Simple formulas for the dynamic stiffness of pile groups - Laboratoire ...
Simple formulas for the dynamic stiffness of pile groups - Laboratoire ...
Simple formulas for the dynamic stiffness of pile groups - Laboratoire ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICSEarthquake Engng Struct. Dyn. 2002; 00:1–6 Prepared using eqeauth.cls [Version: 2002/11/11 v1.00]<strong>Simple</strong> <strong><strong>for</strong>mulas</strong> <strong>for</strong> <strong>the</strong> <strong>dynamic</strong> <strong>stiffness</strong> <strong>of</strong> <strong>pile</strong> <strong>groups</strong>Reza Taherzadeh 1 , Didier Clouteau 1,∗ , Régis Cottereau 21 <strong>Laboratoire</strong> MSSMat, École Centrale Paris, CNRS UMR 8579, Grande Voie des Vignes, 92295Châtenay-Malabry, France2 International Center <strong>for</strong> Numerical Methods in Engineering (CIMNE), Universitat Politècnica deCatalunya, Jordi Girona 1-3, 08034 Barcelona, SpainSUMMARY<strong>Simple</strong> <strong><strong>for</strong>mulas</strong> are derived <strong>for</strong> <strong>the</strong> <strong>dynamic</strong> <strong>stiffness</strong> <strong>of</strong> <strong>pile</strong> group foundations subjected to horizontaland rocking <strong>dynamic</strong> loads. The <strong>for</strong>mulations are based on <strong>the</strong> construction <strong>of</strong> a general model <strong>of</strong>impedance matrices as <strong>the</strong> condensation <strong>of</strong> matrices <strong>of</strong> mass, damping and <strong>stiffness</strong>, and on <strong>the</strong>identification <strong>of</strong> <strong>the</strong> values <strong>of</strong> <strong>the</strong>se matrices on an extensive database <strong>of</strong> numerical experimentscomputed using coupled Finite Element-Boundary Element (FE-BE) models. The <strong>for</strong>mulationsobtained can be readily used <strong>for</strong> design <strong>of</strong> both floating <strong>pile</strong>s on homogeneous half-space and endbearing<strong>pile</strong>s, and are applicable <strong>for</strong> a wide range <strong>of</strong> mechanical and geometrical parameters <strong>of</strong> <strong>the</strong> soiland <strong>pile</strong>s, in particular <strong>for</strong> large <strong>pile</strong> <strong>groups</strong>. For <strong>the</strong> seismic design <strong>of</strong> a building, <strong>the</strong> use <strong>of</strong> <strong>the</strong> simple<strong><strong>for</strong>mulas</strong> ra<strong>the</strong>r than a full computational model is shown to induce little error on <strong>the</strong> evaluation <strong>of</strong><strong>the</strong> response spectra and time histories. Copyright 2002 John Wiley & Sons, Ltd.key words: Soil impedance matrix; <strong>pile</strong> group foundation; design <strong><strong>for</strong>mulas</strong>; lumped-parametermodels; hidden variables models1. INTRODUCTIONWhatever <strong>the</strong> mode <strong>of</strong> vibration, <strong>the</strong> <strong>dynamic</strong> <strong>stiffness</strong> <strong>of</strong> a <strong>pile</strong> group cannot be computedby simply adding <strong>the</strong> <strong>stiffness</strong>es <strong>of</strong> <strong>the</strong> individual <strong>pile</strong>s. Depending on <strong>the</strong> mechanical andgeometrical parameters <strong>of</strong> <strong>the</strong> soil and <strong>pile</strong>s, <strong>the</strong> <strong>dynamic</strong>al behavior <strong>of</strong> each <strong>pile</strong> can be heavilyinfluenced by that <strong>of</strong> its neighbors [1]. Among o<strong>the</strong>r phenomena, it is clear that <strong>the</strong> <strong>dynamic</strong>resonance <strong>of</strong> <strong>the</strong> soil constrained within a cluster <strong>of</strong> <strong>pile</strong>s cannot be modeled when <strong>the</strong> complex<strong>dynamic</strong> interaction between <strong>the</strong>se <strong>pile</strong>s is neglected.The main approach to solve this strongly coupled problem is <strong>the</strong> use <strong>of</strong> full numerical models,taking into account <strong>the</strong> soil and <strong>the</strong> <strong>pile</strong>s with equal rigor. This is however a computationallyvery demanding approach, in particular <strong>for</strong> large number <strong>of</strong> <strong>pile</strong>s, and has only been attemptedat, to <strong>the</strong> knowledge <strong>of</strong> <strong>the</strong> authors, by Kaynia [2], using <strong>the</strong> Boundary Element (BE) method.All o<strong>the</strong>r numerical methods in <strong>the</strong> literature seem to include some simplifying assumptions.∗ Correspondence to: didier.clouteau@ecp.frCopyright 2002 John Wiley & Sons, Ltd.
2 R. TAHERZADEH, D. CLOUTEAU & R. COTTEREAUFor example, <strong>the</strong> axisymetrical Finite Element (FE) model [3], <strong>the</strong> Ring-Pile model [4], or <strong>the</strong>closely-spaced plates model [5] can be used, when <strong>the</strong> geometrical layout <strong>of</strong> <strong>the</strong> <strong>pile</strong> group allows<strong>for</strong> it. The latter two approaches consist in grouping <strong>the</strong> <strong>pile</strong>s in concentric circles or soil-<strong>pile</strong>strippedupright plates, respectively, both allowing <strong>for</strong> an easier evaluation <strong>of</strong> <strong>the</strong> interactioneffects. Ano<strong>the</strong>r approach consists in replacing <strong>the</strong> <strong>pile</strong> group by a single equivalent uprightbeam [6]. In any case, <strong>the</strong>se approaches are not adapted to <strong>the</strong> needs <strong>of</strong> civil and structuralengineers, who need to design <strong>pile</strong> foundations with little recourse to computational tools.A more interesting approach <strong>for</strong> design purposes consists in providing analytical <strong><strong>for</strong>mulas</strong>,whose structure is usually derived from physical considerations, and with tabulized parameters,depending on <strong>the</strong> geometrical and mechanical parameters <strong>of</strong> <strong>the</strong> soil and <strong>pile</strong>s. The simplesttype <strong>of</strong> such approaches is based on Winkler’s spring model <strong>for</strong> <strong>the</strong> soil, <strong>for</strong> which radiationdamping and inertial effects are neglected [7, 8, 9]. A relatively simple method was proposed byGazetas and Dobry [18] <strong>for</strong> estimating <strong>the</strong> damping characteristics <strong>of</strong> horizontally loaded single<strong>pile</strong> in layered soil. Following Wolf’s approach [10] <strong>for</strong> <strong>the</strong> modeling <strong>of</strong> soil-structure interaction(SSI), o<strong>the</strong>r researchers [11, 12, 13, 14, 15, 16] have replaced <strong>the</strong> soil-<strong>pile</strong> system by a onedegree-<strong>of</strong>-freedom (DOF) mass, with a damper and a spring. Inertial effects and radiationdamping are <strong>the</strong>re<strong>for</strong>e taken into account to some extent but <strong>the</strong> general <strong>dynamic</strong>al behavior,and in particular <strong>the</strong> interaction between <strong>the</strong> different <strong>pile</strong>s, is heavily simplified. To improve<strong>the</strong>se models, Dobry and Gazetas [17] proposed an approximate <strong>for</strong>mulation accounting <strong>for</strong><strong>the</strong> interaction between <strong>the</strong> <strong>pile</strong>s, by modeling <strong>the</strong> waves emanating from each excited <strong>pile</strong>.The additional term is <strong>the</strong>re<strong>for</strong>e based on <strong>the</strong> computation <strong>of</strong> <strong>the</strong> propagation <strong>of</strong> a wave,supposed to be cylindrical, emanating from a single excited <strong>pile</strong> in a homogeneous domain.The method was fur<strong>the</strong>r refined by Gazetas and co-workers [19, 20, 21] to attempt to modelmultiple reflections within <strong>the</strong> <strong>pile</strong> group in layered soil. However, a few attempts have beenmade at accurately modeling <strong>the</strong> large <strong>pile</strong> group foundation, in particular <strong>for</strong> <strong>the</strong> complexfrequency-dependance <strong>of</strong> end-bearing <strong>pile</strong> foundations. (Konagai [6] provides <strong><strong>for</strong>mulas</strong> validonly <strong>for</strong> sway, Mylonakis and Gazetas [21] provide <strong><strong>for</strong>mulas</strong> valid <strong>for</strong> all movements but only<strong>for</strong> a group <strong>of</strong> nine <strong>pile</strong>s and Nikolaou et al. [40] provide <strong>for</strong> kinematic <strong>pile</strong> bending <strong>for</strong> a group<strong>of</strong> twenty <strong>pile</strong>s).Despite <strong>the</strong> significant progress in <strong>pile</strong> <strong>dynamic</strong>s [22], <strong>the</strong>re is <strong>the</strong>re<strong>for</strong>e still a need <strong>for</strong> simpleengineering procedures <strong>for</strong> <strong>the</strong>ir design, following <strong>the</strong> example <strong>of</strong> <strong>the</strong> code provisions developed<strong>for</strong> <strong>the</strong> seismic design on spread footings [23, 24]. This paper aims at providing such <strong><strong>for</strong>mulas</strong>,to be used <strong>for</strong> both small and large <strong>pile</strong> <strong>groups</strong>, as well as <strong>for</strong> both floating <strong>pile</strong> <strong>groups</strong> onhomogeneous