Ground Response Analysis 7.1 INTRODUCTION 254 ‘One ofthe mostimportant and most commonly encountered problems in geotechnical earth- ‘quake enginecring is the evaluation of ground response. Ground response analyses ae used to predict ground surface motions for development of design response spectra, to evaluate dynamic stresses and strains for evaluation of liquefaction hazards, and to determine the earthqualke-indvced forces that can lead to instability of earth and earth-retaining structures. Under iden conditions, a complete ground response analysis would model the ruptare ‘mechanism at the source ofan earthquake, the propagation of stress waves through the earth to the top of bedrock beneath a particular site, and would then determine how the ground surface motion is influenced by the sols that lie above the bedrock. Tn reality, the mech nism of fault rupture isso complicated and the nature of energy transmission between the source and the ste so uncertain that this approach isnot practical for common engineering applications. In practice, empirical methods based on the characteristics of recorded earth- {guakes are used to develop predictive relationships of the types discussed in Chapter 3, ‘These predictive relationships are often used in conjunction with aseismic hazard analysis to predict bedrock motion characteristics at the site. The problem of ground response anal- Ysis then becomes one of determining the response of the soil deposit to the motion ofthe beadrock immediately beneath it, Despite the fact that seismic waves may travel through tensSec.7.2 OneDimensional Ground Response Analysis 255, of kilometers of rock and often less than 100 m of sil, the sol plays a very important role in determining the characteristics of the ground surface motion. ‘The influence of local soil conditions on the nature of earthquake damage has been recognized for many years, Since the 1920s, seismologists and, more recently, geotechnical earthquake engineers have worked toward the development of quantitative methods for pre= dicting the influence of local soil conditions on strong ground motion. Over the years, ‘number of techwiques have been developed for ground response analysis. The techniques are often grouped according to the dimensionality of the problems they can address, although many ofthe two- and three-imensional techniques are relatively straightforward ‘extensions of corresponding one-dimensional techniques. This chapter deseribes the most ‘commonly used methods for one-,tWo-, and three-dimensional ground response, and intro- ‘duces the problem of soil-structut interaction 7.2 ONE-DIMENSIONAL GROUND RESPONSE ANALYSIS ‘When a fault ruptures below the earth's surface, body waves travel away from the source in all directions. As they reach boundaries between different geologic materials, they are reflected and refracted. Since the wave propagation velocities of shallower materials are generally lower than the materials beneath them, inclined rays that stike horizontal layer boundaries are tsually reflected to a more vertical direction. By the time the rays reach the aground surface, multipfe refractions have often bent them to a nearly vertical direction (Fig- ire 7.1). One-dimensional ground response analyses are based on the assumption that all boundaries are horizontal and that the response ofa soil deposit is predominantly caused by SH-waves propagating vertically from the underlying bedrock. For one-dimensional _ground response analysis, the soil and hedrock surface are assumed to extend infinitely in the horizontal d rection, Procedures based on this assumption have een shown to predict ground response that isin reasonable agreement with measured response in many cases. eS aut ‘Sicil ayore Figure74_ Refraction procs that rodees nearly veal wave propagation nea the round suze [Before describing any ofthe ground response models it is necessary to define several terms that are commonly used to describe ground motions, With reference to Figure 7.23, the motion a the surface of a soil deposit isthe free surface motion. The motion atthe base ofthe soil depos also the top of bedrock) is called a bedrock motion. The motion ataloca- tion where bedrock is exposed at the ground surface i called a rock outcropping motion. If the soil deposit was not present (Figure 7.2b), the motion at the top of bedrock would be the Ihedrock outeropping motion.256 Ground Response Analysis Chap. 7 Fre sutace motion x Fock ‘roto? = Bodo Bost rroten utering rotor ° » Figure 7.2 Ground response nomenclature) ol overying beac: (m0 oi oveting bedi. Ver sce xagperato, 7.2.1 Linear Approach The mannerin which transfer functions can be used to compute the response of single- sdepree-of-freedom systems is ilustrated in Appendix