eccentric compression loading. A collapsed rural concrete bridge supported on timber piles was used as a prototype in this study. Samples of the collapsed piles were retrieved and experimentally tested under compression and combined compression and flexure. The experimental timber pile response was used to calibrate an numerical model of the full timber piles of the prototype bridge, including material and geometric nonlinearity as well as soil-structure interaction. The numerical results illustrated that the pile strength was significantly reduced under eccentric load compared to concentric load. It was concluded that the effect of compression-flexure interaction on bridge timber piles must be checked during design, especially in the case of simply-supported superstructures where loading on one span may lead to eccentric loading on a timber pile group. INTRODUCTION Timber piles have been commonly used as a foundation system since the middle of the 18th century (Timber Pile Council, 2002). The interaction of combined compression and flexure loads in wood was first investigated in 1924 by the Army and Navy for airplane construction (Newlin and Trayer, 1956). Experimental testing and analysis based on Euler buckling illustrated that the material capacity was a function of load eccentricity as well as the slenderness of the section. In 1982, Zahn

2 99 tested specimens under varying eccentricities. Over 400 tests were performed on long specimens in order to calibrate a finite element analysis for longer specimens. Zahn derived interaction equations to include second-order effects coupling the axial force and bending moment (Zahn, 1988). These equations reasonably captured experimental results and were adopted by current National Design Specifications (AFPA, 2005). Previous experimental testing of wood under eccentric loading has concentrated on small sawn lumber, with limited research on large cylindrical specimens typical of timber piles. As part of a forensic investigation of a bridge collapse, the authors conducted an experimental and computational investigation of bridge timber piles under eccentric loads. Large cylindrical samples for experimental testing were obtained from the collapsed structure and the structure was used as a prototype structure to extend the experimental results to a numerical model of the full length pile. PROTOTYPE STRUCTURE The forensic investigation was motivated by the collapse of a county bridge shown in Figure 1, which was used as a prototype for the study on effect of eccentricity in timber piles. The bridge was constructed in 1976 to service local agricultural traffic based on 1973 AASHTO HS-20 load criteria (AASHTO, 1973), which did not explicitly account for potential load eccentricity. However, since the deck was simply supported, the foundation was susceptible to eccentric loads under unsymmetrical loading conditions such as that shown in Figure 2. The prototype bridge (Figure 1) consisted of three 12.8 m (42 ft) spans that traversed a small stream that flowed between bents 2 and 3 (the bents and piles were numbered increasing to the east and south respectively). The deck was 432 mm (17 in) thick precast pretensioned concrete beams supported on concrete pile caps connected by 19 mm (3/4 in) dowels cast in the pile cap and inserted into 51 mm (2 in) holes in the deck beam. This connection provided negligible rotational restraint between the deck beams and pile cap. The timber piles were embedded 305 mm (12 in) in the cast-in-place pile cap, providing the ability to transfer axial force and moment (Figure 3 and Figure 4). The circular timber piles each had a 254 mm (10 in.) nominal diameter and were 8.5 m (28 ft) long, embedded approximately 5.2 m (17 ft) below the stream mudline. The timber piles were cut from oak, although the specific species of timber was not specified in the design documents. The original bridge plans required the timber piles to have a 214 kn (48 kip) nominal capacity. As part of the regular maintenance of the bridge, in 2000 the top 2 m (6 ft) of two piles of bent 3 were removed due to severe deterioration and replaced by new round timber posts.

3 100 Figure 1. Prototype bridge (photo looking south) Figure 2. Unsymmetrical load on bridge Bent 3 failed in 2008, collapsing the east and middle spans. The pile cap of bent 3 rested in the riverbed. The west span between bents 1 and 2 remained standing. The pile to pile cap connection of bent 3 remained intact after the collapse, maintaining its connection capable of transferring axial force and moment (Figure 4). The piles were fractured longitudinally approximately 1 to 2 m (3 to 6 ft) below the pile cap. Since the deck was not continuous at the pile cap there was a large rotation between the two deck segments, cracking the asphalt surface (Figure 1). A geotechnical investigation ruled out geotechnical causes of the collapse (Borello et. al., 2009). The superstructure was also mainly intact, indicating that the pile foundation most likely initiated the collapse.