half-space and end-bearing <strong>pile</strong> <strong>groups</strong>. The main novelty <strong>of</strong> this paper is that<strong>the</strong> <strong><strong>for</strong>mulas</strong> are valid over a range <strong>of</strong> parameters larger than <strong><strong>for</strong>mulas</strong> previously availablein <strong>the</strong> literature (see above references). They can be used <strong>for</strong> large numbers <strong>of</strong> <strong>pile</strong>s. This ismade possible by <strong>the</strong> use <strong>of</strong> a very general <strong>dynamic</strong> model <strong>for</strong> <strong>the</strong> representation <strong>of</strong> <strong>stiffness</strong>impedance matrices, <strong>the</strong> hidden state variable model (Sec. 3.1). The parameters appearing inthis model are <strong>the</strong>n fitted using an extensive database <strong>of</strong> full coupled FE-BE computations<strong>of</strong> soil-<strong>pile</strong> systems (Sec. 3.2). The sway and rocking <strong>of</strong> <strong>the</strong> foundation are accounted <strong>for</strong>, ina large range <strong>of</strong> parameters <strong>of</strong> <strong>the</strong> soil and <strong>pile</strong>s, and <strong>the</strong> <strong><strong>for</strong>mulas</strong> are given independently<strong>for</strong> floating (Sec. 4.1) and end-bearing <strong>pile</strong>s (Sec. 4.2). Fur<strong>the</strong>r, <strong>for</strong> <strong>the</strong> seismic design <strong>of</strong> abuilding, <strong>the</strong> use <strong>of</strong> <strong>the</strong> simple <strong><strong>for</strong>mulas</strong> ra<strong>the</strong>r than a full computational model is shown toinduce little error on <strong>the</strong> evaluation <strong>of</strong> <strong>the</strong> response spectra and time histories (Sec. 5).The reader interested in a fast use <strong>of</strong> <strong>the</strong> <strong><strong>for</strong>mulas</strong> can refer directly, <strong>for</strong> a floating <strong>pile</strong>group on homogeneous half-space (respectively, an end-bearing <strong>pile</strong> group) to Table II (resp.,Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
SIMPLE FORMULAS FOR THE DYNAMIC STIFFNESS OF PILE GROUPS 3Figure 1. Definition <strong>of</strong> <strong>pile</strong> group foundation.Table IV), <strong>the</strong> coefficients <strong>of</strong> which are to be used in Eq. (9) and (8) (resp., Eq. (14) and (10)).In <strong>the</strong>se equations, both <strong>the</strong> <strong>dynamic</strong> <strong>stiffness</strong> and <strong>the</strong> frequency are normalized, as describedin Sec. 2.2.2. THE IMPEDANCE MATRIX OF A PILE GROUP: DEFINITION AND NOTATIONSIn this section, we introduce <strong>the</strong> main notations and define <strong>the</strong> impedance matrix <strong>of</strong> a <strong>pile</strong>group. The normalizations, that will be used throughout <strong>the</strong> paper, <strong>of</strong> both <strong>the</strong> impedanceand <strong>the</strong> frequency, are also introduced.2.1. NotationsIn all <strong><strong>for</strong>mulas</strong>, <strong>the</strong> indices “s” and “p” will refer to <strong>the</strong> soil and <strong>the</strong> <strong>pile</strong>s, respectively. Whenconsidering two layers <strong>of</strong> soil, <strong>the</strong> top layer will still be denoted “s” while <strong>the</strong> bedrock willbe denoted “b”. E s , ν s , G s , and ρ s (respectively, E p , ν p , G p , ρ p , and E b , ν b , G b , ρ b ) hencedenote Young’s modulus, Poisson’s ratio, <strong>the</strong> shear modulus, and <strong>the</strong> unit mass <strong>of</strong> <strong>the</strong> soil(respectively, <strong>of</strong> <strong>the</strong> <strong>pile</strong>s, and <strong>of</strong> <strong>the</strong> bedrock). V s and β s (resp., V b and β b ) denote <strong>the</strong> shearwave velocity and <strong>the</strong> hysteretic damping <strong>of</strong> <strong>the</strong> soil (resp., <strong>of</strong> <strong>the</strong> bedrock). All <strong>the</strong> <strong>pile</strong>s ina group are supposed identical, with a diameter d, a length l p , an inertial moment I p , and<strong>the</strong>y are separated from each o<strong>the</strong>r by a distance s (see Fig. 1). They are rigidly attached toa mass-less square cap with a half-width B f , which is supposed to have no contact with <strong>the</strong>soil. Fur<strong>the</strong>r we define L 0 = (E p I p /E s ) 0.25 , closely related to <strong>the</strong> critical <strong>pile</strong> length definedby several authors [25, 26] and an equivalent radius <strong>for</strong> <strong>the</strong> cap R f = 2B f / √ π.Two cases will be considered in this paper: (1) floating <strong>pile</strong> <strong>groups</strong> on homogeneous halfspace,and (2) end-bearing <strong>pile</strong> <strong>groups</strong>. In <strong>the</strong> <strong>for</strong>mer case, <strong>the</strong> soil is a homogeneous half-space,while, in <strong>the</strong> latter case, <strong>the</strong> soil is composed <strong>of</strong> a layer <strong>of</strong> thickness H, resting over a bedrock,in which <strong>the</strong> tips <strong>of</strong> <strong>the</strong> <strong>pile</strong>s are embedded (l p > H).2.2. Impedance matrixThe impedance matrix or <strong>dynamic</strong> <strong>stiffness</strong> matrix Z(ω) <strong>of</strong> a <strong>pile</strong> group relates <strong>the</strong> vector <strong>of</strong><strong>for</strong>ces and moments applied on <strong>the</strong> rigid cap at <strong>the</strong> top <strong>of</strong> <strong>the</strong> <strong>pile</strong>s to <strong>the</strong> resulting vector <strong>of</strong>Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
4 R. TAHERZADEH, D. CLOUTEAU & R. COTTEREAUdisplacements and rotations at <strong>the</strong> same point. Since <strong>the</strong> rigid cap is supposed to be square,symmetry considerations ensure that <strong>the</strong> impedance matrix, in <strong>the</strong> basis <strong>of</strong> <strong>the</strong> rigid bodymovements <strong>of</strong> <strong>the</strong> <strong>pile</strong> cap, is not full. More precisely, only <strong>the</strong> diagonal terms, and <strong>the</strong> couplingterms between <strong>the</strong> sway along one <strong>of</strong> <strong>the</strong> axes <strong>of</strong> symmetry and <strong>the</strong> rotation around <strong>the</strong> o<strong>the</strong>rone, are non-null. Besides, <strong>the</strong> two horizontal axes <strong>of</strong> symmetry are equivalent, so that <strong>the</strong>corresponding terms in <strong>the</strong> impedance matrix are equal. Hence, only five elements should beconsidered: horizontal, rocking, pumping, torsion, and horizontal-rocking coupling. Finally, asis usually done in earthquake engineering, we disregard <strong>the</strong> pumping and torsion terms as lesssignificant, and <strong>the</strong>re<strong>for</strong>e write <strong>the</strong> impedance as[ ]Zh (ω) ZZ(ω) = hr (ω), (1)Z hr (ω) Z r (ω)where Z h (ω), Z r (ω) and Z hr (ω) are, respectively, <strong>the</strong> horizontal, rocking, and coupling elements.To simplify <strong>the</strong> comparisons between different soils and/or foundation sizes, it is customaryin foundation design [27, 29] to normalize <strong>the</strong> impedance matrix by a constant static <strong>stiffness</strong>matrix K 0 , given here by[ ]Gs RK 0 = f 00 G s Rf3 . (2)It is also classical to use a dimensionless frequency a 0 = ωR f /V s , where ω is <strong>the</strong> circularfrequency. The impedance matrix is finally writtenZ(a 0 ) = K 1/20¯Z(a 0 )K 1/20 , (3)where <strong>the</strong> tilde symbol¯refers to normalised quantities.Note that <strong>the</strong> imaginary part <strong>of</strong> <strong>the</strong> impedance matrix is sometimes plotted in <strong>the</strong> literatureafter a division by ω. While this is interesting in <strong>the</strong> simple cases where that imaginary partremains more or less linear, such as <strong>for</strong> circular rigid foundations over homogeneous soils, it isnot relevant here, so that that custom will not be followed in <strong>the</strong> plots <strong>of</strong> this paper.3. PRINCIPLE OF THE COMPUTATION OF SIMPLE FORMULASAs described in <strong>the</strong> introduction, <strong>the</strong>re is an engineering need <strong>for</strong> simple <strong>for</strong>mulations <strong>of</strong> <strong>the</strong>impedance matrix <strong>of</strong> large <strong>pile</strong> <strong>groups</strong>. The goal is to avoid using complicated and timeconsumingcomputational models, while retaining <strong>the</strong>ir accuracy. The usual approach to <strong>the</strong>derivation <strong>of</strong> <strong>the</strong>se simple <strong>for</strong>mulations consists in two steps:1. choose a structure <strong>for</strong> <strong>the</strong> <strong>for</strong>mulation, usually based on physical considerations andsimplifying assumptions2. set <strong>the</strong> values <strong>of</strong> <strong>the</strong> parameters <strong>for</strong> a range <strong>of</strong> soil and <strong>pile</strong> group characteristics, usuallyprovided in a tabulized <strong>for</strong>m.In <strong>the</strong> literature (see <strong>the</strong> introduction <strong>for</strong> references), <strong>the</strong> first step is basically a decision <strong>of</strong> <strong>the</strong>designer, which, however, conditions in a large part <strong>the</strong> quality <strong>of</strong> <strong>the</strong> <strong>for</strong>mulation. We <strong>the</strong>re<strong>for</strong>echoose here a more rational approach, in which no specific structure is chosen a priori, andmake use <strong>of</strong> <strong>the</strong> so-called ”hidden variables model” <strong>of</strong> <strong>the</strong> impedance matrix [35, 36]. Theidentification <strong>of</strong> <strong>the</strong> parameters <strong>of</strong> <strong>the</strong> final <strong><strong>for</strong>mulas</strong> is <strong>the</strong>n per<strong>for</strong>med by regression from aCopyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