B (Section B.5.4.2). An important lass of techniques for ground response analysis is also based on the use of transfer func- lions. For the ground response problem, transfer functions can be used to express Various response parameters, such as displacement, velocity, acceleration, shear stress, and shear strain, to an input motion parameter such as bedrock acceleration. Because it relies on the Principle of superposition, this approach is limited to the analysis of linear systems. Non- linear behavior can be approximated, however, using an iterative procedure with equivalent linear soil properties, The mathematical aspects ofthe transfer function approach were described in Section B.S.4.2 of Appendix B. Although the calculation involve manipulation of complex num- bers, the approach itself is quite simple. A known time history of bedrock (input) motion is represented aa Fourier series, usually using the FFT (Section A.3.4) Each term in the Fou- rier series ofthe bedrock (input) motion is then multiplied by the transfer function to pro- ‘duce the Fourier series ofthe ground surface (output) motion. The ground surface (output) ‘motion can then be expressed in the time domain using the inverse FFT. ‘Thus the tansfer function determines how each frequency in the bedrock (input) motion is amplified, or ‘deamplified, by the soil deposit 7.2.1.1 Evaluation of Transfer Functions ‘The key to the linear approach isthe evaluation of transfer functions. Inthe following sections transfer functions are derived for a series of successively more complicated geo- technical confitions. Although the simplest ofthese may only rarely be applicable to actual, problems, they Mustrate some of the important effects of soil deposits on ground motion ‘characteristics without undue mathematical complexity. The more complex are capable of| describing the most important aspects of ground response and are very commonly used in ‘geotechnical earthquake engineering practice.Sec.7.2 Che Dimensional Ground Response Analysis 257 Uniform Undamped Soil on Rigid Rock. First, consider a uniform layer of isotropic, linear clastic soil overlying rigid bedrock as shown in Figure 7.3. Harmonie horizontal mation of the bedrock will produce vertically propagating shear waves in the ‘overlying soil. The resulting horizontal displacement can be expressed, using the results of lz, 1) = Act! Bele) ay ‘where «is the circular frequency of ground shaking, k the wave number (= @Wy,) and A and B the amplitudes of waves traveling in the ~z (upward) and +2 (downward) directions, respectively. At the free surface (2 =0), the shear stress, and consequently the shear strain, ‘must vanish; that is, 10,0) = 6y(0,1) = 6 au(0,1) sae a 2) Substituting (7.1) into (7.2) and differentiating yields Gik(ACKO— Bee" = Gik(A Bye = 0 3) ‘which is satisfied (nontrivially) when A = B, The displacement can then be expressed as 2 Which describes a standing wave of amplitude 2A cos kz. The standing wave is produced by the constructive interference of the upward and downward traveling waves and has a fixed shape with respect to depth. Equation (7.4) can be used to define a transfer function that describes the ratio of displacement amplitudes at any two points inthe soil layer. Choosing these two poinss tobe the top and bottom ofthe soil layer gives the transfer function (a) = 2A lat = 2A cos ke elt aa) fo = te.) 2c 1 1 o FO) (imED ” Theos kHe™ ~ Soak” cos taHy =) ‘The modblus ofthe transfer funtion is the amplification function [rol = (REF OIF+ Tima? = __ gw fos AT] ‘which indicates that the surface displacement is always atleast a large as the bedrock dis- placement (since the denominator can never be greater than 1) and, at certain frequencies, is ' j atest) Figure7.3. Lina lati sil dopos of thickness under by ig bedrock258 Ground Response Analysis Chap. 7 ‘much larger. Thus IF (ais the ratio of the free surface motion amplitude to the bedrock ‘motion amplitade (or, since the bedrock is rigid inthis case, the bedrock outeropping motion). As oly, approaches 1/2 + nr, the denominator of equation (7.6) approaches zero, which implies tha ininite amplification, or resonance, will occur (Figure 7.4. Even this very sim- ple mode! ilustrates thatthe response ofa soil deposit is highly dependent upon the frequency ‘of the base metion, and thatthe frequencies at which strong amplification occurs depend on ‘the geometry (hickness) and material properties (s-wave velocity) of the soil layer. ca % z 2 ™ te Se Tee meg °F 2H cd mae Fore? lence of rer on say sate respon of unaped earelaie ter Fsample 7 ‘Compute he time history of acceleration tthe surface ofthe linear elastic sil deposit shown in Figure tain response to the F-W component of the Gilroy No.1 (ock) motion (Figure 3.1. Figure E719 Solution Computation ofthe ground surface motion from the bedrock motion canbe accor: lished inthe allowing eres of ep, 1. Obain the time history ofaceleration ofthe input motion. In this case the inp motion isthe E-W componentof the Giloy No. | (rock) motion shown in Figure E7.1b. The Gil toy No I record consists of 2000 acceleration values at 0.02-se ineval, 2% Compute the Fourier series ofthe bedrock (input) motion, The Fourier sisi complex "alae its one-sided Fourier amplitude spectrum (Section A.3 of Appendix A) is shown in Figure E7. 