You are on page 1of 18
May, 1968 sus Journal of the | SOIL MECHANICS AND FOUNDATIONS DIVISION | __ Proceedings of the American Society of Civil Engineers ‘THE SEISMIC COEFFICIENT IN EARTH DAM DESIGN H. Bolton Seed,! M..ASCE, and Geoffrey R. Martin? INTRODUCTION | © The design of earth structures to eafely withstand the destructive effects | of earthquakes constitutes a complex analytical problem. Sudden ground dis~ Placements during earthquakes induce large inertia forces in embankments. ‘As 8 result, any one slope of an embankment will be subjected to inertia forces ‘hich alternate in direction many timeeduring an earthquake and its neces bary to determine the effects of these pulsating stresses, superimposed on the ‘initial dead-load stresses, on the embankment configuration. Because a satisfactorily designed embankment is stable under the initial |) dead-load stresses and the inertia forces induced by the earthquake are tran sient in nature, the ground motions will induce permanent deformations only | when they induce stresses exceeding the yield strength of the soll—that is, | mene no ety neon ‘sulficiently low, For rigid-plastic materials Or soils that show no appreciable plastic deformation until the peak strength {e developed, this will oceur whenever the factor of safety becomes less than nity. During an earthquake, the inertia forces in certain zones of an embankment may be sufficiently lazge to dropthe factorof safety below unity several times, ‘buy only for brief perieas of time, During such periods, permanent displace~ | ments willoceur but the movement will be arrested when the magnitude of the acceleration decreases or is reversed, The over-all effect of a series of large, but brief, inertia forces may well be a cumulative displacement of 2 section of the embankment; however, once the ground motions generating the inertia forces have ceased, no further deformation will oceur unless the soit strength has been decreased significantly. ole, Dlacusdion open iatil Ootobor 1, 1056. To extend the oostng date one mouth ‘a vrlifen request must he filed with the Exeoutive Secretary, ASCE, ‘This paper is part ‘Sr tho copyrighted Journal of the Soit Mochanies end Foundations Division, Proceedings St the Araorican Society ot Civik Engineers, Vol. 92, No, SHS, May, 1986, Manaaoript ‘Yas cubmitted for review for pooethls publication on October 26, 1963. 1 Prof. of Civ. Engrg, Univ. of California, Berkeley, Calif. | 2 Degkaner in cies ergng., inv. of Auckland, Auckland, New Zealand, 25 28 May, 1966 sM3 | ‘Thus, the magnitude of the deformations that develop will depond on the time history of the inertia forces, and a logieal method of desiga requires (1) fa determination of the variation of inertia forces with time and @2) an assess | ‘ment of the embankment deformations Induced by these forces. "The important concept that the effects of earthquakes on embankment sta- bility should be ascessed in terms of the deformations they produce, rather than the minimum factor of safety developed, was first proposed by N. W. Newmark. Newmark! and Seed® subsequently presented methods of analysis | ‘based on this concept which presume a knowledge of the time history of the | inertia forces acting on an embankment during the earthquake, t ‘Deformation approaches of this type are considerably more involved then | the conventional method of stability analysis for static loading conditions or | the method of seismic analysis which includes a horizontal static force in the | stability computations to represent inertia effects; both methods simply tn | volvo the computation of factor of safety for the slope under consideration. I the actor of safety determined by these analyses approaches unity, the em- | ‘bankment is considered to be unsafe, ‘Actually, euch an approach ie entirely warranted for thestatic loading con~ | ition which represents, in effect, a special and simple case of the general | problem of predicting embankment deformations. Under static loading, if the factor of safety becomes unity, the displacements will necessarily become 0 large as to constitute failure because there is no possibility of removing the driving forces rapidly enough to arrest the movement. During an earthquake, however, the drivingforees which may reduce thefactor of safety below unity, and thereby cause permanent deformations, are of a transient nature and are rapidly removed both during the earthquake (as the inertia forces reverse di reetions)andat the endof the earthquake (when the inertia forces are no long=t resent). ‘Thus, the deformation does not necessarily increase to an extent ‘which may be considered to constitute failure, Under these conditions, the de formation that develops will dependon the entire time history of stress appli~ cations and eannot be determined simply from a knowledge of the minimm factor