SIMPLE FORMULAS FOR THE DYNAMIC STIFFNESS OF PILE GROUPS 5database <strong>of</strong> FE-BE computations. Note that <strong>the</strong> hidden state variable model that is chosenhere might appear more ma<strong>the</strong>matically than previous <strong>for</strong>mulations found in <strong>the</strong> literature.However, remember that this structure is just an intermediate state that allows to find <strong>the</strong>final <strong><strong>for</strong>mulas</strong> that will eventually be used by <strong>the</strong> engineers. In section 4, we will see someanalogies between <strong>the</strong> <strong><strong>for</strong>mulas</strong> proposed, and mechanical systems created as sets <strong>of</strong> spring,dampers and masses. The differences with lumped parameters is that with <strong>the</strong> hidden statevariables model, <strong>the</strong> equivalent mechanical model comes out naturally as a consequence <strong>of</strong> <strong>the</strong>regression, ra<strong>the</strong>r than being chosen a priori.3.1. The hidden variables modelThe construction <strong>of</strong> <strong>the</strong> hidden variables model <strong>of</strong> an impedance matrix is based on <strong>the</strong>supposition that, besides <strong>the</strong> n Γ physical DOFs on which <strong>the</strong> impedance is defined (typically<strong>the</strong> rigid-body modes <strong>of</strong> <strong>the</strong> cap <strong>of</strong> <strong>the</strong> <strong>pile</strong> group), <strong>the</strong>re exists n I additional DOFs thatrepresent some internal resonance phenomena inside <strong>the</strong> soil and <strong>the</strong> <strong>pile</strong> group. The resonancemodes corresponding to <strong>the</strong>se DOFs cannot be physically identified, as only <strong>the</strong>ir influence on<strong>the</strong> impedance matrix is observable, so <strong>the</strong> DOFs are referred to as “hidden”, or ”inner”.With respect to <strong>the</strong> n = n Γ + n I DOFs, matrices <strong>of</strong> mass M, damping D, and <strong>stiffness</strong> Kcan <strong>the</strong>n be identified, and <strong>the</strong> <strong>dynamic</strong> <strong>stiffness</strong> matrix S(a 0 ) is defined asS(a 0 ) = (K − a 2 0M) + ia 0 C. (4)The impedance matrix corresponding to <strong>the</strong> hidden variables model is <strong>the</strong>n <strong>the</strong> condensationon <strong>the</strong> n Γ physical DOFs <strong>of</strong> <strong>the</strong> <strong>stiffness</strong> matrix S(a 0 ). More specifically, introducing <strong>the</strong> blockdecomposition <strong>of</strong> Eq. (4),[SΓ (a 0 ) S c (a 0 )S T c (a 0 ) S I (a 0 )]=<strong>the</strong> impedance matrix is defined as([ ] [ ]) [ ]KΓ K cK T − a 2 MΓ M c CΓ Cc K0 I M T + ia cc M 0I C T , (5)c C IZ(a 0 ) = S Γ (a 0 ) − S c (a 0 )S −1I(a 0 )S T c (a 0 ). (6)As <strong>the</strong> hidden variables are not necessarily physical DOFs, but ra<strong>the</strong>r state variables in<strong>the</strong> background <strong>of</strong> <strong>the</strong> physical model, <strong>the</strong> matrices M, D, and K are really generalized mass,damping and <strong>stiffness</strong> matrices and do not correspond a priori with <strong>the</strong> classical mass, dampingand <strong>stiffness</strong> matrices, or to those obtained through <strong>the</strong> application <strong>of</strong> some modal reductiontechnique. Ano<strong>the</strong>r equivalent <strong>for</strong>m <strong>of</strong> <strong>the</strong> hidden variables model can be derived [35], where<strong>the</strong> hidden parts <strong>of</strong> <strong>the</strong> matrices are diagonal, and with no coupling in mass. In that case, <strong>the</strong>impedance can be written∑n hZ(a 0 ) = (K Γ − a 2 (ia 0 C l c + K l0M Γ ) + ia 0 C Γ −c)(ia 0 C l c + K l c) T(kI l − a2 0 ml I ) + ia 0c l Iwhere C l c and K l c are <strong>the</strong> l th columns <strong>of</strong> C c and K c , and m l I , cl I and kl I are <strong>the</strong> diagonalelements <strong>of</strong> M I , C I and K I .The main interest <strong>of</strong> this hidden variables model is its generality. Its structure makes itsuitable <strong>for</strong> <strong>the</strong> representation <strong>of</strong> any type <strong>of</strong> impedance matrix, provided that an appropriatenumber <strong>of</strong> hidden variables is used. Note that <strong>the</strong> numerical identification <strong>of</strong> <strong>the</strong> matrices M,l=1(7)Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
6 R. TAHERZADEH, D. CLOUTEAU & R. COTTEREAUC, and K is entirely per<strong>for</strong>med from <strong>the</strong> knowledge only <strong>of</strong> <strong>the</strong> impedance matrix, and that<strong>the</strong> number <strong>of</strong> hidden variables can be automatically chosen based on a precision criteria <strong>for</strong><strong>the</strong> approximation <strong>of</strong> <strong>the</strong> impedance matrix [39, 35, 36].Contrary to <strong>the</strong> lumped-parameter models <strong>of</strong> <strong>the</strong> impedance matrix [28], in which <strong>the</strong>identification <strong>of</strong> <strong>the</strong> mechanical elements yields negative values <strong>of</strong> <strong>the</strong> springs, dashpots, and/ormasses, in <strong>the</strong> hidden state variable model, <strong>the</strong> causality and stability <strong>of</strong> <strong>the</strong> soil impedancematrix are directly related to <strong>the</strong> positivity <strong>of</strong> M, K and C. In o<strong>the</strong>r words, in comparisonwith lumped parameter models, <strong>the</strong> diagnosis <strong>of</strong> unphysical models is very natural .In <strong>the</strong> next section, <strong>the</strong> numerical method that is used to derive <strong>the</strong> reference impedancematrices, and to identify <strong>the</strong> parameters <strong>of</strong> <strong>the</strong> <strong><strong>for</strong>mulas</strong>, is described. The methodology <strong>for</strong><strong>the</strong> identification <strong>of</strong> <strong>the</strong> hidden variables model <strong>of</strong> a given impedance matrix is also describedin App. 6.3.2. The reference FE-BE modelWe suppose, <strong>for</strong> <strong>the</strong> reference computations, that both <strong>the</strong> soil and <strong>the</strong> <strong>pile</strong>s behave linearlyand that <strong>the</strong> contact between <strong>the</strong> <strong>pile</strong>s and <strong>the</strong> soil is continuous in all directions, without anyslippage or gap. The elasto<strong>dynamic</strong> equations are <strong>the</strong>re<strong>for</strong>e linear. The numerical approachused to derive <strong>the</strong> reference results <strong>for</strong> <strong>the</strong> calibration <strong>of</strong> <strong>the</strong> simple <strong>for</strong>mulations is basedon an efficient FE-BE coupling technique that is described in detail in [30, 31] and is brieflyrecalled below.The soil is separated into two blocks: one, bounded and containing <strong>the</strong> <strong>pile</strong>s, which is modeledby <strong>the</strong> FE method, and <strong>the</strong> o<strong>the</strong>r, surrounding <strong>the</strong> previous one, which is modeled by <strong>the</strong> BEmethod (see Fig. 2). Within <strong>the</strong> FE block, <strong>the</strong> <strong>pile</strong>s are modeled as Bernouilli beam elements.The two blocks are <strong>the</strong>n assembled using <strong>the</strong> Craig-Bampton coupling technique [32], so as tolower <strong>the</strong> computational cost, which may reach high levels <strong>for</strong> large <strong>pile</strong> <strong>groups</strong>. This numericalmodel was already validated <strong>for</strong> <strong>stiffness</strong> problems taken from <strong>the</strong> literature (in particular [2])and <strong>the</strong> results are given in [31]. However, <strong>the</strong>se validation results only concerned floating<strong>pile</strong> <strong>groups</strong> on homogeneous half-spaces so that we present here a comparison, on a particularexample, <strong>of</strong> <strong>the</strong> FE-BE model with <strong>the</strong> BE approach described in [34].We <strong>the</strong>re<strong>for</strong>e consider a 4 × 9 <strong>pile</strong> group embedded in a soil with two layers (see Fig. 2).The <strong>pile</strong>s have a Young’s modulus <strong>of</strong> 25 GPa, a diameter <strong>of</strong> d = 1.3 m and are separated bys = 2.6 m. The first layer <strong>of</strong> soil is H = 9.5 m-thick, and is <strong>for</strong>med <strong>of</strong> a very s<strong>of</strong>t saturatedorganic clay with S-wave velocity V s = 80 m/s, unit mass ρ s = 1.5 Mg/m 3 and Poisson’s ratioν s = 0.49. The lower layer <strong>of</strong> soil is a stiff sand with S-wave velocity V d = 300 m/s, unit