1c. The Fourier amplitude spectrum is defined for Frequencies upto 2A 25 Hz, but most ofthe energy inthe bedrock motion iat fequenties les than Sto 10 Ha 3. Compute the transfer function that relates the ground surface (outpt) motion o the bed *oek (input) motion. From equation (7-5), the transfer function (Figure ET.) forthe case oF endamped sil is real valued. The wansfer function has values oft belo fequenciesSec.7.2 One Dimensional Ground Response Analysis 258 ‘of about 10 Hz. However, at fequencies tha approach the fundamental frequency ofthe sil deposit f= v,/4H= 2625 Hr, the transfer function begin totake on large values 4. Compe the Fourier series ofthe ground surface (output) motion a the product ofthe transfer function and the Fourier series ofthe bedrock (inp) motion. At frequencies less ‘than So 10 Hz, the Fourier spectrum ofthe ground surface motion i virally the same ‘as thtof the bedrock motion. Aldnough the transfer Function indicates that frequencies shove 20 Hz or so will be amplified strongly, the input motion is weak in tha frequency ange. The one-sided amplitde spectrum is shown in Figure B7 Le. Examination of this Fourie“amplitue spectrum indicates thatthe ground surface motion has somewhat more high equeney motion, bis generally similar to that ofthe bedrock mation. ‘5. Obianthe time history of the ground surface motion by inverting is Fourier series. As illustrated in Figure E71f, the time history of ground surface motion has a somesvhat _greater content of high-frequency components, but i generally’ similar to the time his {ory ofbedrock motion ° 5 10 15 20 2 Frequency (H) @ zw 4 ai + are $$ —_________ens o HE iB® rr a a a rea s eceleraton , Bok o 5 0 1 2 2 9% a 4 Tie (sec) Figure brab-t260 Ground Response Analysis Chap. 7 Because the soil layer had a high natural frequency (intentionally selected to avoid the ‘extemely lage transfer function valves that exst near the natural frequency of an undamped Jaye) its effect on th ground surface mosion was relatively small Uniform, Damped Soll on Rigid Rock. Obviously, the type of unbounded amplification predicted bythe previous analysis cannot physically occu. The previous anl- 4s asued no isipation of energy, or damping, in te soi. Since damping i resent in all ‘materials, more realistic results canbe obtained by repeating the analysis with damping Assuming the sil to ave the shearing characteristics ofa Kelvin-Voigt sli, the wave equation can be writen [equation (5.94) a, Pu 9 Ba Pu 2 GE yy Pu 0. Par * Cet Mactan a As shown in equation (5.94), the solution to this wave equation is ofthe form (a0) = Ael@rr#9 4 Bellon ‘where k* isa cemplex wave number with real part ky and imaginary part . Repeating the previous algebraic manipulations withthe complex wave number, the transFet function for the ease of damped soil over rigid rock can be expressed as L 1 PO) = SEUFH ~ caste) ae Since the frequeney-independent complex shear modulus (Section 5.51) is given by G* = (1+ 128), the complex shear wave velocity can be expressed as B . (OBD. Bers = 90 +18) 09) for small & Thea the complex wave number can be writen, again for small , as be So og RO =) (7.10) and finally, the ransfer function, as 1 1 ‘osk(1 16) ~ cos ToH{1 +7] Using the identity eos(x-+ i9] = Joos" snh¥y, the amplification function can be oxpreeed as Fyo) ay {Fx(@)| = ——_____ aay [eos *kH + sinh? Ek ‘Since sinh*y ~ y* for small y, the amplification function can be simplified to [F(@) ~ 1 ___ - —__1___ (7.13) eos H+ GEM) feos? (OHTo,) + ECORI DT For smal daing ratios, equation (7.13) indicate that amplification by a damped soi layer also varies with frequency. The amplification will each a local maximum whenever kil = 2+ macbat will never reach a value of infinity since (for & > 0) the denominator willSec.7.2. OneDimensional Ground Response Analysis 261 always be greater than zero. The frequencies that correspond to the local maxima are the ‘natural frequencies of the soil deposi. The variation of amplification factor with frequency is shown for different levels of damping in Figure 7.5. This amplification factor is also equal to the ratio ofthe free surface motion amplitude to the bedrock (or bedrock outcropping) ‘motion amplitude. Comparing Figures 7.4 and 7.5 shows that damping affects the response at high frequencizs more than at lower frequencies. 