of safety. INVESTIGATION OF SBISMIC STABILITY USING ‘PSEUDO-STATIC ANALYSES Past practice and most current practice i the design of embankments | against earthquake forces involves the compation of the minimum factor of | Safety againat slicing slong ® critical potential failure surface when a static, | horiaintal force of some magnitude is Included in the analysis. ‘The analysis | js treated a a static problem and the horizontal force ie expressed as the | product of aseiomic eoefiient, ky and the weight, W, of the potentias sliding Tass. If the factor of exety approaches tity, the section is generally cone Sidored! unsafe, although there iz no generally recognized limit for the miai~ | Fewnark, al, “Eartha Blots on Data and Enmbanients* proses et to ASH sraatal Engag Ce Sam #raeinca, aly oxen 1969, | oeuasis Nk Sbincto & Barca oo Dard eva Bmentoent,” Gales | sigue talon of ke, Engrs Londen, bagand, Vou X¥, No. dene, 1986. | fe By va Ncthod Fo Bartels Cogn of Barth Damo,” Jeu) tino Soh Sianlance snd Pounoatons Division, ASCE, Vl 9, No. ic | I Ae, Tansey, 186, sm ‘SEISMIC CORFFICIENT a mum acceptable factor of safety. In effect, the dynamic effects are replaced by a static force and the approach might therefore be termed a pseudo-statie method of analysis. ‘Meaning of Setsmic Coefficient in Pseudo-Static Anciyses,—To understand this approach, itis necessary to ascertain the meaning of *seismic coatficient.* Clearly, until’a meaning is attached to this term,an appropriate value for use {in design cannot be selected, ‘There appear to be three schools of thought on this question (as of 1966); 1, Some engineers believe that the product of the seismic coefficient and the weight of the potential sliding mass represents the maximum inertia force developed on the mass during the design earthquake. If this is #0, then the application of this force (which might actually act for a fraction of & second) as a staticforce is extremely conservative—other things being equal. Clearly, 4 large force ean develop for an extremely short periodot time without eaus- ing significant deformations, Because time is an essential factor in the de— ‘velopment of deformations, the application of a transient force as a static force would grossly overestimate its effect, In effect, adoption of this approach is equivalent to accepting the concept ‘that thedevelopment of any permanent deformation, no matter how small, con- stitutes failure of the embankment. However, this is rarely the case, Usually ‘designer will accept some limiting deformation during a major earthquake, in which ase theuseot a statie seismic cosificient to represent the maximum transient inertia force will be unduly conservative, 2. A second group of engineers consider that the selsmic coefficient is a means of designating the magnitude of a static force which is equivalent in ef- fects (i.e, produces the samo deformations of the embankment) to the actual dynamic iiertia forces induced by the earthquake. In this case—other factors being equal—the equivalent static seismic force would be considerably less ‘than the maximum transient inertia force. ‘But how would the seismic coefficient denoting this equivalent static force ‘be determined? Tt would seem that the determination of an appropriate value ould necessarily involvetwo steps,namely, (a) a determination of the degree of distortion or instability of tho embankment reaulting from the eazthquake ground motionsand (b) a determination of the statie force that woula produce |. the same degree of distortion or instability, Consideration of these two steps leads to the following conclusions; (1) Because step (a) is the ultimate objec ‘uve of the analysis, there would seemto be little point in ever malking the de termination required in step )); @2) the determination of the embankment deformations in step (@) could only be made once the time-history of inertia forces had been assessed; the time history of forces would thus be requized even if a static seismic coefficient were ultimately used in the analysis; (3) even if step (a) were completed and there were some valid reason to make step ®), there is no analysis presently available for making the determination of embankment deformations under static loading conditions required in step b); and (Q){t 18 moredifficult to estimate the end reault of both stope than to eeti- ‘mate or determine the results of step (a)alone. Thus, again, there would seem tobe little point in trying to make the determination required in step 1k would appear that any attempt to select a final value of such a seismic coetticient without going through step (@) and without a large backlog of ex- perience to guide the selection could have lilile reliable basis, In fact, at the oe 28 May, 1966 Ms present stage of inowledge and with little experience available to serve as a tguide, it would be virtually impossible. '3,’A third and more pragmatic approach, is to regard the seismic coefi!