massρ d = 2 Mg/m 3 and Poisson’s ratio ν d = 0.4, in which <strong>the</strong> <strong>pile</strong>s penetrate 6 m. In both layers<strong>the</strong> hysteretic damping is taken as β s = β d = 0.05. As seen on Fig. 2, <strong>the</strong> agreement between<strong>the</strong> results in <strong>the</strong> two numerical approaches is very good.It should be noted that <strong>the</strong> frequency-dependance <strong>of</strong> <strong>pile</strong> <strong>groups</strong> is particularly sensitive to<strong>the</strong> number <strong>of</strong> <strong>pile</strong>s and to its character <strong>of</strong> floating or end-bearing. The <strong>dynamic</strong> <strong>stiffness</strong> <strong>of</strong>single <strong>pile</strong>s and <strong>pile</strong> <strong>groups</strong> with a small number <strong>of</strong> <strong>pile</strong>s is nearly independent <strong>of</strong> frequency [33],while that <strong>of</strong> larger <strong>pile</strong> <strong>groups</strong> may show large variations with frequency. Likewise, <strong>the</strong> behavior<strong>of</strong> end-bearing <strong>pile</strong> <strong>groups</strong> is much more erratic with frequency than that <strong>of</strong> floating <strong>pile</strong> <strong>groups</strong>on homogeneous half-space. These physical results are retrieved with <strong>the</strong> FE-BE approach andan example <strong>of</strong> such comparison is shown in Fig. 3. These results were obtained considering <strong>the</strong>sample number 4 in Tables I and III.Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
SIMPLE FORMULAS FOR THE DYNAMIC STIFFNESS OF PILE GROUPS 7Dynamic <strong>stiffness</strong> [N/m]0−2−4−60 2 4 6 8 102 x Frequency [Hz]Damping [N/m]64200 2 4 6 8 108 x Frequency [Hz]Figure 2. FE model (left) <strong>of</strong> <strong>the</strong> 4 × 9 <strong>pile</strong> group within a block <strong>of</strong> soil andcomparison <strong>of</strong> <strong>the</strong> real (right, up) and imaginary (right, down) parts <strong>of</strong> <strong>the</strong>horizontal impedance computed using <strong>the</strong> FE-BE model (solid line) and <strong>the</strong> BEmodel (dashed line) [34].3060Normalized Real Part [−]20100−10−20Normalized Imaginary Part [−]5040302010−300 1 2 3 4 5 6Dimensionless Frequency [−]00 1 2 3 4 5 6Dimensionless Frequency [−]Figure 3. Real (left) and imaginary (right) parts <strong>of</strong> <strong>the</strong> normalized horizontalimpedance matrix, <strong>for</strong> floating (solid line) and end-bearing (dashed line) <strong>pile</strong><strong>groups</strong>.It is also interesting to note, on Fig. 3, <strong>for</strong> <strong>the</strong> end-bearing <strong>pile</strong> group, that <strong>the</strong> imaginarypart <strong>of</strong> <strong>the</strong> impedance (it is also true <strong>for</strong> <strong>the</strong> rocking term, not shown here) present a small andalmost constant value below some cut-<strong>of</strong>f frequency, which is <strong>the</strong> resonance frequency <strong>of</strong> <strong>the</strong>top layer <strong>of</strong> soil. Indeed, <strong>for</strong> very low frequencies, surface waves cannot build up in that toplayer and take energy away from <strong>the</strong> foundation, so that <strong>the</strong> radiation damping is very low.Above that cut-<strong>of</strong>f frequency, a large peak can be observed on <strong>the</strong> imaginary part (with <strong>the</strong>real part almost cancelling), indicating a resonance within <strong>the</strong> soil, that tends to soak energyaway from <strong>the</strong> foundation.4. COMPUTATION OF SIMPLE FORMULAS FOR PILE GROUPSIn this section, we present <strong>the</strong> derivation <strong>of</strong> <strong>the</strong> simple <strong><strong>for</strong>mulas</strong> in <strong>the</strong> cases <strong>of</strong> <strong>the</strong> floating<strong>pile</strong> <strong>groups</strong> on homogeneous half-space and end-bearing <strong>pile</strong> <strong>groups</strong>, and using <strong>the</strong> ideasCopyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
8 R. TAHERZADEH, D. CLOUTEAU & R. COTTEREAU1050Normalized Real Part [−]50−5−10Normalized Imaginary Part [−]40302010−150 1 2 3 4 5 6Dimensionless Frequency [−]00 1 2 3 4 5 6Dimensionless Frequency [−]Figure 4. Real (left) and imaginary (right) parts <strong>of</strong> <strong>the</strong> horizontal impedancematrix <strong>for</strong> different <strong>pile</strong> separations: s/d = 2 (solid line), s/d = 2.5 (dashed line)and s/d = 3.5 (solid-dashed line). The figures correspond to a 15 × 15 <strong>pile</strong> groupwith E p/E s = 300 and R f /l p = 1.1.2060Normalized Real Part [−]100−10−20−30Normalized Imaginary Part [−]5040302010−400 1 2 3 4 5 6Dimensionless Frequency [−]00 1 2 3 4 5 6Dimensionless Frequency [−]Figure 5. Real (left) and imaginary (right) parts <strong>of</strong> <strong>the</strong> rocking term <strong>of</strong> <strong>the</strong><strong>dynamic</strong> <strong>stiffness</strong> matrix <strong>for</strong> different <strong>pile</strong> separations: R f /l p = 0.7 (solid line),R f /l p = 0.65 (dashed line) and R f /l p = 0.55 (solid-dashed line). The figurescorrespond to a 14 × 14 <strong>pile</strong> group with E p/E s = 375 and s/d = 2.discussed above. Depending on <strong>the</strong> type <strong>of</strong> <strong>pile</strong> group, and on <strong>the</strong> type <strong>of</strong> element <strong>of</strong> <strong>the</strong>impedance matrix, more or less hidden variables are necessary to describe its behavior, and,correspondingly, more or less parameters are needed in <strong>the</strong> <strong><strong>for</strong>mulas</strong>.4.1. Floating <strong>pile</strong> <strong>groups</strong>We first consider floating <strong>pile</strong> <strong>groups</strong> embedded in a homogeneous half-space. In that case,<strong>the</strong> variation <strong>of</strong> <strong>the</strong> <strong>dynamic</strong> <strong>stiffness</strong> with <strong>the</strong> frequency is ra<strong>the</strong>r smooth, as seen in Fig. 4and 5. More precisely, <strong>the</strong> <strong>dynamic</strong> <strong>stiffness</strong> always has a parabolic variation while <strong>the</strong> dampingcoefficient is approximately linear. The parabolic decrease <strong>of</strong> <strong>the</strong> real part seems to indicatethat a mass remains entrapped between <strong>the</strong> <strong>pile</strong>s and vibrates in-phase with <strong>the</strong> cap.The hidden variables model predicts in all cases in <strong>the</strong> database (described in Table I) atwo-DOFs system, one <strong>for</strong> <strong>the</strong> sway and one <strong>for</strong> <strong>the</strong> rocking, and with no hidden variables. NoteCopyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
SIMPLE FORMULAS FOR THE DYNAMIC STIFFNESS OF PILE GROUPS 9Table I. The database <strong>of</strong> soil-<strong>pile</strong> group systems used to derive <strong>the</strong> simple<strong>for</strong>mulations <strong>for</strong> floating <strong>pile</strong> <strong>groups</strong> on homogeneous half-space. The range <strong>of</strong>parameters is 250 ≤ E p/E s ≤ 1500, 2 ≤ s/d p ≤ 3.6 and 0.55 ≤ R f /l p ≤ 2 andconstant hysteric damping β s = 0.05.Sample Piles E p d E s B f l p s V s[-] [GPa] [m] [GPa] [m] [m] [m] [m/s]1 8 × 8 30 1 0.1 15 18 3.3 1402 11 × 11 30 1 0.1 20 18 3.3 1403 16 × 16 30 1 0.1 30 18 3.3 1404 13 × 13 40 1 0.08 20 24 2.8 1305 13 × 13 30 1 0.08 20 24 2.8 1306 13 × 13 20 1 0.08 20 24 2.8 1307 15 × 15 25 1.3 0.08 20 18 2.5 1308 15 × 15 25 1 0.08 20 18 2.5 1309 15 × 15 25 0.7 0.08 20 18 2.5 13010 18 × 18 30 1 0.02 25 14 2.6 6011 18 × 18 30 1 0.08 25 14 2.6 13012 18 × 18 30 1 0.1 25 14 2.6 14013 16 × 16 30 1 0.08 15 28 2.0 13014 16 × 16 30 1 0.08 15 20 2.0 13015 16 × 16 30 1 0.08 15 14 2.0 130that <strong>the</strong> coupling term is negligible. On Fig. 6, a schematic drawing <strong>of</strong> a system correspondingto such impedance is presented. The superstructure is subjected to <strong>the</strong> seismic horizontal <strong>for</strong>cef s . The elements <strong>of</strong> <strong>the</strong> normalized impedance can <strong>the</strong>re<strong>for</strong>e be written⎧⎪⎨˜Z h (a 0 ) = (˜k h − a 2 0 ˜m h ) + ia 0˜c h˜Z r (a 0 ) = (˜k r − a⎪⎩2 0 ˜m r ) + ia 0˜c r˜Z sr (a 0 ) = 0, (8)where <strong>the</strong> values <strong>of</strong> ˜k h , ˜c h , ˜m h , ˜k r , ˜c r and ˜m r depend on <strong>the</strong> case considered. Remember that<strong>the</strong> definition <strong>of</strong> <strong>the</strong> normalized frequency a 0 is given in Sec. 2.2 and that <strong>the</strong> normalizedvalues in <strong>the</strong>se <strong><strong>for</strong>mulas</strong> (8) must be scaled by <strong>the</strong> static <strong>stiffness</strong> to yield <strong>the</strong> actual value <strong>of</strong><strong>the</strong> impedance matrix, as described in Sec. 2.2.In previous works, <strong>the</strong> leading parameters <strong>for</strong> this type <strong>of</strong> <strong>pile</strong> <strong>groups</strong> were identified to be<strong>the</strong> ratio <strong>of</strong> Young’s moduli E p /E s and <strong>the</strong> normalized separation <strong>of</strong> <strong>the</strong> <strong>pile</strong>s s/d [2, 37], or <strong>the</strong>factor L 0 = (E p I p /E s ) 0.25 related to <strong>the</strong> active <strong>pile</strong> length [38, 25, 26]. We decide here to use asleading parameters <strong>the</strong> normalized radius <strong>of</strong> <strong>the</strong> foundation R f /l p and a normalized active <strong>pile</strong>length ratio L 0 /s. We <strong>the</strong>re<strong>for</strong>e provide equations <strong>of</strong> <strong>the</strong> parameters χ ∈ {˜k h , ˜c h , ˜m h , ˜k r , ˜c r , ˜m r }Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