2 a #6 ge ga : Zo ° ° 3 10 15 2 a Pewe7s atti tnt di Te than equal hel ep cn aY(Sem) ne hue ony Since the peak amplification factor decreases with increasing natural frequency, the greatest amplification factor will occur approximately a the lowest natural frequency, also known athe funddamentel frequency ‘The period of vibration corresponding o the fundamental frequency is called the charac- teristic site period, ns) 1, = 28 MH 16 ‘The characterise pid, hich deen only onthe hikes ander Wav veaity ofthe soi provi ery inden th pero of aon attic not, Pca amplicon canbe expend At each natal frequency. tnding wave develops ia the si: Normalized Aefomed stapes r mteshapes refs te al equa shown Pgs 76.Note that he i esplacenens raps al eps ie unanetl mode bt notin he higher nodes. At eguence above te fundamental feteny at fhe ll ‘cpt my bo moving none coon whi anater pais moving epee dce ton This phenomenon mt be cms in thevalueon etl fee al ees reid forsee sab sales Che262 Ground Response Analysis Chap, 7 ae Normalzadeiplacomont Siace 1905 0s : V | < VEEL: ° rock igure Displacement pes fr sanding waves at undamenl (r= 0), second =| ‘anid ¢r=2) natal agence fora sol ayer wih § = 5%, Dipset ae mare by maxim dseaceren a tefnlament eer. Example 7.2 The sitext which the Gilroy No.2 (oil) eanhquake motion shown in Figure 3.1 was recorded isundercin hy some 40 of soil underlain by shale and secpentnite bedrock. Te shear wave ‘eocity ofthe sil varies from about 1000 fuses at desis ess than sbout 13040 about 2000 Fuse at seater depths, Assuming an average shear wave velocity of 1500 fee, an average unit weigh of 125 Iva damping aio of 5%, ad rgid bedrock (Figure B7.28), compute the ‘round surface motion that would occur te bedrock was subjected tothe E-W component of the Gilroy No.1 (oek) motion, igi Beco Figure E720 Solution Computation ofthe ground surface maton from the bedrock motion canbe accomn- Dlised ir the same five steps as described in Example 71 1. Ottain the time history of acceleration of the ip motion, This stp i denied ta step 1 Example 7.1: the motion is shown in Figure E72. 2. Campa the Fourier series ofthe bedrock (input) motion. This step i als iden step 2 in Example 7.1; the result is shown in Figure E72c, ‘3. Compute the transfer fonction that relates the ground surface (output) motion tothe ed rock input) motion, From equation (7.11), the easter funetion forthe case of damped soils complex valued, The modulus ofthe transfer function is shown in Figure E7 2, ‘The shape of the transfer faction indicates that significant amplification will occur at several natural frequencies, and that higher frequencies (sreater than about 10H) will besuppressed. caltoSec. 7.2 One-Dimensional Ground Response Analysis 263 ‘4 Compute the Fourier series ofthe ground surface (outp) motion asthe product ofthe tearsfr function andthe Four series ofthe bedrock (input) motion. The Fourier spectrum of ground surface mesion (Figure E7.2e) shows amplification at the natural frequencies of be soil deposit and litle high frequency motion. ‘5. Obtin the time history of the ground surface mation by inverting the Fourier series. The time history of ground surface motion (Figure 7.21 indicates that peak acelerations at {he ground surface and bedrock levels are similar, but the frequency contents re differ ent. Because the ground surface motion is weighted toward lower frequencies, the peak ‘velocity and displacement athe pround surface ae likely tobe considerably greater han at badeook, ‘Therigid bedrock analysis predicts a peak ground surface accleation of 0.452 g, which isconsidembly sreate than the peak acceleration of 0.3229 actually recorded at the Gilroy No 2 (oil) tain, Uniform, Damped Soil on Elastic Reck. ‘The preceding two sections developed expressions for amplification factors for sols overlying rigid bedrock. If the bed ‘ock:srigid its motion willbe unaffected by motions in, or even the presence of, the overlying. 