4 101 Figure 3. Timber pile foundation (Bent 2) Figure 4. Typical pile-to-pile cap connection after collapse After the collapse, several specimens of the bridge piles were retrieved. These specimens were tested at the University of Illinois under compression and combined compress and flexure. These tests results were then extended numerically to predict the strength of a single pile. EXPERIMENTAL TESTING Table 1 summarizes the properties of eight experimental specimens. Six specimens were retrieved from three piles of Bent 3 and two of the specimens were cut from Bent 2. The specimens were designated as being either above or below the riverbed or from one of the two posts inserted into the piles from previous repairs. The specimens were each cut to a length of 914 mm (36 in). The specimens were tested at two moisture content conditions, air-dried (under laboratory conditions) and water-saturated (submerged in water until their weight stabilized), to evaluate the influence of moisture content on the pile mechanical properties. The specimens were subjected to either monotonic or cyclic compression loading or monotonic combined compression and flexure. The predicted strength of the specimens was calculated per the National Design Specification (NDS) for Wood Construction Section 6, Round Timber Poles and Piles (AFPA, 2005) utilizing Load and Resistance Factor Design (LRFD) procedures without the use of a resistance factor. The resulting "cross-section strength" was 18.3 MPa (2640 psi). Based on the NDS column stability curve, the compressive strength reduction due to flexural buckling was negligible for the experimental specimens. Tests were conducted per ASTM D198 (ASTM, 2005) on a 2.7 MN (600 kip) MTS uniaxial hydraulic frame. For compression tests, a spherical head was placed below the specimen to prevent unintentional loading eccentricities, as shown in Figure 5. The monotonic tests were initially conducted under displacement control to provide a wood fiber strain rate of mm/mm per minute, which was increased after substantial post-peak softening.

5 102 Table 1. Test Specimen Matrix Elevation c Moisture Specimen Bent Pile Test Type Loading Number a Number b Content Plan SP1 3 1 Post Air-dried Compression Monotonic SP2 2 4 Above Air-dried Compression Monotonic SP3 2 4 Above Saturated Compression- Flexure Monotonic SP4 3 4 Above Air-dried Compression Monotonic SP5 3 1 Post Air-dried Compression Cyclic SP6 3 1 Below Saturated Compression Monotonic SP7 3 1 Below Saturated Compression Cyclic SP8 3 2 Below Saturated Compression- Flexure Monotonic a Bents are numbered increasing to the east. b Piles are numbered increasing to the south. c Elevation describes the original position of the specimen as either above the water line, below the water line, or a repair post. If the prototype bridge was loaded unsymmetrically (Figure 2), the piles would be subjected to combined compression and bending. Therefore, compressionflexure tests were conducted to determine the response of timber piles under such loading condition. To apply the load eccentrically, specimens were bolted to 38 mm (1 1/2 in) thick steel plates (loading plates) on each end as shown in Figure 6. The plates were loaded through rollers placed 76 mm (3 in) eccentric to the centroidal axis, thereby inducing a constant moment equal to the product of the applied load and 76 mm (3 in) combined with the axial force. Cyclic tests were conducted to evaluate potential pile deterioration due to repeated traffic loads cyclic tests were conducted. The specimens were cycled between the approximate dead load and increasingly higher loads, until the peak strength was reached. Table 2 summarizes the compression test results. The test-to-predicted ratio describes the experimental strength normalized by the NDS (AFPA, 2005) specifications strength, which yielded a mean of 1.14 and standard deviation of The mean measured specimen strength of 1108 kn (249 kips) was 5 times larger than the required pile capacity of 214 kn (48 kips). Borello et al. (2009) provides further experimental details.

