BEHAVIOR OF A TIMBER BRIDGE PILE EMBEDDED ...

BEHAVIOR OF A TIMBER BRIDGE PILE EMBEDDED IN A REINFORCED CONCRETE CAP

B. L. Deam1, J. B. Mander2, and T. Rahardjo3

ABTRACT An experimental study is used to characterize the behavior of Pinus Radiata bridge piles embedded in a reinforced concrete cap beam when subjected to earthquake induced motion. A nonlinear model of the pile and connection is developed from experimentally measured properties of wood coupons. Incremental Dynamic Analysis (IDA) is used to identify the most critical earthquake ground motion and the intensity required to achieve the level of lateral drift that corresponds to the drift capacity of the pile. It is shown that the model used for the IDA needs to provide a good estimate of the wood properties in order to select a suitable ground motion and define its intensity. Experimental results for an 80 % of full scale bridge pile are presented for bidirectional quasi-static and pseudo dynamic tests. The predicted response is shown to have good agreement with the hysteretic response for the bi-directional pseudo dynamic test. It is concluded that for performance based design purposes, the chord rotation of these pile foundation connections should be limited to 0.03 radians. INTRODUCTION Pinus Radiata is the principal timber used for commercial and domestic construction in New Zealand. However, Australian hardwoods such as Iron Bark and Jarrah have been used extensively for timber bridges within New Zealand in the past. The seismic performance of timber bridges constructed using Pinus Radiata needs to be assessed before they can be approved for use within New Zealand’s more seismically active regions. Shama and Mander (2004) experimentally investigated the behavior of North American grown Douglas Fir piles embedded in concrete caps that were subjected to reversed cyclic loading. A principal objective of their research was to ascertain the depth that the wooden pile needed to be embedded into the concrete cap to provide good seismic performance. They concluded that the minimum embedment depth needed to be approximately 1.5 times the pile diameter to provide the rotational restraint required to fully utilize the flexural strength of the pile. 1

Leicester Steven EQC Lecturer in Earthquake Engineering, Dept. of Civil Engineering, University of Canterbury, Private Bag 4800, Christchurch, NZ 2 Professor, Dept. of Civil Engineering, University of Canterbury, Private Bag 4800, Christchurch, NZ 3 Former Graduate Student, Dept. of Civil Engineering, University of Canterbury

They developed a method of estimating the lateral drift of the pile and cap assemblage and noted that compression perpendicular-to-the-grain had a significant effect on the performance of timber pile-to-concrete cap connection under lateral loading. The research presented herein describes a prototype bridge structure and tests conducted on a reduced-scale Pinus Radiata pile and cap assemblage. Several test methods, including quasi earthquake displacement and pseudo dynamic tests, were used to investigate the seismic performance of the assemblages when they were loaded in two perpendicular directions. Samples were extracted from the piles to provide material properties for further computational modeling. The computational models used the Incremental Dynamic Analysis (IDA) procedure to identify the critical ground motions used in quasi earthquake displacement and pseudo dynamic tests. The experimental results are then evaluated and compared to the computational estimations. MATERIAL PROPERTY TESTS Shama and Mander (2004) proposed a method of calculating the lateral displacement at the top of a pile from the non-linear stress-strain relationships for the wood. They found that the lateral resistance increased with pile embedment depth and that flexural bending dominates the behavior of slender piles when the embedment depth was greater than the pile diameter, dp. They defined slender piles as those with M/Vdp > 2.5, where M = pile moment and V = pile shear. Samples were cut from a CCA treated Pinus Radiata pile and tested to obtain the properties for the experimental work described later. The test methods followed ASTM D143 – 94 (ASTM, 1994) for modulus of rupture, parallel to the grain compression, perpendicular to grain compression and parallel to grain tension. Five specimens were tested using each method. The test results are plotted using grey lines in Figure 1.

Figure 1. Stress-strain responses of the clear-wood samples extracted from the pile (light lines) and stress-strain relationships fitted to these (dark lines) for a) compression parallel, b) tension parallel and c) perpendicular compression. d) Predicted pile lateral-load vs drift relationships.