08) pe ee ° 5 ‘ 1s Fa 2 Frequency he g 0s © ee fies sh o 6 a8 aaa “me ace268 Ground Response Analysis Chap. 7 soil. tact asa fixed end (Section 5.4.1) boundary. Any downward-traveling waves nthe soil will be eompetely reflected back toward the ground surface by the rigid layer, thereby rap- ping al ofthe elastic wave energy within the soil layer. Ifthe nck is elastic, however, downward-traveling tress waves that reach the soil rock boundary will be reflected only partially; part of their energy will be transmitied through the boundary to continue traveling downward through the rock. Ifthe rock extends {o great depth (large enough that waves reflected from any deeper material boundaries do ‘not return tthe soil-rock boundary soon enough, or with sufficient amplitude, to influence the response of the soil deposit) the elastic energy of these waves will effectively be removed fromthe soil layer. Ths is form of radiation damping, and it causes the free sur- {ace motion amplitudes to be smaller than those for the case of rigid bedrock. ‘Consider the case of a soil layer overlying a halfspace of elastic rock (Figure 7.7). If the subscripts sand r refer to soil and rock, respectively the displacements due to vertically propagating s-waves in each material can be written as Ue t) = AsetOr Hed 4 Bel O-KS) (17) eyo) = Aeon tis) + Bef ert) 7) ‘The fe surfice effect, as before, requires that A, = B,, and compatibility of displacements and continuity of stresses atthe soil-rock boundary require that, =0) 7.18) 4G, =) = 446,=0) (19) Substituting equations (7.17) into equation (7.18) yields ue, =H) =u Aye eM) = A+B, (7.20) ‘From equaticn (7.19) and the definition of shear stress (t= G ud) AGG Khe gay Figue 7.7. Nomeeltre forthe ae oi ayer overSec.7.2. One Dimensional Ground Response Analysis 265 or (7.22) Mento Paver (7.23) Pr Where vj, and v5, are the complex shear wave velocities of the soil and rock, respectively, and a is the complex impedance ratio (see Section 5.4.1). Solving equations (7.20) and (7.22) simultaneously gives, LALA + oe 1 —atyeky (7.240) itty (7240) B, = LA ate 4 (1 +02) ‘Suppose hat a vertically propagating shear wave of amplitude, A, traveled upward Uhrough the rock. the soil was not present, the free surface effect atthe rock outerop would ‘produce a bedreck outcropping motion of amplitude 24, With the sol present, the free sur- face motion amalitude would be 4A Trae. (eet Defining the transfer function, F;, as the ratio ofthe sol surface amplitade to the rock out- crop amplitude, 2A, = 2 BO) = oa 725) ras aH hich, using Euler's law, can be rewriten as Fy) = a 1 026) cos GH +i of sin GH ~ cos(@Hivi,)+i af sin(@iilv) The modulus of F(@) cannot be expressed ina very compact form when soil damping exis To illustrate the important effet of bedrock elstcity, however, the amplification factor for undamped soil can be expressed as [Fst@.g = 0)| = 1 27 feostk,H + oF sin*k, "Note that resonance cannot oceur (the denominator is always greater than zero, even when the sol is undamped). The effect ofthe bedrock stiffness, as reflected by the impedance ratio, ‘on amplification behavior is illustrated in Figure 7.8. Not the similarity between the effects ‘of soil damping and bedrock elasticity by comparing the shapes of the amplification factor ‘curves in Figure 7.8 and those in Figure 7.5, The elasticity of the rock affects amplification266 Ground Response Analysis Chap. 7 Impedance rato = 0.0 Figure 74 Effect of mpolane rato oa amplifies ator fr ase of uname il similarly tothe damping ratio of the soil—both prevent the denominator from reaching zero. ‘Thisradiation damping effect has significant practical importance, particularly in the eastern United States, where bedrock is substantially harder than that typically found in the western states. The ster bedrock means that greater amplification may occur in the east and that design ctteria established on the basis of empirical evidence from western earthquakes may bbe somewhat unconservative inthe east. Example 7.3 ‘Repeat Example 7.2 assuming that the bedrock is wot rigid. Assume a shear wave velocity of, 5000 fe, a unit weight of 160 bt, nd 25 damping for bedrock atthe Gilroy No, 2 site (ewe £7.30, Joe ap = l. se 60 Hay Figere £7 8 Solution Computation ofthe ground surface motion from the bedrock motion canbe accom plished nthe same five steps desribed in Example 72. The only diference shat the ansfer funetior inthis example wil inclode the eects of bedrock compliance. 1. laain the ime history of acceleration ofthe input motion. This step is identical to step in Examples 7.1 and 7.2; the motion is shown in Figure E7 3b 2 Compute the Fourier series ofthe bedrock (input) motion, Again, this tp is identical vo ‘ep 2 in Examples 7.1 and 7.2: the results shown in Figure E7 3c 3. Compute the transfer funetion that relates the ground surface (output motion tothe bed- tock (input) moti, From Equation 7.25 the ransfer function for thegase of dampedSec. 