6 103 Figure 5. Compression test setup Figure 6. Compression-Flexure setup

7 104 Table 2. Compression Tests Results Specimen SP1 SP2 SP4 SP5 SP6 SP7 Minimum Diameter Predicted Stress Predicted Strength Measured Ultimate Stress Measured Ultimate Strength Test/ Predicted Ratio mm (in) MPa (psi) kn (kips) MPa (psi) kn (kips) (2640) 17.3 (2515) (10.82) (243) (231.4) (2640) 22.4 (3255) (9.91) (204) (250.9) (2640) 20.2 (2924) (10.84) (244) (270) (2640) 23.4 (3390) (10.84) (244) (313) (2640) 19.0 (2752) (9.47) (186) (193.8) (2640) 22.1 (3206) (9.61) (191) (232.5) 1.21 Mean 20.7 (3007) 1108 (249) 1.14 Std. Dev. 2.3 (335) 178 (40) 0.13 The load-displacement response of the monotonic compression tests are shown in Figure 7a. The initial stiffness was approximately linear for all specimens, while the peak load varied slightly. This could be attributed to cross-sectional area variation among specimens. The four specimens tested monotonically yielded an average peak stress of 19.7 MPa (2861 psi). The post-peak softening response was also nearly linear, exhibiting a ductile response without sudden failure. The slight increases in the load during this softening branch of the curve, as observed in SP2 and SP4, occurred when the imposed displacement rate was increased. After significant deformation, the specimens exhibited local buckling of the fibers and occasional longitudinal splitting. Figure 7b presents the load-displacement response of the cyclic compression tests. The pre-peak behavior is relatively linear elastic, designated by the specimen tracing the loading path when unloaded on each cycle. As illustrated in the figure, no signs of cyclic degradation in strength or stiffness due to repeated loading were observed. The monotonic load-displacement curve roughly envelopes the cyclic response of a comparable specimen. The two specimens, SP3 and SP8, that were tested in combined compression and flexure, were equipped with two extensometers placed symmetrically about the axis of bending. As expected, one side of the specimen experienced net tension while the other experienced compression. Figure 8 presents the applied load versus crosshead displacement response of the two specimens. The response was similar to the monotonic compression tests, exhibiting approximately linear pre-peak and post-peak behavior.

8 SP2 SP4 SP1 SP2 SP4 SP SP5 SP5 SP7 Load (kn) SP1 SP6 Load (kn) SP Actuator Displacement (mm) (a) Monotonic tests Actuator Displacement (mm) (b) Cyclic tests Figure 7. Response of compression specimens Table 3 summarizes the compression-flexure test results. In the table, the predicted strength is calculated by solving the NDS (AFPA, 2005) interaction equation for the maximum permissible load with a 76 mm (3 in) eccentricity. The mean strength of the two specimens was 449 kn (101 kips), showing a 60% reduction from the concentrically-loaded specimens. This result illustrates that the timber pile strength is sensitive to even modest load eccentricity. The mean test-to-predicted ratio of 0.75 may be attributed to a reduced specimen cross-section due to damage and longitudinal splitting, which slightly increased the calculated cross-sectional area and moment of inertia used to compute the predicted strength SP3 SP3 SP8 Load (kn) SP Actuator Displacement (mm) Figure 8. Average load vs. displacement of specimens tested in compressionflexure