10 R. TAHERZADEH, D. CLOUTEAU & R. COTTEREAUFigure 6. A schematic drawing <strong>of</strong> a simple model <strong>for</strong> floating <strong>pile</strong> group.in <strong>the</strong> <strong>for</strong>m( ) λ1( ) λ2 Rf L0χ = λ 0 , (9)l p swith <strong>the</strong> values <strong>of</strong> λ 0 , λ 1 and λ 2 being provided <strong>for</strong> each <strong>of</strong> <strong>the</strong> parameters. A multiple regressionanalysis was <strong>the</strong>n conducted with respect to <strong>the</strong> two quantities R f /l p and L 0 /s, and lead to<strong>the</strong> values described in Table II. The regression coefficient R is also indicated in <strong>the</strong> same tableto provide an indicator <strong>of</strong> <strong>the</strong> accuracy <strong>of</strong> <strong>the</strong> regression analysis.In general terms, <strong>the</strong> <strong><strong>for</strong>mulas</strong> in Table II corroborate <strong>the</strong> observed results that, <strong>for</strong> shortseparations between <strong>the</strong> <strong>pile</strong>s, and <strong>for</strong> weak soils, both normalized <strong>dynamic</strong> <strong>stiffness</strong> anddamping increase. Note that, as indicated by <strong>the</strong> zeros in Table II, <strong>the</strong> influence <strong>of</strong> <strong>the</strong> ratioR f /l p on <strong>the</strong> horizontal impedance is negligible, while it is ra<strong>the</strong>r important <strong>for</strong> <strong>the</strong> rockingterm. Besides <strong>the</strong> uni<strong>for</strong>m presentation in Table II, <strong>the</strong> reader may also find an expanded,non-normalized, version <strong>of</strong> <strong>the</strong> same <strong><strong>for</strong>mulas</strong> in App. 6, <strong>for</strong> easier reading.4.2. End-bearing <strong>pile</strong> <strong>groups</strong>We <strong>the</strong>n consider end-bearing <strong>pile</strong> <strong>groups</strong>. As stated earlier, <strong>the</strong>ir <strong>dynamic</strong>al behavior is muchmore complicated than that <strong>of</strong> floating <strong>pile</strong> <strong>groups</strong> on homogeneous half-space. The structure<strong>of</strong> <strong>the</strong> approximation <strong>for</strong> <strong>the</strong> impedance matrix is <strong>the</strong>re<strong>for</strong>e difficult to guess a priori and weuse <strong>the</strong> hidden variables model in a very general setting. Note that, as <strong>the</strong> coupling term isnegligible in <strong>the</strong> cases considered, <strong>the</strong> hidden variables model was identified independently on<strong>the</strong> horizontal and rocking terms <strong>of</strong> <strong>the</strong> impedance matrix.The identification <strong>of</strong> <strong>the</strong> hidden variables model <strong>for</strong> all <strong>the</strong> cases in <strong>the</strong> database describedin Table IV suggest <strong>the</strong> consideration <strong>of</strong> three hidden variables <strong>for</strong> <strong>the</strong> sway and none <strong>for</strong> <strong>the</strong>rocking. Besides, no coupling in <strong>the</strong> <strong>stiffness</strong> <strong>for</strong> <strong>the</strong> first hidden variable and no coupling in <strong>the</strong>damping <strong>for</strong> <strong>the</strong> two o<strong>the</strong>rs seemed to be necessary. The chosen structure <strong>for</strong> <strong>the</strong> end-bearingCopyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
SIMPLE FORMULAS FOR THE DYNAMIC STIFFNESS OF PILE GROUPS 11Table II. Coefficients <strong>for</strong> <strong>the</strong> horizontal and rocking elements <strong>of</strong> <strong>the</strong> impedance <strong>of</strong>a floating <strong>pile</strong> <strong>groups</strong> on homogeneous half-space.( ) λ1 Rf(χ = λL0) λ20 l p s λ0 λ 1 λ 2 R [%]˜k h 6.8 0 0.3 80˜c h 5 0 0.5 90˜m h 0.4 0 1.6 86˜k r 8 -0.6 0.4 83˜c r 5 -0.5 0.2 72˜m r = ˜m h leq/4 2 0.7 -1 0.4 96<strong>pile</strong> <strong>groups</strong> is <strong>the</strong>re<strong>for</strong>e written, as a special case <strong>of</strong> Eq. (7) <strong>for</strong> <strong>the</strong> hidden variables model,⎧⎪⎨˜Z h (a 0 ) = (˜k h − a 2 a0 ˜m h ) + ia 0˜c h +2 0˜c2 1˜k−2 2˜k−2 3(˜k 1−a 2 0 ˜m1)+ia0˜c1 (˜k 2−a 2 0 ˜m2)+ia0˜c2 (˜k 3−a 2 0˜Z ˜m3)+ia0˜c3r (a 0 ) = (˜k r − a 2 0 ˜m r ) + ia 0˜c r,⎪⎩ ˜Z sr (a 0 ) = 0(10)and represented as a set <strong>of</strong> mass, springs, and dampers in Fig. 7. The previous observation <strong>for</strong><strong>the</strong> coupling with <strong>the</strong> hidden variables can be translated in Fig. 7 by <strong>the</strong> fact <strong>the</strong> mass m 1 islinked to <strong>the</strong> foundation by a dashpot while <strong>the</strong> masses m 2 and m 3 are linked to it throughsprings. This fact arises from <strong>the</strong> presence <strong>of</strong> <strong>the</strong> cut-<strong>of</strong>f frequency <strong>of</strong> <strong>the</strong> top layer <strong>of</strong> soil thatwas discussed in Sec. 3.2.More physical remarks can be made in <strong>the</strong> different frequency ranges defined by <strong>the</strong>resonance frequencies a 0α <strong>of</strong> <strong>the</strong> masses m α representing <strong>the</strong> hidden variables. In <strong>the</strong> lowfrequency range (a 0 ≪ a 01 ), a first-order expansion gives˜Z h (a 0 ) = ˜k 0 + ia 0 (˜c h + ˜c 2 + ˜c 3 ). (11)It is worth noticing that <strong>the</strong> slope <strong>of</strong> <strong>the</strong> imaginary part ˜c 0 + ˜c 1 + ˜c 2 + ˜c 3 is not small sinceit allows to quickly reach <strong>the</strong> level <strong>of</strong> <strong>the</strong> hysteretical damping. In <strong>the</strong> range <strong>of</strong> resonance <strong>of</strong>mass m 1 (a 0 − a 01 ≪ 2ζ 1 a 01 , with ζ α = c α /(2 √ k α m α )), and supposing that all <strong>the</strong> resonancefrequencies are far enough from each o<strong>the</strong>r (a 0 − a 02 ≫ 2ζ 2 a 02 and a 0 − a 03 ≫ 2ζ 3 a 03 ), onehas[˜Z h (a 0 ) = ˜k h + ˜k 1 − ˜k a 2 022a 2 02 − − ˜k a 2 ]033 a2 0 a 2 03 − a2 0[ ( ) a2 2 ( )+ ia 0 ˜c h − ˜c 1 + ˜c 02a2 2]2a 2 02 − + ˜c 033 a2 0 a 2 03 − (12)a2 0which means that around a 01 , <strong>the</strong> mass m 1 has <strong>the</strong> same displacement as <strong>the</strong> foundation, sothat <strong>the</strong>re is no damping contribution from ˜c 1 . For a 0 ≪ a 02 , masses m 2 and m 3 are also linkedto <strong>the</strong> foundation but <strong>the</strong> dashpots ˜c 2 and ˜c 3 introduce some damping. The equivalent slopeCopyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
12 R. TAHERZADEH, D. CLOUTEAU & R. COTTEREAUFigure 7. A schematic drawing <strong>of</strong> a simple model <strong>for</strong> end bearing <strong>pile</strong> group.around a 01 tends to ˜c 0 + ˜c 2 + ˜c 3 = β eq /a 01 which is actually small as expected to model <strong>the</strong>sole hysteretical damping β eq . Usually ˜c 1 ≫ a 01 β eq . In <strong>the</strong> range <strong>of</strong> resonance <strong>of</strong> <strong>the</strong> mass m 2(an equivalent <strong>for</strong>mula can be derived <strong>for</strong> mass m 3 ), a large imaginary part is brought on byk α2ζ α, which corresponds to <strong>the</strong> peaks observed in Fig. 3. Finally, at high frequency (a 0 ≫ a 03 ),one has()˜Z h (a 0 ) = ˜k h − ˜c2 1− a 2 0 ˜m h + ia 0˜c h (13)˜m 1which classically corresponds to all <strong>the</strong> masses m 1 , m 2 and m 3 being fixed. One can see ˜c 1as <strong>the</strong> radiative damping which occurs only above a 01 since <strong>for</strong> this frequency we have shownthat <strong>the</strong> damping is only β eq /a 01 . Thus this model reproduces <strong>the</strong> cut-<strong>of</strong>f frequency at <strong>the</strong>resonance frequency <strong>of</strong> <strong>the</strong> layer.Once <strong>the</strong> structure <strong>of</strong> <strong>the</strong> approximation has been decided, a multiple regression analysisis per<strong>for</strong>med on <strong>the</strong> same leading parameters as be<strong>for</strong>e, plus <strong>the</strong> ratio (ρ b V b )/(ρ s V s ) to yield<strong>the</strong> <strong><strong>for</strong>mulas</strong> presented in Table IV. Note that several coefficients appear as zeros in <strong>the</strong> table,which means that <strong>the</strong> parameters modeled do not have any influence on <strong>the</strong> <strong>for</strong>mula. Note alsothat, as be<strong>for</strong>e, <strong>the</strong> <strong><strong>for</strong>mulas</strong> are presented in a non-normalized manner in App. 6 <strong>for</strong> easierreading. The general <strong><strong>for</strong>mulas</strong> <strong>for</strong> <strong>the</strong> parameters areχ = λ 0(RfH) λ1( ) λ2( ) λ3 L0 ρb V b. (14)s ρ s V sIt is particularly interesting to note that, although <strong>the</strong> <strong><strong>for</strong>mulas</strong> were derived from ra<strong>the</strong>rma<strong>the</strong>matical considerations (<strong>the</strong> hidden variables model and a regression analysis), <strong>the</strong>yyield a very good evaluation <strong>of</strong> <strong>the</strong> resonance frequencies <strong>of</strong> <strong>the</strong> soil layer. Indeed, <strong>the</strong> firstfundamental frequency <strong>of</strong> <strong>the</strong> soil layer ω s 01 = 2πV s /(4H) and √ k 1 /m 1 = 1.4V s /H (see App. 6<strong>for</strong> non-normalized <strong><strong>for</strong>mulas</strong>) coincide. Likewise, <strong>the</strong> second fundamental frequency <strong>of</strong> <strong>the</strong> soilCopyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