A modified concrete stress-strain model (Mander et al., 1988) was fitted to the coupon test results. A timber pile can fail either by i) tensile wood fracture; or by ii) crushing of the wood perpendicular to the grain within the embedded length (Shama and Mander, 2004). The expected lateral strength of the timber pile specimen was calculated as 25.2 and 57.1 kN for the two respective failure modes, so tensile bending fracture was expected to govern. Moment-curvature analysis was used to predict the lateral load vs drift behavior of the pile. Monotonic pushover curves for the upper 95th and lower 5th percentile bounds of the strength vs drift behavior are shown in Figure 1(d). These were calculated from moment-curvature analyses that used the material properties obtained from the coupon tests. EXPERIMENTAL PILE SPECIMEN A long multi-span bridge across an estuary was chosen as the prototype for this research. The simply-supported bridge spans were assumed to be 9 m long, with a 10 m wide deck as shown in Figure 2. The superstructure (deck) was assumed to have a distributed weight of 5 kPa, and the span ends were assumed to be supported by multi-pile pier bent substructures. The reinforced concrete pile cap in each substructure was assumed to translate both laterally and longitudinally with minimal rotation in either direction. The firm soil surrounding the pile bases was assumed to provide little rotational restraint for the piles. The lateral stiffness of the abutment was assumed to be three times the lateral stiffness of the pier substructures. 10.0

9.0

9.0

Axial Load Ball joint

300

Section A-A

Pile

300

900x900x750 Base Block

A Transverse Section Through Bridge

240 2.10

Loading

2.65

Frame

Lateral Load

360

A

Axial Load

Test Assemblage

Figure 2. The prototype bridge structure and the reduced scale test assemblage An interior pile from the pier bent (Section A-A, Fig. 2) was chosen for further study. A reduced-scale (and upside down) model was required to fit within the laboratory equipment. Thus a 2.1 m long, 240 mm diameter pile was chosen to physically model the timber bridge, with a scale factor of 0.8. The wood pile specimen was grouted into a 300 mm diameter cylindrical cavity within the top of a 750 mm high reinforced concrete base block. The timber pile specimen was held in place while low-strength concrete grout was poured into the 360 mm deep (ie. 1.5 × 240 mm) cavity recommended by Shama and Mander (2004). A 72 kN axial force was applied to the assemblage using a Universal Testing Machine to provide constant stress and strain similitude between the prototype and specimen. Bi-directional lateral loads were applied using hydraulic actuators attached to a head block in both the east-west and north-south directions. These reacted against L-shape frames that were attached to the

concrete base block. The actuator forces were measured using in-series load cells. Lateral pile movements were measured from the other two sides of the pile, using long gauge lengths to minimize the errors created by vertical movement at the attachment points. REVERSE-CYCLIC LATERAL LOAD TESTS To advance beyond previous research (Shama and Mander, 2004 and Mander et al., 2000) that only used unidirectional cyclic loading, more complex experiments that incorporate bidirectional cyclic loading effects were conducted using three methods as described below. Cyclic Loading Tests The first tests were uni-directional cyclic loading tests (eg. Park, 1989), with two cycles to a drift amplitude of ±1.5 % and a third cycle to ±2 %. This sequence was used for separate tests in the east-west (E-W) direction and then the north-south (NS) direction. The load-drift responses for the two tests are given in Figure 3 below. The energy absorbed by the assemblage was converted to equivalent viscous damping. The absorption in the 2 % drift cycle in each direction was found to be equivalent 5 % of critical viscous damping.

Figure 3. Lateral load vs drift responses for uni-directional cyclic loading tests A concurrent bi-lateral loading pattern was used for the third quasi-static reverse-cyclic test. The specimen top (observed from above) traced the perimeter of a 4-leafed clover calculated using the parametric expression r = a sin 2θ relating the radial distance, r, to the angle, θ. The results of this test, with a drift amplitude, a = 2 %, are given in Figure 4, where the 4leafed clover plan view (a) of the deformations is flanked by 3rd angle projection “elevations” showing the lateral load and drift in each direction, (b) and (c), and across the diagonal (d). The bi-directional loading load-drift plots (b) and (c) were similar to those of Figure 2, showing that there were no adverse performance effects introduced by the bi-directional loading. The specimen absorbed an equivalent of 5.25 % viscous damping.

Quasi Earthquake Displacement Test The Quasi Earthquake Displacement (QED) testing method (Dutta et al., 1999) applies a horizontal displacement pattern that simulates the effect of real ground motion. The QED technique uses a non-linear time history simulation to predict responses (seismic displacements and forces) of the prototype structure. The prototype structure is first modeled with an inelastic analysis program and suitable ground motion records to develop a realistic displacement-time sequence of the assemblage.

a) Load-drift responses for QED test (black) and theoretical pushover predictions (grey) from Figure 1(d)

b) Time-drift responses Figure 4. Bi-directional quasi-static loading results for 4-leaf clover displacement pattern

Figure 5. Results for the QED test.