7.2. One-Dimensional Ground Response Analysis 267 a I « | 8a asa a Tier) xz” onbe A oO ° 3 ‘o = Fy 3s eavncy © @ PFs ° 0 2 7 1 % 25 reqerey tt) 1° $28 20 lei Eo | ° 3 ry ‘5 20 2 requ ty 0; ——____ "ent © CO os a ee) Tres) Frewe #73 ‘oils complex valued. The modulus ofthe wansfr function, shown in Figure E734 indestes that ess amplification wil occur than in the ease of rigid bedrock. ‘4 Compute the Fourier series of the ground surface (output) motion asthe product ofthe transfer function and the Fourier series ofthe bedrock (input) motion, The Fourier series ofthe ground surface motion (Figure E73) shows less amplification than inthe case of ‘5. Obtain the time histor ofthe ground surface motion by inverting the Fourier series. The time history of ground surface motion Figure E7.30 indicates thatthe peak acceler. tions atthe ground surface i lower than the peak acceleration a the Bedrock level; the frequency contents are aso diferent ‘The compliant bedrock analysis predicts peak ground surface acceleration of 0.339%, ‘hich agrees well with the peak acceleration of 03224 recorded atthe Gilroy No.2 (sol) ta tion. The gocd agreement betwen peak accelerations, however, doesnot ean tha his simple analysis has peice al aspects ofthe Gilroy No.2 (sil) motion, Comparison ofthe Fourier “amplitude spectrum of the predicted motion (Figure E7.3e) with that ofthe recorded mation (Figure 3.136) shows significant differences in frequency content.268 Ground Response Analysis Chap. 7 Layered, Damped Soil on Elastic Rock. While the uniform elastic layer ‘models are Lseful for illustration of the influence of soil conditions on several ground ‘motion charceterstcs, they are seldom suitable for analysis of practical ground response problems. Real ground response problems usually involve soil deposits with layers of dif- ferent sifness and damping characteristics with boundaries at which elastic wave energy Will be reflected and/or transmitted. Such conditions require the development of transfer functions forlayered soil deposits. Consider a soil deposit consisting of N horizontal layers where the Nth layer is bed= rock (Figure 7.9). Assuming that each layer of soil behaves as a Kelvin—Voigt solid, the ‘wave equatin is ofthe form given in equation (5.94). The solution tothe wave equation can be expressed in the form wet) = AcK@rse'd 4 pears 28) Where and 3 represent the amplitudes of waves traveling inthe (upward) and +2(down- ward) ditections, respectively. The shear stress is then given by the product of the complex shear modulus, G*, and the shear strain, so ee 4 =e GH a(G iam) 5, a 28) (7.29) Introd.cing a local coordinate system, Z, for each layer, the displacement atthe top and bottom ef layer m will be ti(Zn = 941) = (Ay + Bye (7303) lig Zn. hes) = (Ame + By (7300) Displacements at layer boundaries must be compatible (i.e. the displacement athe top of 4 particular liyer must be equal to the displacement atthe bottom of the overlying layer. Applying the compatibility requirement to the boundary between layer m and layer m+ 1, that is, c actuals aan a oy Gu SP Figure 79 Nomenclature for yee sil depot on lnc bookSec.7.2. One-Dimensional Ground Response Analysis 269 yields Aust Baer = Agent Bye he asi) The shear stresses atthe top and bottom of layer m are Yn(Zn= 0.1) = HGA(Ay~ Bye! (7.324) Tal Zin = fgg) = TG (Agel — ByeHitey ele (7.32) Since stresses must be continuous a layer boundaries, Tn(Zn = hf) = Tesi Zot = O51) kG hap eoikshe Aas ~Baos = Grae (Agee Betty 33) Adding (7.31) and (7.33) and subtracting (7.33) from (7.31 gives the recursion formulas Amar = pAm(1 + Ones + 1B, (1 a5) tite (74a) Buys = Ama de™n + 18,1 + oye (7.340) ‘here ay isthe complex impedance ratio at the boundary between layers m and m+ | ay, = —*2Ga Palme (7.35) Kur1Gnet Pret mt At the ground surface, the shear stress must be equal to zero, which requires [from equation (7.32a)] that Ay = By. Ifthe recursion formulas of equation (7-38) are applied repeatedly for alllayers fom I tom, functions relating the amplitudes in layer m to thoxe in layer I canbe expressed by An = dy @)Ay 7.362) By, = byl @)By 7.360) ‘The transfer function relating the displacement amplitude at layer to that a layer jis given by co) = {t] _ alo) +b) ao) ~ [id - seen) 030 Because |il = lil = «|u| for harmonic motion, equation (7.37) also describes the amplification of accelerations and velocities from layer‘ o layer Equation (737) in cates that the mosion in any layer can be determined from the motion in any other layer. Hence ifthe motion at any one point inthe soil profile is known, the motion at any other Point can be contibuted. This result allows avery useful operation called deconvolution (Section 7.2.14) 0 be performed Example 74 ‘AS par of comprehensive investigation of round mation estimation ehniqes the Elec Power Reseach Istute perfoieda dui subsrace investigation the se the Gioy210 Ground Response Analysis Chap. 7 ‘No.2 oi) recording sation (EPRI, 1993) A rough approximation othe measured shear wave ‘velocity profile is listed below. Avenge Shear Depth Range ft) __ Wave Vela (ice) om ‘oo 20-45 0 470 1500 70-130 1000 10-510 >st0 ‘Assum ng, asin Examples 72 and 7.3, an average soil unit weight of 125 It and 5% soil damping, compote the expected pround surface