SIMPLE FORMULAS FOR THE DYNAMIC STIFFNESS OF PILE GROUPS 13Table III. The database <strong>of</strong> soil-<strong>pile</strong> group systems used to derive <strong>the</strong> simple<strong>for</strong>mulations <strong>for</strong> end-bearing <strong>pile</strong> <strong>groups</strong>. The range <strong>of</strong> parameters is 125 ≤E p/E s ≤ 750, 2.8 ≤ s/d p ≤ 4.4, 1 ≤ R f /H ≤ 2.1 and 3 ≤ V b /V s ≤ 8 andconstant hysteretic damping β s = 0.05.Samples Piles E p d E s B f l p H V s V b s[-] [GPa] [m] [GPa] [m] [m] [m] [m/s] [m/s] [m]1 8 × 8 30 1 0.1 15 16 14 140 830 3.32 11 × 11 30 1 0.1 20 16 14 140 830 3.33 16 × 16 30 1 0.1 30 16 14 140 830 3.34 13 × 13 40 1 0.08 20 22 20 130 620 2.85 13 × 13 30 1 0.08 20 22 20 130 620 2.86 13 × 13 20 1 0.08 20 22 20 130 620 2.87 15 × 15 25 1.3 0.2 37 20 18 200 780 4.38 18 × 18 30 1 0.04 32 20 18 90 440 3.19 18 × 18 30 1 0.1 32 20 18 140 700 3.110 14 × 14 30 1 0.08 26 26 24 130 1000 3.311 14 × 14 30 1 0.08 26 20 18 130 1000 3.312 14 × 14 30 1 0.08 26 16 14 130 1000 3.313 13 × 13 30 1 0.06 26 18 16 110 620 3.614 13 × 13 30 1 0.06 26 18 16 110 620 3.615 13 × 13 30 1 0.06 26 18 16 110 620 3.6layer ω s 02 = 6πV s /(4H) is very well approximated by <strong>the</strong> third resonance <strong>of</strong> <strong>the</strong> simple model√k3 /m 3 = 5.1V s /H.5. IMPACT OF THE FORMULAS ON THE EVALUATION OF DESIGN QUANTITIESIn this last section, we discuss <strong>the</strong> accuracy <strong>of</strong> <strong>the</strong> proposed <strong><strong>for</strong>mulas</strong> on two practical cases.More particularly, <strong>the</strong> accuracy <strong>of</strong> <strong>the</strong> predicted transfer functions, spectral acceleration ontop <strong>of</strong> a building and relative displacement between top and bottom <strong>of</strong> <strong>the</strong> building, using<strong>the</strong> proposed <strong><strong>for</strong>mulas</strong>, is demonstrated. In a second test, we compare <strong>the</strong> accuracy <strong>of</strong> ourproposed <strong>for</strong>mula with ano<strong>the</strong>r one from <strong>the</strong> literature.5.1. Case 1For this validation, a 10 ×10 end-bearing <strong>pile</strong> group is used, with <strong>pile</strong>s with d p = 1 m, l p = 22m, s = 5 m, and connected by a 1.1 m-thick, rigid, cap with B f = 25 m. The mechanicalproperties <strong>of</strong> <strong>the</strong> <strong>pile</strong>s are E p = 30 GPa, ν p = 0.25, and ρ p = 2500 kg/m 3 . This <strong>pile</strong> <strong>groups</strong>tands in H = 20 m-thick soil layer, with properties E s = 60 MPa, ν s = 0.4 and ρ s = 1750Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
14 R. TAHERZADEH, D. CLOUTEAU & R. COTTEREAUTable IV. Coefficients <strong>for</strong> <strong>the</strong> horizontal and rocking elements <strong>of</strong> <strong>the</strong> impedance<strong>of</strong> an end-bearing <strong>pile</strong> group.( ) λ1 Rf(χ = λL0) λ2( ) λ3ρb V b0 H s ρ sV sλ 0 λ 1 λ 2 λ 3 R [%]˜k 0 = ˜k h − ˜k 2 − ˜k 3 10 0.5 0.35 0 93˜c h = ˜c 0 + ˜c 1 1 -0.5 0.5 0.5 60˜m h = ˜m 0 0.5 -1 0 0 82˜k 1 2.6 1 0 0 65˜c 1 1.9 -1.5 0 0 75˜m 1 1.4 -1 0 0 80˜k 2 1.25 0.35 -1 0.5 60˜c 2 0.04 0 -1 1 62˜m 2 0.08 -1 -1 0.5 80˜k 3 16.1 3 3 -0.5 75˜c 3 3 2 3 -1.5 70˜m 3 0.6 1 3 -0.5 75˜k r 15 0.5 1 1 98˜c r 17 0.5 2 -0.5 95˜m r = ˜m 0 leq/4 2 1.6 -1.5 1 -0.5 90kg/m 3 . The mechanical properties <strong>of</strong> <strong>the</strong> underlying half-space are E b = 1.5 GPa and ν b = 0.3and ρ b = 2000 kg/m 3 . The real and imaginary parts <strong>of</strong> <strong>the</strong> impedance are shown on Fig. 8, bothas computed using <strong>the</strong> numerical FE-BE model, and using <strong>the</strong> simple <strong><strong>for</strong>mulas</strong> <strong>of</strong> Eq. (10). Theagreement between <strong>the</strong> two approaches is good, in particular <strong>for</strong> <strong>the</strong> shaking term, considering<strong>the</strong> important variability in frequency. Note that <strong>the</strong> <strong>pile</strong> group considered here was not used<strong>for</strong> <strong>the</strong> regression analysis that determined <strong>the</strong> parameters in Table IVWe now turn to <strong>the</strong> observation <strong>of</strong> <strong>the</strong> accuracy <strong>of</strong> <strong>the</strong> proposed <strong>for</strong>mulations <strong>for</strong> <strong>the</strong>estimation <strong>of</strong> engineering quantities <strong>of</strong> interest. We <strong>the</strong>re<strong>for</strong>e consider a 60 m high-building(20 floors), with floors <strong>of</strong> 22.5 m ×22.5 m, and 6 columns ×6 columns. The slab weight perunit area is 500 kg/m 2 and <strong>the</strong> characteristics <strong>of</strong> <strong>the</strong> beams and columns are, respectively,EI = 5.1 MN.m 2 and EI = 1 MN.m 2 .We first consider <strong>the</strong> estimation <strong>of</strong> transfer functions in two different cases: (1) using <strong>the</strong>entire, 6 × 6, impedance matrix computed from <strong>the</strong> FE-BE model, and considering both <strong>the</strong>kinematic and inertial interaction, and (2) using only <strong>the</strong> horizontal and rocking elements <strong>of</strong><strong>the</strong> impedance matrix computed with <strong>the</strong> proposed <strong>for</strong>mula (10) and neglecting <strong>the</strong> kinematicinteraction. For both cases, <strong>the</strong> displacement field is decomposed on a basis that contains <strong>the</strong>rigid body modes <strong>of</strong> <strong>the</strong> building (l m ), which coincide with those <strong>of</strong> <strong>the</strong> foundation, and <strong>the</strong>flexible modes <strong>of</strong> <strong>the</strong> building on a rigid basis (φ n ):u(ω,x) = ∑ mc m (ω)l m (x) + ∑ nα n (ω)φ n (x) = [ c α ] { LΦ}(15)Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
SIMPLE FORMULAS FOR THE DYNAMIC STIFFNESS OF PILE GROUPS 15Real part [N/m]0−10 1 2 3 4 51 x Frequency [Hz]Real part [N/m]100 1 2 3 4 52 x Frequency [Hz]Imaginary part [N.m]100 1 2 3 4 52 x Frequency [Hz]Imaginary part [N.m]100 1 2 3 4 52 x Frequency [Hz]Figure 8. Comparison between <strong>the</strong> real (up) and imaginary (down) parts <strong>of</strong> <strong>the</strong>horizontal (left) and rocking (right) elements <strong>of</strong> <strong>the</strong> impedance matrix <strong>for</strong> a 10×10end-bearing <strong>pile</strong> group computed using <strong>the</strong> simplified <strong><strong>for</strong>mulas</strong> (10) (dashed line)and <strong>the</strong> FE-BE model (solid line).where L is <strong>the</strong> matrix <strong>of</strong> <strong>the</strong> rigid body modes <strong>of</strong> <strong>the</strong> structure and Φ is <strong>the</strong> matrix <strong>of</strong> <strong>the</strong>eigenmodes <strong>of</strong> <strong>the</strong> structure clamped at its base. The response <strong>of</strong> <strong>the</strong> structure, taking intoaccount soil-structure interaction, is <strong>the</strong>n computed using <strong>the</strong> following <strong>for</strong>mula([ Z(ω) 00 0]+ (1 + 2iβ)[ 0 00 Λ] [− ω 2 MΓM ΩΓM ΓΩI]) { c(ω)α(ω)} { } Z(ω)c0 (ω)=0where <strong>the</strong> diagonal matrix Λ contains <strong>the</strong> squares <strong>of</strong> <strong>the</strong> lowest circular frequencies <strong>of</strong> <strong>the</strong>structure on fixed base and I is <strong>the</strong> identity matrix arising from <strong>the</strong> orthogonality <strong>of</strong> <strong>the</strong>eigenmodes with respect to <strong>the</strong> mass matrix. Γ stands <strong>for</strong> <strong>the</strong> rigid body modes and Ω <strong>for</strong> <strong>the</strong>eigenmodes on fixed base, while c 0 is <strong>the</strong> kinematic interaction. The differences between <strong>the</strong>two models with respect to this <strong>for</strong>mulation are <strong>the</strong> impedance matrix Z(ω) and <strong>the</strong> kinematicinteraction factor takes equal to ∆u i (ω) with ∆ having null components but a unitary <strong>for</strong><strong>the</strong> sway term. Besides, it is worth noticing that <strong>the</strong> simplified model does not correspond to<strong>the</strong> physical model sketched on Fig. 7 subjected to an uni<strong>for</strong>m acceleration a i . Indeed, inertial<strong>for</strong>ces are not applied on mass m 1 , m 2 and m 3 since <strong>the</strong>se masses are in <strong>the</strong> soil and have <strong>the</strong>irinertial <strong>for</strong>ces already balanced in <strong>the</strong> soil.The resonance frequencies <strong>of</strong> <strong>the</strong> soil are computed at f s 01 = 1.55 Hz and f s 02 = 4.6 Hz.Assuming a horizontal harmonic base motion at <strong>the</strong> bedrock, <strong>the</strong> horizontal transfer functionat <strong>the</strong> free surface and at <strong>the</strong> top <strong>of</strong> <strong>the</strong> building are represented on Fig. 9. It clearly shows <strong>the</strong>effect <strong>of</strong> <strong>the</strong> soil-structure interaction, as well as <strong>the</strong> ability <strong>of</strong> <strong>the</strong> <strong><strong>for</strong>mulas</strong> (10) to compute<strong>the</strong> resonance frequency <strong>of</strong> <strong>the</strong> coupled system (<strong>the</strong> peaks <strong>of</strong> <strong>the</strong> dotted and dash-dotted lineson Fig. 9 coincide almost exactly).We <strong>the</strong>n consider two real recordings <strong>of</strong> earthquakes, with different frequency contents (seeFig. 10) and peak ground accelerations (PGA) at about 0.3 g. On Fig. 11 a comparison is given<strong>of</strong> <strong>the</strong> spectral acceleration on top <strong>of</strong> <strong>the</strong> building computed in <strong>the</strong> two cases considered earlier<strong>of</strong> <strong>the</strong> FE-BE model supposing inertial and kinematic interaction and <strong>the</strong> simple <strong><strong>for</strong>mulas</strong> (10)neglecting <strong>the</strong> kinematic interaction. On Fig. (12), a comparison is given <strong>of</strong> <strong>the</strong> time histories<strong>of</strong> <strong>the</strong> relative displacements between <strong>the</strong> top and <strong>the</strong> base <strong>of</strong> <strong>the</strong> building. On both figures,<strong>the</strong> agreement between <strong>the</strong> two approaches is very good.(16)Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