In the present study, one of the 1989 Loma Prieta Earthquake records (with PGA intensity of 0.2g) was identified as the critical ground motion using Incremental Dynamic Analysis. The use of IDA for bridges is described in a companion paper (Mander et al., 2006). The horizontal displacement pattern for the QED test was obtained using time history analysis on the prototype structure using RUAUMOKO 3D (Carr, 2004). The specimen response and the applied displacement pattern are given in Figure 5. This test used a maximum drift of 4.7%. During the test, the specimen prematurely failed by tensile bending ar a drift of 2.5% in one direction, but still survived at 4.7% drift with maximum load of 21.2 kN in the other direction. The partial failure of the specimen was attributed to faulty

installation of the instrumentation with unnecessarily large threaded rods that acted as stressraisers. The load-drift response of the specimen (Figure 5(a)) shows good agreement with the monotonic responses predicted from the material properties (Figure 1(d)). Pseudo Dynamic Test Although the QED method provides a more realistic displacement sequence than the customary reverse-cyclic loading method, it is still an open loop control method with a predetermined displacement sequence. It has no feedback to correct for unanticipated strength changes during the test. The accuracy of the method is therefore the same as the accuracy of the computational model used to generate the displacement sequence. Unexpected strength changes can only be accommodated using a closed loop control method like the pseudo dynamic test method. The pseudo dynamic (PD) test, first proposed by Takanashi et al (1975), combines the simplicity of static testing with almost the same realism as shaking table tests. The response behavior of the specimen proceeds step-by-step. It uses the restoring force measured from the test specimen itself to provide more accurate results than is possible using a priori selected restoring forces (ie in QED testing). The PD test method is thus a numerical time integration technique similar to that used for the dynamic analysis of structures, but it uses the force measured on the test specimen rather than a mathematical model of the specimen stiffness. It proceeds step-bystep, moving the specimen to its new position and then returning the measured force to the analysis algorithm so it can calculate the next position to move the specimen to (Machida and Mutsuyoshi, 1992). This cycle repeats throughout the test. Therefore during a PD test, the specimen behaves as though it is subjected to an earthquake. The PD test was conducted with a lengthened time scale to minimize the error arising from delayed actuator response. However, the time lengthening can lead to rate-effect inaccuracy. The relaxation in timber strength is well known (eg. Deam, 2001). The specimen resistance is conservatively lower that it would be without the lengthened time scale The PD test also requires an estimate of the (usually viscous) damping provided by the remainder of the structure before the test begins. The effect of this assumption is less important for structures like bridges where the majority of the energy is dissipated as hysteretic damping rather than viscous damping. A biaxial PD test was conducted on the partially damaged timber pile specimen, using the same earthquake record as the QED test, but with the earthquake intensity reduced to PGA = 0.1g. The specimen response is show in Figure 6. A second QED test was conducted on the damaged specimen with the same PGA intensity to compare the PD and QED test methods. The drift responses of the two tests (Figure 7) shows that while the periods were reasonably similar, the drift amplitudes were somewhat different.

Figure 6. Specimen responses for the first PD test with a PGA intensity of 0.1g.

Figure 7. Comparative drift responses for the second QED and the PD tests

A second biaxial PD test was carried out using another earthquake (PGA = 0.3g) selected using IDA. Figure 8 shows that damage in the E-W direction also influenced the performance in the N-S direction, as evidenced by the reduction in resistance at ~ 2.5 % drift amplitude. The pile performed surprisingly well, in spite of the significant reduction in resistance and the -0.3 % residual drift at the end of the test.

Figure 8. Specimen responses for the second PD test with a PGA intensity of 0.3g.

DISCUSSION The comparison of analytical prediction in IDA and the PD test outcomes are given in Figure 9. Although Incremental Displacement Analysis (IDA) is an effective tool to predict the maximum drift of the structure under an earthquake with a certain PGA, it is inevitable to have discrepancies between analytical predictions shown in IDA curves and actual responses in the PD test since the specimen was subjected to strength degradation due to partial damage from previous tests while the analytical prediction in IDA assumes the structure is still mostly in the elastic range. The discrepancies are further exacerbated by the uncertainties in modeling due to different values of parameters used in modeling as mentioned above.