~ clent as anempirical constant that serves to designate a lateral force, the use of which tends to produce a more conservative section for the embankment than ‘would otherwise be adopted. ‘This is a perfectly reasonable objective but, un~ fortunately, it 1g nol always realized, For example, i a designer accepts a low factor of safety, say 1.1, for earthquake design conditions, then the inch= ‘sion of a seismic coefficient less than 0.1 in the analyses will often elimincte the need to modify the embankment section from that required to satisfy sta:ic stability requirements. Whon this situations accepted, the embankment eross Section is the same as sf would have been if there were no seismic hazard at the site, Ih this event, either the embankment is not designed to resist the earthquake at all or the designer is convinced that it is unnecessary to make fany concession toward increasedstability due to the fact that the embankment 4s to be constructed in a seismically active rogion. When adopting an empiz~ {cal approach, it should also be borne In mind that selecting or specifying a design selemic coefficient without simultaneously specifying the method of ‘analysis with which itis to be used may be relatively meaningless. Thus, for ‘example, 2 designer may be requlred to design for a seismic coetficient of 0.15. However, he is rarely restricted in the method of stability analysis, al ‘though miner modifications inanalytical procedures may havoalmost as much ‘etfect onthe computed factor of safety as the inclusion or exclusion of the seismic force, Therefore, specification of a seismic coefficient without spec tying the method of analysis may aot achieve the desired effect, Selection of Seismic Coefficient in Pscudo-Static Analyses.—Whichever of the above three concepts of the meaning of the seismic coefficient an engineer may adopt, he is necessarily faced with the problem of assigning a value to it for designpurposes. Various mothodsare available for this purpose including the adoption of empirical rules,the assumption of rigid-body response,or the use of elastic-response analyses. The applicability of these various approaches ‘will now be examined. Use of Empirical Velues.—Most engineers in the United States, who use a pseudo-statie method of seismale stability analysis, adopt some empirical value for the designseismic coefficient; generally, this value is in the range of 0.05 to 0.15. Although there is good reason to use some value of this typo to di- Serentiate between the seismicity and foundation conditions at different sites oF to study the advantages of different dam sections, there appears to be no pablished justification for using values in the range of 0.05 to 0.15 as a basis {or selecting or approving final design slopes; nor is any logical argument for the use of these particular values offered by most engineers adopting them. Generally, a value is selected simply because It Is used by other engineers and, although there is some belief tha! such values may have a rational basis, ‘most engineers are unaware where this can be found. It appears, therefcre, that confined use of these empirical valuos has given them some semblence | of m authoritative design crterion—yet no one seems to know why the values | ‘were selected in the first place. ‘The following are possible reasons for using seismic coefficients in the range of 0,05 0 0.15: ne SEISMIC COEFFICIENT 29 1, The values are of the same order of magnitude as the value suggested by Westergaard® for determining water pressures ontheface of concrete dams during earthquakes quite diferent problem from embankment design. 2. The values are of the same order of magnitude as seismic coeificlonts presently and formerly recommended for structural design of buildings in seismically active regions. Here again, however, structural enginoero using these values recogaize that there will bconsierable plastic deformations of ‘uileings during a major earthquake eventhough the initial factor of safety may bo of the order of 2or 3. Factors of safety in earth structures are consider aby leas than this and plastic distortions during an earthquae are therefore likely to bo not only greater but also more serious In thelr consequences, 5. Tthas been suggested that values of 0.1 and lower have considerable ap= peal because their use, in conjunction withthe acceptance of a lower factor of safety or earthquake conditions, often will not require any change from a de= sign providing a reasonable factor of safety for static stability. ‘Tavs, an ene gineer who is convinced that earth dame aze inherently stable structures and requireno special analyses toaseure their stability during earthquakes, might bbe subconsciously influenced in his chotee