9 106 Table 3. Compression-Flexure Test Results Specimen Predicted Strength Ultimate Strength Test/Predicted Ratio kn (kips) kn (kips) SP3 703 (158) 533 (119.8) 0.76 SP8 494 (111) 366 (82.3) 0.74 Mean 449 (101) 0.75 Std. Dev. 120 (27) 0.01 NUMERICAL MODEL DESCRIPTION A numerical model was created for each specimen, as well as a single model to represent an in-situ pile of the prototype bridge using the nonlinear finite element program OpenSees (OpenSees, 2009). The specimen and full-scale pile models included geometric, material and soil-structure interaction nonlinearity. The constitutive behavior of each fiber was defined using a uniaxial material. To calibrate the material model used in the full-scale pile analysis, two numerical models were developed for the experimental testing; compression loading and combined compression and flexure loading. The spherical head of the compression tests was modeled as a pinned boundary condition while the surface bearing was represented as a fixed condition. The boundaries for the combined compression-flexure tests were modeled as eccentric pinned supports. Two uniaxial materials were calibrated from the experimental response of the timber specimens, representing the pile above and below the riverbed, assumed to have air-dried and water-saturated moisture contents, respectively. The compression branch was modeled using specimens SP4 and SP6 responses for above and below the riverbed, respectively, as nonlinear up to a peak stress. The post-peak response was modeled as linear softening followed by perfectly plastic behavior to represent residual stresses equal to one-half of the peak stress. Due to the similarities between the compressive behaviors of concrete and tested timber, the compression branch of the uniaxial material model was based on the OpenSees Concrete02 material model (OpenSees, 2009). Tensile response was calibrated using the combined compression-flexure tests of specimens SP3 and SP8 for above and below the riverbed, respectively. Stressstrain response was modeled as linear elastic up to fracture based on the response of wood reported in the literature (Gurfinkel, 1973). PROTOTYPE SINGLE PILE For the case of a statically determinate superstructure, common for simplysupported spans, the superstructure provides negligible resistance to collapse of the foundation. Therefore, it was deemed acceptable to model the foundation independently of the superstructure. A nonlinear numerical model including soil response from the prototype structure was developed for a single pile. This model was utilized to predict the ultimate pile strength under concentric and eccentric loading conditions.

10 107 The pile model was divided into two regions, 3.3 m (11 feet) above the riverbed, and 5.2m (17 feet) below the riverbed based on the respective calibrated uniaxial material models discussed above. The top of the pile was connected to an 864 mm (34 in) rigid link to represent the concrete pile cap as shown in Figure 9. The deck and cross-bracing between the piles were assumed to provide adequate stiffness to prevent longitudinal and transverse translation of the pile cap, respectively. However, the deck-to-pile cap connection was insufficient to restrain the pile cap against rotation. Therefore, the top node of the rigid link was constrained against horizontal translation but allowed to rotate (Figure 9). A sensitivity analysis illustrated that the parameters selected for pile diameter, live load eccentricity and initial out-of-plumb ratio were reasonable. Figure 9. Prototype single pile numerical model Lateral (transverse) nonlinear soil springs were placed every 152 mm (6 in) to represent the resistance of the soil against pile buckling. The properties of these springs were determined from a geotechnical site investigation performed by the Illinois department of Transportation and Borello et. al (2009). The nonlinear geotechnical soil springs [p-y curves (API, 1987; O Neill and Reese, 1999)] were approximated with a tri-linear constitutive formulation. Elastic-perfectly plastic vertical springs [t-z curves (Olson, 1990)] were used to model skin friction and end bearing resistance of the soil, and were placed at every node below the riverbed. Displacements of 2.5 mm (0.1 in) and 3.0 mm (0.12 in) were required to reach maximum side resistances and maximum end bearing [i.e., plastic response], respectively. The load was applied to the top of the pile cap in two steps. First, the tributary dead load of 144 kn (32 kips) for the prototype bridge was applied concentrically, since it was applied symmetrically to the structure. When considering the live load supported by the bridge, there are two possible loading cases: (1) symmetrical, when the two spans attached to a bent are loaded equally; and (2)