16 R. TAHERZADEH, D. CLOUTEAU & R. COTTEREAU1412Transfer Function [−]10864200 1 2 3 4 5Frequency [Hz]Figure 9. Transfer function at ground surface free field (solid line), <strong>of</strong> <strong>the</strong> structurewithout SSI (dashed line), and <strong>of</strong> <strong>the</strong> structure with SSI (dashed-dotted line), allcomputed with <strong>the</strong> FE-BE approach, and transfer function <strong>of</strong> <strong>the</strong> structure withSSI computed with <strong>the</strong> proposed <strong><strong>for</strong>mulas</strong> (dotted line). The figures correspondsto a structure resting on a 10 × 10 end-bearing <strong>pile</strong> group.Acceleration [m/s 2 ]Acceleration [m/s 2 ]30−30 4 8 12 16Time [s]30−30 4 8 12 16Time [s]Spectral acceleration [m/s 2 ]10864200 4 8 12 16Frequncy [Hz]Figure 10. Ground acceleration (left) and 5%-damped response spectra (right)recorded in Aegion (Greece) in 1995 (top and solid line), and in Friuli (Italy) in1976 (bottom and dashed line).5.2. Case 2In this example, we compare our simplified <strong>for</strong>mulations 10 <strong>for</strong> <strong>the</strong> impedance <strong>of</strong> <strong>the</strong> endbearing<strong>pile</strong> group introduced in Sec. 3.2 with <strong>the</strong> simplified <strong><strong>for</strong>mulas</strong> proposed in [34, 21]. Weuse an equivalent with <strong>of</strong> R f = 8 m, because <strong>the</strong> <strong><strong>for</strong>mulas</strong> in Table IV are proposed <strong>for</strong> circularor square foundation. Fig. 13 shows this comparison, along with <strong>the</strong> value <strong>of</strong> <strong>the</strong> BE solution.Our <strong><strong>for</strong>mulas</strong> seems to behave at least as well as <strong>the</strong> previously available one. Remember thatits range <strong>of</strong> application, in in particular in terms <strong>of</strong> <strong>the</strong> numbers <strong>of</strong> <strong>pile</strong>s is much larger.Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
SIMPLE FORMULAS FOR THE DYNAMIC STIFFNESS OF PILE GROUPS 17108Spectral acceleration [m/s 2 ]7.552.5Spectral acceleration [m/s 2 ]64200 2 4 6 8 10Frequency [Hz]00 2 4 6 8 10Frequency [Hz]Figure 11. Comparison <strong>of</strong> <strong>the</strong> acceleration response spectra at <strong>the</strong> top <strong>of</strong> <strong>the</strong>building <strong>for</strong> <strong>the</strong> Friuli earthquake (left) and <strong>the</strong> Aegion earthquake (right), using<strong>the</strong> complete FE-BE model (solid line) and <strong>the</strong> simple <strong>for</strong>mulation (dashed line).The figures corresponds to a structure resting on a 10×10 end-bearing <strong>pile</strong> group.0.120.03Relative displacement [m]0.080.040−0.04−0.08Relative displacement [m]0.020.010−0.01−0.02−0.120 6 12 18 24 30Time [s]−0.030 6 12 18 24 30Time [s]Figure 12. Comparison <strong>of</strong> <strong>the</strong> relative displacements between <strong>the</strong> top and <strong>the</strong> base<strong>of</strong> <strong>the</strong> building <strong>for</strong> <strong>the</strong> Friuli earthquake (left) and <strong>the</strong> Aegion earthquake (right),using <strong>the</strong> complete FE-BE model (solid line) and <strong>the</strong> simple <strong>for</strong>mulation (dashedline). The figures corresponds to a structure resting on a 10 × 10 end-bearing <strong>pile</strong>group.6. CONCLUSION<strong>Simple</strong> <strong>for</strong>mulations have been derived <strong>for</strong> <strong>the</strong> <strong>dynamic</strong> <strong>stiffness</strong> matrices <strong>of</strong> <strong>pile</strong> groupfoundations subjected to horizontal and rocking <strong>dynamic</strong> loads. These <strong>for</strong>mulations were foundusing a large database <strong>of</strong> impedance matrices computed using a FE-BE model. They can bereadily employed <strong>for</strong> design <strong>of</strong> large foundations on <strong>pile</strong>s and are shown to yield very accuratevalues <strong>of</strong> <strong>the</strong> estimated quantities <strong>of</strong> interest <strong>for</strong> building design. The <strong>for</strong>mulations have beenderived both <strong>for</strong> floating <strong>pile</strong> <strong>groups</strong> on homogeneous half-space and end-bearing <strong>pile</strong> <strong>groups</strong>in a homogenous stratum. They can be used <strong>for</strong> large <strong>pile</strong> <strong>groups</strong> (n ≥ 50), as well as <strong>for</strong> alarge range <strong>of</strong> mechanical and geometrical parameters <strong>of</strong> <strong>the</strong> soil and <strong>the</strong> <strong>pile</strong>s. They providea first step towards code provisions specifically focused on <strong>pile</strong> footings.Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
18 R. TAHERZADEH, D. CLOUTEAU & R. COTTEREAU22Dynamic <strong>stiffness</strong> [N/m]3 x 109 Frequency [Hz]10Damping [N/m]2.5 x 109 Frequency [Hz]1.51−10.5−20 1 2 3 4 5 600 1 2 3 4 5 6Figure 13. Real (left) and imaginary (right) parts <strong>of</strong> <strong>the</strong> horizontal impedancematrix, <strong>for</strong> a 36 end-bearing <strong>pile</strong> group computed using <strong>the</strong> simplified<strong><strong>for</strong>mulas</strong> (10) (solid line) and BE solution (dashed line) and simplified analyticalsolution <strong>of</strong> [8] (dotted line).REFERENCES1. Wolf JP. Foundation analysis using simple physical model. Prentice-Hall: Englewood Cliffs, New-Jersey,1994.2. Kaynia AM. Dynamic <strong>stiffness</strong> and seismic response <strong>of</strong> <strong>pile</strong> <strong>groups</strong>. Research Report R82-03,Masshachussetts Institute <strong>of</strong> Technology, ; 1982.3. Waas G, Hartmann HG. Seismic analysis <strong>of</strong> <strong>pile</strong> foundations including soil-<strong>pile</strong>-soil interaction. Proceedings<strong>of</strong> <strong>the</strong> 8th World Conference on Earthquake Engineering San Francisco, July 1984; 555–62.4. Takemiya H. Ring-Pile analysis <strong>for</strong> a grouped <strong>pile</strong> foundation subjected to base motion. StructuralEngineering/Earthquake Engineering 1986; 3(1):195–202.5. Ohira A, Tazoh T, Dewa T, Shimizu K, Shimada M. Observation <strong>of</strong> earthquake response behaviours <strong>of</strong>foundations <strong>pile</strong>s <strong>for</strong> road bridge. Proceedings <strong>of</strong> <strong>the</strong> 8th World Conference on Earthquake Engineering,San Francisco, July 1984; 3:577–584.6. Konagai K, Ahsan R, Maruyama D. <strong>Simple</strong> expression <strong>of</strong> <strong>the</strong> <strong>dynamic</strong> <strong>stiffness</strong> <strong>of</strong> grouped <strong>pile</strong>s in swaymotion. Journal <strong>of</strong> Earthquake Engineering 2000; 4(3):355–376.7. Crouse CB, Cheang L. Dynamic testing and analysis <strong>of</strong> <strong>pile</strong> group foundation. In Dynamic Response <strong>of</strong>Pile Foundations-Experiment, Analysis and Observation, Nogami T. (ed). ASCE: New-York, 1987; 79–98.8. Mylonakis G, Nikolaou A, Gazetas G. Soil-Pile-Bridge seismic interaction: kinematic and inertial effects.Part I: s<strong>of</strong>t soil. Earthquake Engineering and Structural Dynamics 1997; 26(3):337–359.9. Hutchinson TC, Chai YH, Boulanger RW, Idriss IM. Inelastic seismic response <strong>of</strong> extended <strong>pile</strong>-shaftsupportedbridge structures. Earthquake Spectra 2004; 20(4):1057–1080.10. Wolf JP. Soil-structure-interaction analysis in time domain. Prentice-Hall: Englewood Cliffs, New-Jersey,1988.11. Levine MB, Scott RF. Dynamic response verification <strong>of</strong> simplified bridge-foundation model. Journal <strong>of</strong>Geotechnical Engineering 1989; 115(2):1246–1260.12. Spyrakos CC. Assessment <strong>of</strong> SSI on <strong>the</strong> longitudinal seismic response <strong>of</strong> short span bridges. EngineeringStructures 1990; 12(1):60–66.13. Harada T, Yamashita N, Sakanashi K. Theoretical study on fundamental period and damping ratio <strong>of</strong> bridgepier-foubdation system. In Proceedings <strong>of</strong> <strong>the</strong> Japan Society <strong>of</strong> Civil Engineers 1994; 489(1-27):227–234.14. Chaudhary MS, Parakash S. Dynamic soil structure interaction <strong>for</strong> bridge abutment on <strong>pile</strong>s. InGeotechnical Special Publication 1998;2:1247–1258.15. Spyrakos CC, Loannidis G. Seismic behavior <strong>of</strong> a post-tensioned integral bridge including soil-structureinteraction (SSI). Soil Dynamics and Earthquake Engineering 2003; 23(1):53–63.16. Tongaonkar NP, Jangid RS. Seismic response <strong>of</strong> isolated bridges with soil-structure interaction. SoilDynamics and Earthquake Engineering 2003; 23(4):287–302.17. Dobry R, Gazetas G. <strong>Simple</strong> method <strong>for</strong> <strong>dynamic</strong> <strong>stiffness</strong> and damping <strong>of</strong> floating <strong>pile</strong> <strong>groups</strong>. Geotechnique1988; 38(4):557–574.Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