response when the bedrock is subjected tthe Gilroy No.1 (rock) motion. Solution As in the previous examples of this chapter, this problem requires evaluation of he ‘transfer function that relates the ground surface motion othe bedrock motion, Because of ml: tiple reflections within the layered system, the transfer function [equation (7.36 frthisexan- ple is considerably more complicated than forthe single-layered cases of the previous examples, While the ransfer fusion can be evaluated by hand, it has also been coded in he computer program SHAKE (Schnabel eta, 1972). SHAKE was used, with constant sil tif ness and damping ratio, to obtain the transfer function shown in Figure E74. Asin the pev ‘ous examples, the Fourier series ofthe ground surface motion Figure E740) was computed a: the produto the transfer function andthe Fourier eres ofthe bedrock mation Inversion of his Foutier series produces the tim history of ground surface aeveleration (Figure E74). Examination of Figure E74e shows thatthe transfer function foe the layered system is indeed more complicate than the wansfer functions for the singl-layered cases of Examples 71 “72, and 7.3. Tho spikes inthe transfer funeton a frequencies of about 35 and 55 Hz help pro dae a pak aceleraton of 0.4994 thas considerably largerthan the peak acceleration of 3225 ‘that was actully recorded atthe Gilroy No. (sil) tation, Diferences in the frequency contents ‘ofthe redictod and recorded motions canbe seen by comparing Figures E7.4d and 3.138. 7.2.1.2 Equivalent Linear Approximation of Nonlinear Response Since he nonlinearity of soil behavior is well known, the linear approach must be modified to provide reasonable estimates of ground response for practical problems of, interest. As ciscussed in Chapter 6, the actual nonlinear hysteretic stress-strain behavior of cyclically oxded soils can be approximated by equivalent linear soil properties. The equiv alent linear shear modulus, G, is generally taken as a secant shear modulus and the equiv alent linear éamping rato, 6, as the damping ratio that produces the same energy loss in single eycle asthe actual hysteresis loop. The stran-dependent nature ofthese equivalent linear properties was deseribed in Section 6.4.2. Since the linear approach requires that G and & be constant for each soil layer, the problem becomes one of determining the values that are consistent with the level of strait induced in each layer. To address this problem, an objective definition of strain level is ‘needed. The laboratory tests from which modulus eduction and damping ratio curves (e.g. those shown in Figures 6.47 and 6,50) have been developed used simple harmonic loadingSec.7.2 OreDimensional Ground Response Analysis an Frequency Hz) 0 2% % 4 6 60 70 80 9 Time (ese) igure E74 ‘and characterized the strain level by the peak shear strain amplitude. The time history of shear strain for atypical earthquake motion, however, is highly regular witha peak ampli- tude that may only be approached by few spikes inthe record. Figure 7.10 shows both har. ‘monic (asin a typical laboratory test) and transient (asin atypical earthquake) shear strain time histories that have the same peak eyclic shear strain, Clearly, the harmonic record rep resents a more severe londing condition than the transient record, although their peak values are identical. Asa result, itis common to characterize the stain evel ofthe transient record interms of an effective shear strain which has been empirically found to vary between about [50 and 70% of te maximum shear strain. The computed response is not particularly sen- sitive to this percentage, however, and the effective shear strain is often taken as 65% ofthe peak strain, Since the computed strain level depends on the values ofthe equivalent linear prop- ctes, an iterative procedure is required to ensure that the properties used in the analysis are ‘compatible with the computed strain levels in all layers. Referring to Figure 7.11, the iter- alive procedure operates as follows:Shear strain 2m Ground Response Analysis Chap. 7 Fire 7.10 Two shear arin ime histories with etl pak she stan. Fr the tanseot motion ofa seta enthqake the effective shee Sain usally taken as ‘athe peak sean 2 te ” co a Sheers (og sae) er stain ons Fee 711 tertn toward ssincompe tr mln naming in equal! eral Using nal exis inde at es aml pred ameflectve shew stain. Bec hs rains pester thon oie onmepondng 1 and an tera ere Tee nex keaton espace {© nd hat ae eampatbwith 7 The equal incr rl epee ant de parmatry ccd sin omg vl of Gand re one. 4, Initial estimates of G and & are made for eacl layer. The initially estimated values usually correspond to the same strain level; the low-strain values are often used for ‘he inital estimate, 2. The