Figure 9. a) IDA curves showing the fractile intervals of expected displacement; and b) comparison between pre test analytical prediction using IDA and the Pseudo Dynamic test result The theoretical pushover curve generated from the fundamental analysis proposed by Shama and Mander (2004) effectively envelopes the experimental cyclic loops as shown in Figure 9(b). However, sudden failure was evident in the present tests due to a flaw in the experiment set up where screws were placed in the timber 100 mm above the concrete base. These acted as crack initiators when the extreme wood fibers were under tension. This led to the premature tensile fracture under high applied moments (at about 2.6% drift). Therefore, greater concern should be raised and more specimens need to be tested. If a similar type of failure happens, the limitation of displacement capacity is required. Previous research on the sound (unflawed) Douglas Fir timber piles by Shama and Mander (2004), showed that the timber piles can be sufficiently flexible enough to undergo deformation up to 12% drift without significant strength degradation. This seems to less likely outcome for NZ grown Pinus Radiata. However, the experiment needs to be repeated with a sound specimen. CONCLUSIONS Based on the experimental investigation described herein the following conclusions can be drawn: 1. Providing the structure is mostly elastic and can be computationally modeled accordingly, there is little difference between the merits of the QED and PD testing procedures. However if the structural element is highly non-linear and with behavior difficult to predict a priori,

then the PD testing procedure has the advantage of directly capturing all non-linear attributes including sudden failure. 2. The theoretical model developed in the previous study by Shama and Mander (2004) gives a reliable approach that provides an appropriate estimation of cap-to-pile connection performance including the expected range of behavior. 3. Compared to North American grown Douglas Fir pile-to-cap connection tested previously by Shama and Mander (2004), the New Zealand grown Pinus Radiata connection is evidently less ductile. Therefore, a designer needs to consider the displacement limits of the connection when designing Pinus Radiata pile-to-cap connections. It is suggested that the chord rotation of such piles be limited to 0.03 radians. ACKNOWLEDGEMENTS The first author gratefully acknowledges the financial support of the New Zealand Earthquake Commission (EQC). REFERENCES ASTM, 1994. Standard Methods of Testing Small Clear Specimens of Timber, American Society for Testing Materials Standard D143 – 94, Vol. 4.10. Carr A.J, 2004, RUAUMOKO: Inelastic Dynamic Computer Program, Computer Program Library, Department of Civil Engineering, University of Canterbury, Christchurch, New Zealand. Deam B.L, 2001, Seismic Ratings for Degrading Structural Systems, Bulletin of the New Zealand Society for Earthquake Engineering, 34(3):219-220 Dutta A, Mander J.B, and Kokorina T, 1999, Retrofit for Control and Repairability of Damage, Earthquake Spectra, 15 (4):666-670 Machida A and Mutsuyoshi H, 1992, Development of Accurate Pseudodynamic Test Method for R/C Structures, Tenth World Conference on Earthquake Engineering pp2647-2652 Mander, J., Priestley, M.J. and Park, R., 1988, Theoretical stress-strain model for confined concrete, ASCE Journal of Structural Engineering. 114(8):1804-1826. Mander, J.B., Allicock, D.R., and Friedland, I.M., (2000) Seismic Performance of Timber Bridges. Transportation Research Record 1740, Paper No. 00-1231, National Academy Press, Washington DC. pp 75-84. Mander, J.B., Dhakal, R.P. and Mashiko, N. 2006. Incremental dynamic analysis applied to seismic risk assessment of bridges. Submitted to the 100th Anniversary Conference commemorating the 1906 San Francisco Earthquake. Park, R. 1989. Evaluation of Ductility of Structures and Structural Assemblages from Laboratory Testing. Bulletin of the New Zealand National Society for Earthquake Engineering 22(3):155-166. Shama A.A and Mander J.B, 2004, Behavior of Timber Pile-To-Cap Connection Under Cyclic Lateral Loading, ASCE Journal of Structural Engineering, 130(8):1252-1262, August 2004. Takanashi K, K. Udagawa, M. Seki, T.Okada, and H. Tanaka, 1975, Nonlinear earthquake response analysis of structures by a computer-actuator on-line system, Bulletin of Earthquake Resistant Structure Research Center 8, Institute of Industrial Science, University of Tokyo, Tokyo, Japan.