11 108 unsymmetrical, when the spans are unequally loaded. In the case of unsymmetrical loading, when the spans are simply supported, the live load will be eccentric to the supporting piles (Figure 2). For the prototype structure, assuming that the deck beams were bearing uniformly on the pile cap, they would apply the live load at an eccentricity of 197 mm (7.75 in). Therefore, the model was subjected to concentric compression for the symmetrical loading case (Case 1) and eccentric compression for the unsymmetrical case (Case 2). Live loads (axial and moment) were applied monotonically, up to failure, in a static analysis under displacement control to capture material softening behavior. Table 4 presents the results from the full-scale pile numerical analysis under concentric (Case 1) and eccentric (Case 2) loads. Under increasing loading, the pile response begins to soften, and it experiences a peak strength at which failure of the pile is predicted. The results clearly illustrate that Case 2 governs the foundation performance. The live load carrying capacity (total pile capacity minus the dead load) is reduced 77% by the unsymmetrical load. The total capacity in Case 2 is below the design strength of 214 kn (i.e., 48 kips per the design drawings). Table 4. Numerical Total and Live Load Strength of Single Pile Concentric Loading (Case 1) Eccentric Loading (Case 2) a kn (kips) kn (kips) Total Pile Capacity 389 (87) 201 (45) Dead Load 142 (32) 142 (32) Live Load Capacity 247 (55) 58 (13) CONCLUSIONS This paper summarizes experimental and numerical work conducted to study the effect of load eccentricity on timber pile strength in the context of a prototype structure. The prototype structure provided experimental specimens and example parameters for the numerical study. Experimental testing of six samples exhibited a mean capacity of 1108 kn (249 kips) under monotonic compression. The mean capacity of two similar specimens tested with a 77 mm (3 in) eccentric load was 449 kn (101 kips), a 60% reduction compared to the concentrically loaded specimens. The experimental testing was used to calibrate a numerical model of a single full-scale pile of the prototype structure including geometric, material and soilstructure interaction nonlinearity. The live load capacity under symmetrical and unsymmetrical loading was determined to be 247 kn (55 kips) and 58 kn (13 kips), respectively. This shows that a slight eccentricity can drastically reduce the capacity of timber piles. Therefore, it is recommended that during the design of bridges with timber piles, the superstructure should be analyzed carefully to determine the possibility of inducing eccentricity in the foundation, especially under unsymmetrical loads.

12 109 REFERENCES American Forest and Paper Association (AFPA) (2005). National Design Specification for Wood Construction. Washington, DC: American Forest and Paper Association. American Petroleum Institute (API) (1987). Recommended Practice for Planning, Designing, and Constructing Fixed Offshore Platforms, API Recommended Practice 2A (RP2A), Seventeenth Edition, April. The American Association of State Highway Officials (AASHTO) (1973). Standard Specifications for Higway Bridges Eleventh Edition. ASTM International (ASTM) (2005). D198 Standard Testing Methods of Static Tests of Lumber in Structural Sizes. West Conshohocken, Pennsylvania: American Society of Testing and Materials. Borello, D. J., Andrawes, B., Hajjar, J. F., Olson, S. M., Hansen, J., and Buenker, J. (2009). Forensic Collapse Investigation of A Concrete Bridge with Timber Piers. Springfield, Illinois: Illinois Center for Transportation. Gurfinkel, G. (1973). Wood Engineering. New Orleans, Louisiana: Southern Forest Products Association. Newlin, J. A., and Trayer, G. W. (1956). Stresses in Wood Members Subjected to Combined Column and Beam Action. Madison, Wisconsin: Forest Products Laboratory. Olson, R.E. (1990). Axial Load Capacity of Steel Pipe Piles in Sand. Proceedings of the Offshore Technology Conference, Houston, Texas. Richardson, Texas: Offshore Technology Conference O Neill, M. W. and Reese, L. C. (1999). Drilled Shafts: Construction Procedures and Design Methods, Report No. FHWA-IF , Washington, DC: Federal Highway Administration. OpenSees (2009). Open System for Earthquake Engineering Simulation. Berkeley, California: University of California, Berkeley. Timber Piling Council. (2002). Timber Pile Design and Construction Manual. Vancouver, Washington: American Wood Preservers Institute. Wood, L. W. (1950). Formulas for Columns with Side Loads and Eccentricity. Madison, Wisconsin: Forest Products Laboratory. Zahn, J. J. (1982). Strength of Lumber Under Combined Bending and Compression. Madison, Wisconsin: Forest Products Laboratory. Zahn, J. J. (1988). Combined-Load Stability Criterion for Wood Beam-Columns. Journal of Structural Engineering, ASCE, Vol. 114, No. 11, pp ACKNOWLEDGEMENTS This work was supported by the Illinois Center for Transportation (ICT) under project no. R27-SP12; the Illinois Department of Transportation (IDOT); the U.S. DOT, the Federal Highway Administration (FHWA); and the University of Illinois at Urbana-Champaign. The authors thank the members of the project Technical Review Panel from IDOT, the FHWA, the ICT, as well as personnel from IDOT and DeKalb County Highway Department for their assistance with the research.