SIMPLE FORMULAS FOR THE DYNAMIC STIFFNESS OF PILE GROUPS 1918. Gazetas G., Dobry R. Horizontal response <strong>of</strong> <strong>pile</strong>s in layered soil. Journal <strong>of</strong> Geotechnical Engineering1984; 110(1):20–34.19. Gazetas G, Makris N. Dynamic <strong>pile</strong>-soil-<strong>pile</strong> interaction. Part I: Analysis <strong>of</strong> axial vibration. EarthquakeEngineering and Structural Dynamics 1991; 20(2):115–132.20. Makris N, Gazetas G. Dynamic <strong>pile</strong>-soil-<strong>pile</strong> interaction. Part II: Lateral and seismic response. EarthquakeEngineering and Structural Dynamics 1992; 21(2):145–162.21. Mylonakis G, Gazetas G. Lateral vibration and internal <strong>for</strong>ces <strong>of</strong> grouped <strong>pile</strong>s in layered soil. Journal <strong>of</strong>Geotechnical and Geoenvironmental Engineering, ASCE 1999; 125(1):16–25.22. Pender M. Aseismic <strong>pile</strong> foundation design analysis. Bulletin <strong>of</strong> <strong>the</strong> New Zealand National Society <strong>for</strong>Earthquake Engineering 1993; 26(1):49-160.23. ATC: Tentative provisions <strong>for</strong> <strong>the</strong> development <strong>of</strong> seismic regulations <strong>of</strong> buildings: cooperative ef<strong>for</strong>t with<strong>the</strong> design pr<strong>of</strong>ession, building code interests and <strong>the</strong> research community. Washington, D.C, 1978.24. NEHRP, National Earthquake Hazards blackuction Program. Recommended provisions <strong>for</strong> <strong>the</strong> development<strong>of</strong> seismic regulations <strong>for</strong> new buildings. Washington, D.C, 1997.25. Randolph M. Response <strong>of</strong> Flexible Piles to Lateral Loading. In Geotechnique, 1981; 31(2):247–259.26. Poulos HG, Hull TS. The role <strong>of</strong> analytical geomechanics in foundation engineering. In FoundationEngineering: Current Principles and Practices, ASCE, 1989; 2:1578–1606.27. Gazetas G. Analysis <strong>of</strong> machine foundation vibrations: state <strong>of</strong> <strong>the</strong> art. In Soil Dynamics and EarthquakeEngineering, 1983; 2(1):2–42.28. Wolf JP. Consistent lumped-parameter models <strong>for</strong> unbounded soil: physical representation. In EarthquakeEngineering and Structural Dynamics, 1991; 20(1):12–32.29. Sieffert JG, Cevaer F. Handbook <strong>of</strong> impedance functions. Surface foundations. Ouest Éditions, 1992.30. Clouteau D, Taherzadeh R. Soil, <strong>pile</strong> group and building interactions under seismic loading. In Proceedings<strong>of</strong> <strong>the</strong> First European Conference on Earthquake Engineering and Seismology, Geneva, Switzerland,September 2006, in CDROM.31. Taherzadeh R, Clouteau D, Cottereau R. Identification <strong>of</strong> <strong>the</strong> essential parameters <strong>for</strong> <strong>the</strong> lateral impedance<strong>of</strong> large <strong>pile</strong> <strong>groups</strong>. In Proceedings <strong>of</strong> <strong>the</strong> 4th International Conference on Geotechnical EarthquakeEngineering, Thessaloniki, Greece, June 2007, in CDROM.32. Craig RJ, Bampton M. Coupling <strong>of</strong> substructures <strong>for</strong> <strong>dynamic</strong> analyses. AIAA Journal 1968; 6(7):1313–1319.33. Miura K, Kaynia AM, Masuda K, Kitamura E, Seto Y. Dynamic behaviour <strong>of</strong> <strong>pile</strong> foundations inhomogeneous and non-homogeneous media. Earthquake Engineering and Structural Dynamics 1994;23(2):183–192.34. Gazetas G, Hess P, Zinn R, Mylonakis G, Nikolaou A. Seismic response <strong>of</strong> a large <strong>pile</strong> group. In Proceedings<strong>of</strong> <strong>the</strong> 11th European Conference on Earthquake Engineering, Paris, September 1998, in CDROM.35. Cottereau R, Clouteau D, Soize C. Construction <strong>of</strong> a probabilistic model <strong>for</strong> impedance matrices. ComputerMethods in Applied Mechanics and Engineering 2007; 196(17-20):2252–2268.36. Cottereau R, Clouteau D, Soize C. Probabilistic impedance <strong>of</strong> foundation: Impact <strong>of</strong> <strong>the</strong> seismic design onuncertain soils. Earthquake Engineering and Structural Dynamics 2008; 37(6):899-918.37. Gazetas G. Seismic response <strong>of</strong> end-bearing single <strong>pile</strong>s. International Journal <strong>of</strong> Soil Dynamics andEarthquake Engineering 1984; 3(2):82–93.38. Poulos HG. Behavior <strong>of</strong> laterally loaded <strong>pile</strong>s: PART II - group <strong>pile</strong>s. Journal <strong>of</strong> <strong>the</strong> Soil Mechanics andFoundations Division, ASCE, 1971;733–751.39. Cottereau R. Probabilistic models <strong>of</strong> impedance matrices. PhD <strong>the</strong>sis, Ecole Centrale Paris, France,(http://tel.archives-ouvertes.fr/tel-00132950/en/); 2006.40. Nikolaou S., Mylonakis G., Gazetas G. and Tazoh T. Kinematic <strong>pile</strong> bending during earthquake: analysisand field measurement. Geotechnique 2001; 51(5):425-440.APPENDIX AIn this appendix, <strong>the</strong> practical methodology <strong>for</strong> <strong>the</strong> construction <strong>of</strong> <strong>the</strong> reduced matrixS(a 0 ) = K − a 2 0M + ia 0 C is introduced. Three main steps are identified:¢The impedance <strong>of</strong> <strong>the</strong> FE-BE model is computed. More specially a set <strong>of</strong> values {Z(a 0l )}<strong>of</strong> <strong>the</strong> impedance matrix at a finite number <strong>of</strong> frequencies (a 0l ) 1≤l≤L is computed.¢The set <strong>of</strong> values {Z(a 0l )} is interpolated to yield a matrix-valued rational function in<strong>the</strong> <strong>for</strong>m a 0 → N(a 0 )/q(a 0 ), which approximates <strong>the</strong> behavior <strong>of</strong> <strong>the</strong> impedance matrix{Z(a 0 )} <strong>of</strong> <strong>the</strong> model. The function a 0 → N(a 0 ) is a matrix-valued polynomial in (ia 0 ),Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
20 R. TAHERZADEH, D. CLOUTEAU & R. COTTEREAUand <strong>the</strong> function a 0 → q(a 0 ) is a scalar polynomial in (ia 0 ). Many methods can be usedto achieve that goal.¢The identification <strong>of</strong> <strong>the</strong> matrices K, C and M from <strong>the</strong> polynomials a 0 → N(a 0 ) anda 0 → q(a 0 ) is <strong>the</strong>n per<strong>for</strong>med. This step does not involve any approximation and isfur<strong>the</strong>r detailed in [39].APPENDIX BIn this appendix, we present an extended version <strong>of</strong> <strong>the</strong> <strong><strong>for</strong>mulas</strong> presented in Tables IIand IV, in a non-normalized <strong>for</strong>m. For <strong>the</strong> case <strong>of</strong> <strong>the</strong> floating <strong>pile</strong> <strong>groups</strong> on homogeneoushalf-space <strong>the</strong> coefficients appearing in <strong>the</strong> non-normalized version <strong>of</strong> Eq. (8) are⎧ (⎪⎨k h = 6.8G s RL0) 0.3√ f sc h = 5 GsR2 f L 0V s s(17)⎪⎩ (m h = 0.4ρ s Rf3 L0) 1.6sand( ) 0.6 ⎧⎪ ⎨ k r = 8G s Rf3 lp (L0) 0.4R f s√ (c⎪ r = 5 GsR3 fV Rf⎩slL0) 0.2(18)p sm r = 0.7G s Rf 4l ( L0) 0.4p sThe range <strong>of</strong> parameters is 250 ≤ E p /E s ≤ 1500, 2 ≤ s/d p ≤ 3.6 and 0.55 ≤ R f /l p ≤ 2 andconstant hysteric damping β s = 0.05.For <strong>the</strong> case <strong>of</strong> end-bearing <strong>pile</strong> <strong>groups</strong>, <strong>the</strong> coefficients appearing in <strong>the</strong> non-normalizedversion <strong>of</strong> Eq. (10) are⎧( ) 0.5 Rf (⎪⎨k 0 = k h − k 2 − k 3 = 10G s RL0) 0.35f H√ sc h = c 0 + c 1 = GsR2 f H L 0 ρ b V bV s R f s ρ sV s⎪⎩m h = 1 2 ρ sHRf2 , (19)⎧R⎪⎨ k 1 = 2.6G 2 f s Hc 1 = 1.9 GsR2 fV s(HR f ) 1.5 , (20)⎪⎩m 1 = 1.4ρ s Rf 2H ⎧) 0.35 s)sL 0( √Rfk ⎪⎨ 2 = 1.25G s Rρb V bf H(L 0 ρ sV sc 2 = 0.04G s Rf2 ρ b V b, (21)ρ sVs2√⎪⎩ m 2 = 0.08ρ s Rf 2H s ρb V bL 0 ρ sV s⎧Rk 3 = 16.1G 4 (f L0) 3√ρsV ss H ⎪⎨3 s ρ b V bc 3 = 3 Gs R 4 (f L0) 3( ) 1.5 ρsV sV s H 2 s, (22)ρ b V b⎪⎩Rm 3 = 0.6ρ 4 (f L0) 3√ρsV ss H 1 s ρ b V bCopyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls
SIMPLE FORMULAS FOR THE DYNAMIC STIFFNESS OF PILE GROUPS 21and ⎧⎪ ⎨⎪ ⎩k r = 15G s R 3 f√R fc r = 17 GsR4 fV sH(L0m r = 1.6ρ s R 4 f L0sL 0sρ b V bρ sV sρ sV sρ b V b) √ 2 Rfs√H. (23)R f H ρsVsρ b V bThe range <strong>of</strong> parameters is 125 ≤ E p /E s ≤ 750, 2.8 ≤ s/d p ≤ 4.4, 1 ≤ R f /H ≤ 2.1 and3 ≤ V b /V s ≤ 8 and constant hysteretic damping β s = 0.05.Copyright¡2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 00:1–6Prepared using eqeauth.cls