esiimated G and & values are used to compute the ground response, including Lime histories of shear strain foreach layer. ‘3. The effective shear strain in each layer is determined from the maximum shear strain {inthe computed shear strain time history. For layer j Wy = Ray here the superscript refers to the iteration number and R; isthe ratio ofthe effective shear sirain to maximum shear strain. R, depends on earthquake magnitude (Idriss and Su, 1992) and can be estimated fromSec.7.2. One-Dimensional Ground Response Analysis 23 M-1 10 4. From thiseffective shear strain, new equivalent linear values, G*" and" are cho- sen for the next iteration, 5. Steps 2to4 are repeated until differences between the computed shear modulus and damping ratio values in two successive iterations fall below some predetermined value in all layers. Although convergence is not absolutely guaranteed, differences of less than 5 t0 10% are usually achieved in three to five iterations (Schnabel et al., 1972) Ry Even though the process of iteration toward strain-compatible soil properties allows nonlinear soil behavior to be approximated, it is important to remember that the complex ‘esponse method is stil linear method of analysis. The strain-compatible soil properties fare constant throughout the duration ofthe earthquake, regardless of whether the strains at ‘paricular time are small or large. The method is incapable of representing the changes in soil stiffness that actually occur during the earthquake, The equivalent linear approach to ‘one-dimensional ground response analysis of layered sites has been coded into a widely used computer frogram called SHAKE (Schnabel etal, 1972). Example 75 ‘An extensive laboratory testing rogram conducted by EPRI (1993) produced detailed infor- ‘mation onthe modulus reduction and damping characteristics ofthe soils beneath the Giltoy [No.2 (soi cording station. Although the soll conditions varied with depth, rough approx: mation tothe average modulus redvtion and damping characteristes is given below. Sain 6) ot 109 wo? 0 10? ow 1% a0! Gilg 100 1.00 100 099 090 O71 047 024 010 00s ae Sis) 307 300 300353 483 768 123 IS 24 220” 300 Repeat the alysis of Example 74 with the data listed above using the iterative equivalent in- at approach Solution As inthe case of Example 7.4, the transfer funtion Was evaluated using the come pte progran SHAKE (Seluabel etal, 1972). Inthisexamplete hist tration use the Sate stiffness and damping values used in Example 74, Subsequent iterations used stiffness and damping values that were consistent with the modulus reduction and damping behave listed above. After a total of eight iterations, the shear moduli and damping ratios had converged 40 ‘within 1% of stain-compatibie values. Because the strain-compaible shear modull were ‘smaller than the low-sran shear moduli on which the analysis of Example 7-4 was based (the iterations converged to strain a Which G/Gnax values were les than 1.0), the transfer fue: ion Figure 37.Sc) was shifted toward lower requencies. As inthe previous examples, the Foxe Fer series ofthe ground surface motion Figure E7.54) was computed asthe product of the transfer function and the Pouter eres ofthe bedrock motion. Inversion of this Fourier sevies ‘roduced th ime history of ground sutface aceleration shown in Figure E7 Se24 Ground Response Analysis Chap. 7 ¢ $ 0 6 30 5 Frequent 100) >> @ & x = aw ° 3 ‘0 1 20 25 Fraquery i 25 Frequoney (2) i ow 2 9 4 Ss? 60 70 6 Time (se) igure E75 The softer sol behavior indicated by the iterative equivalent linear analysis is clearly ‘reflected nthe ranfer funtion (Figure E7 5), tho Fourie serie of the ground surface motion (Figure E74), andthe time history of ground surface aceleration (Figure E7.3e) The pre dicted peak ground surface acceleration of 0.304 compares well with the peak acceleration of (0.322 resonded atthe Gilroy No. 2 (Sil sation, but comparison ofthe overall mations inthe ‘ime dorin (Figure E7Se versus Figure 3.1) and frequeney domain (Figure B75 versus Fig ‘we 3:13) shows signiNeant differences. ‘Ths aalyis described inthis example wa intended to ilustrate the effects ofthe equiv alent linea approach to approximation of nonlinear oil behavior on computed ground motions. Like Examples 7.21074 itcharactrized the actual soil conditions the Gilroy No.2(sol ste in simp ified manner. More detailed charaeterization (EPRI, 1993) produced significantly better agrement between predicted and recorded mevions. 7.2.1.3 Deconvolution ‘Because the equivalent linear approach utilizes a linear analysis, the response at any point can be related to the response at any other point. Although the transfer functions {developed in Scction 7.1.1.1 related to the computation of ree surface motion from bedrock