Analysis of seismic pile response in liquefiable ground using a constitutive model for large post-liquefaction deformation
Rui Wang,Xing Liu,Jian-Min Zhang
2015-01-01
Abstract:This paper presents a three-dimensional finite ele m nt study on single piles in liquefiable soil using a unified plasticity model for the analysis o f large post-liquefaction shear deformation of sand. The constitutive model is capable of providin g a unified description of the monotonic, cyclic and post-liquefaction behaviour of sand at differen t states, by applying a unique decomposition of volumetric strains and formulations for dilatancy. Applying the model, three dimensional solidfluid coupled finite element analysis of seismic si ngle pile response in liquefied is conducted. The analysis method is first validated in simulation of a centrifuge shaking table test on single pile in liquefiable ground, and hence utilized to investiga te the effects of kinematic and inertial interaction on pile bending moment. Introduction The seismic response of piles in liquefiable ground is a dynamic nonlinear three-dimensional problem. Ongoing research efforts in dynamic couple d formulations for soil skeleton–pore fluid problems (e.g. Zienkiewicz et al, 1999) and constit utive models for sand liquefaction (e.g. Yang and Elgamal, 2003; Boulanger and Ziotopoulou, 2013; Wang et al, 2014) have made threedimensional dynamic continuum methods more effectiv and appealing for the analysis of piles in liquefiable ground (e.g. Finn, 2004; Cheng and J eremic 2009). The dynamic continuum approach has the advantage of providing a more rati on l analysis for the soil-pile kinematic interaction and the structure-foundation inertial i nteraction, and especially the coupling of the kinematic and inertial interactions. The coupling o f the kinematic and inertial forces have traditionally been treated as linear combinations o f the peak kinematic and inertial loads with various coefficients (Liyanapathirana and Poulos, 2 005; Caltrans, 2013), but has shown to be more complicated through experiments (Tokimatsu et al, 2005; Brandenberg et al, 2005). This paper aims to simulate and investigate the sei smic response piles in liquefiable ground through a three-dimensional solid-fluid coupled fin ite element analysis approach, with focus on the combined effects of kinematic and inertial inte raction. A novel constitutive model for the analysis of large-post-liquefaction is presented an d applied in the simulation of piles in liquefiable ground. The simulation method is valida ted against a centrifuge shaking table test on a single pile usystem, and then applied to study th e roles of soil-pile kinematic interaction and the structure-foundation inertial interaction. Assistant Researcher, State Key Laboratory of Hydro science & Engineering, Tsinghua University, Beijing , China, wangrui_05@mail.tsinghua.edu.cn. PhD student, State Key Laboratory of Hydroscience & Engineering, Tsinghua University, Beijing, China. Professor, State Key Laboratory of Hydroscience & E ngineering, Tsinghua University, Beijing, China. Constitutive Formulation One of the most important aspect of three-dimension al dynamic continuum analysis of piles in liquefiable ground is the choice of appropriate con stitutive models for sand. A unified plasticity model for large post-liquefaction shear deformation was formulated to appropriately reflect the cyclic mobility (strongly related to dilatancy) and large post-liquefaction shear deformation. The elastic shear and bulk moduli for the model fol low the formulation of Richart et al (1970). A state parameter (Been and Jefferies, 1985) was intr oduced in the model formulations to provide unified description of sand under different densiti e and confining pressures through compliance with the critical state theory: c e e Ψ = − (1) with e being the current void ratio and ec the critical void ratio. The model operates within the framework of bounding surface plasticity by adopting modifying features from a model by Wang et al (1990) for plas tic modulus and its respective mapping rule (Fig. 1). With the plastic modulus formulated as: ( ) 2 exp( ) exp( ) 1 3 b p m M n H hg G n M ρ θ ρ − Ψ = − Ψ − (2) where h and p n is are model constants, θ is the lode angle, Ψ is a state parameter, M is the critical state stress ratio, m M is the maximum stress ratio history during loading , ρ is the distance between r and in α (Figure 1), and ρ the distance between r and in α (Figure 1). The model decomposes dilatancy in to reversible and irreversible components based on observations of drained cyclic torsional tests (Sha moto et al, 1997), with the rate of each respected component defined as: ( ) ( ) ,1 2 ,2 2 , 0 3 / , 0 re re re d D d p χ < = > d d d r r : n (r r) : n (r r) : n (3) ( ) 2 , , , 1 exp( ) exp( )( exp( ) ) 1 exp( ) d d r d ir ir vd ir d d r mono n D d n n γ αε χ γ γ < − Ψ > = Ψ − < > + < − Ψ > + d r r : n (4) where re D and ir D represent the reversible and irreversible dilatanc y rate. ,1 re d , ,2 re d , ir d , d n , α , , d r γ are material constants for dilatancy. χ is a function enhancing the dilatancy at load reversal. mono γ is the cumulative shear strain since the last stre s versal. Figure 1. Surfaces and mapping rules This unique formulation of dilatancy is vital in re flecting the cyclic mobility of sand and allow the undrained cyclic stress path to reach all the w ay to liquefaction, with reversible dilation generating and releasing during each load cycle and irreversible dilation accumulating asymptotically. At liquefaction, the elastic respon se of sand is considered unchanged while dilation is assumed to continue, thus generating an increasing shear strain each cycle at the state of liquefaction. Table 1 lists the model parameters and explains their physical meanings and calibration methods. Table 1. Details of model parameters. Parameter Physical meaning Calibration method* Value for Fujian sand o G Elastic shear modulus constant Small strain T 200 κ Rebound index Rebound D T 0.006 h Plastic modulus constant D T 1.7 M Stress ratio at critical state in compression D/U T 1.3 ,1 re d Reversible dilatancy generation rate constant D C T 0.45 ,2 re d Reversible dilatancy release rate constant D C T 30 ir d Irreversible dilatancy rate constant U C T 0.6 α Decrease rate constant of irreversible dilatancy U C T 40 , d r γ Reference shear strength length U C T 0.05 p n State constant for bounding surface D T 1.1 d n State constant for reversible dilatancy surface D C T 8.0 c λ Critical state constants D/U T 0.023 0 e 0.837 ξ 0.7 * Note: U is undrained, D is drained, C is cyclic, and T is triaxial (can also be torsional for cyclic tests). FEM Simulation The constitutive model has been implemented into th e FEM framework OpenSees (McKenna and Fenves, 2001) with the tag CycLiqCPSP. Combined with u-p form brick elements, soil liquefaction analysis can be achieve through soil–p re fluid coupled formulations, enabling the FEM simulation seismic pile response in liquefiable ground. Using this FEM approach, a centrifuge shaking table test on a single pile in s aturated sand was simulated. The shaking table test was conducted at the geotech nical centrifuge facility at Tsinghua University under 30 g centrifugal acceleration, wit h a single direction horizontal excitation input at the base of the model. The model was constructed within a laminar box to achieve periodic boundaries. A 6m long square pile was installed ver tically into a ground consisting of two layers of Fujian sand, a 5m medium dense ( Dr=50%) layer sand overlying a 2.5 m dense ( Dr=80%) layer. The piles used were square aluminium piles w ith EI = 47.25 MNm. A HPMC (hydroxypropyl methylcellulose) solution with 30 ti mes the viscosity of water was used as the pore fluid. A pile cap with a 10.8t superstructure on top was connected to the pile head. All parameters and measurements are given in prototype scal . The finite element mesh for the numerical simulatio n of the centrifuge shaking table test is shown in Figure 2, which is only half of the actual physical model due to symmetry. The pile was simulated with second order brick elements and linear elastic isotropic constitutive model, the cross section of the pile in the simulations co nsisted of 6 elements to accurately calculate the bending moment and curvature of the pile. The two l ayers of sand were simulated using up elements and the unified plasticity model for large post-liquefaction shear deformation of sand. The model parameter values used in the simulations are shown in Table 1, the elastic shear modulus parameter ( G0), plastic modulus parameter ( h ) and critical state stress ratio ( M) was obtained from drained triaxial test data, and elast ic bulk modulus parameter ( κ ) was determined via the rebound curves of triaxial consolidation te sts. The critical state parameters ( c λ , e0, ξ ) for Fujian sand reported by Yang and Sze (2011) were us ed. Figure 2. FEM mesh for the simulation of a single p i in liquefiable ground Figure 3 shows typical results of horizontal accele ration and excess pore pressure in the ground from both test measurement and numerical simulation . It can be seen from comparing the two set of results that the numerical simulation well repro duced the seismic response of the liquefiable ground. The maximum input acceleration at the base of the model was -4.95 m/s , occurring at 6.76s, while the maximum acceleration at the ground surface was less than -3 m/s 2 and was greatly deamplified due to the buildup of excess po re pressures and subsequent decrease in effective stress in the ground. In both the centrif uge test and the numerical simulation, the top 4m of sand reached liquefaction after about 10s, with excess pore pressure ratio r u = 1.0. Figure 3. Calculated and measured acceleration and excess pore pressure Figure 4. Calculated and measured pile moment: (a) pile moment histories at three depth, (b) peak pile moment along pile depth The seismic response of the pile is illustrated in terms of pile moment in Figure 4. The calculate