ofa setsmie coeticient by & valve providing the simultaneous advantages of (1) corroborating his belt, Q) ree auiring no design mocifications, 8) satistying the requirement of regulatory agencies for a seismic analysis of some type, and ()satisiying his conscience tat an acceptable check on selamic stability has been made, Because no reasonable basis for adopting seismic coefficients of this order of magnitude (0.05 to 0.15) for earth dam design has o far been advanced, the validity oftheir use cannot be evaluated. Because few major dams have been subjected to major earthquakes, theres Itle experience available to serve a5 2 guide. Anengineer may follow current practice nd justify theusect some value onthis basis, Certainly, this tends to produce & wiiformapproach to the design problem. However if present practice has not really been tested by 2 major earthquake, adherence to a uniform approach may just as easily lead 9 my embankments hengfound inadequate asto their performing satisfactorily, 1 ssentirely possible that empirical values on he order of 0-1 or 0.15 may lead to safe design conditions n many cases but, until some means of judging their validity is developed—either by comparison with more rigorous methods or by a comprehensive seriesof observations oftheir influence on the bebau for of dams subjected to iown carthauake ground mations nd this, would probably involve a determination of how the cain would have behaved ifthe equivalent statieforce had not been inclided inthe design or if seme afferent value had been used)-thelr use must be considered of questionable value in fhe design procedure, Because an assessment of their value by observation of dam performance ts probably economically impossible, a judgment may ‘ultimately have to be rendered by comparing their effects vith those of more rational procedures. For this reason alone the development of & more Fatio= mal design procedure would be of considerable vaiue to the profession Eis interesting to note tha, whereas the design seismic coefficient is typ- ically on the order of 0.1 in the United Slates, somewhat higher values ae used in Japan.7 The design seismic coefficients given in the design criteria —TWestergaard, HM, “Waler Presbures on Dane During Earliquibes? Tan ations, ASCE, vel 9, 1833, Dating Bastia? Tate sera Eertauane caste Design for Civil Engineering Structures, Earth Struchres and Foundations in epan,” Report, Japan Soe. ofCiv: Ea, Tajo, Japs 1880, 30 May, 1966 ams for earth dams established in 1957 by the Japanese National Committee on LargeDams, and commonly followed in Japan, range from 0.12 100.25 depond~ ing on the location of the dam, the type of foundation, and the possible down streameffects of damage caused by anearthquake. Design values are reduced by 50% for conditions of empty reservoir on the grounds that damages oceur~ ring under these conditions would not be serious, Rigid-Body Response Analyses.—It 2 dam is assumed to behave as 2 rigid body, the accelerations will be uniform throughout the dam and equal at all times to the ground accelerations. ‘Taus, it is sometimes argued that the de~ ‘ign seismic coefficient should be equal fo the maximum ground acceleration, ‘The main limitations of this approach are 2s follows: 1, Although low, stiff embankments or embankments in narrow canyons may respond essentially as rigid structures, there is considerable evidence from, field tests in which actual dams have been subjected to forced vibrations by. ‘means of large shaking machines that all earth dams do not behave as rigid bodies but respond in difforent ways to any given series of imposed motions. Hence, all dams will not respond in the same Way to any given earthquake, Furthermore, it would appear that the magnitude of the aecelerations in a dam may be greater or less than those of the foundation, depending on the natural period and damping characteristics of the dam and the nature of the ground motions. 2. The maximum acceleration will only be developed in a dam for a short period of time so that the deformation resulting from it may be small, Al- ‘though it will be supplemented by deformations produced by other accelerations and inertia forces developed during the earthquake, there is no reason to be~ lieve that thetr combined effects will be equivalent to those produced by 2p- plying an inertia force, corresponding to the maximum acceleration inthe dam, a a static foree, ie., a6 if it were acting for an unlimited period of time, Considering these Limitations, there is nothing to gulde the design engineer on the appropriateness of using a selonale coefficient corresponding to the maximum groundacceleration ina pseudo-static analysieot seiomic stability ‘or on how this value might appropriately be modified for afferent structures and ground motions. ‘Elastic Response Analyses.—The deficiencies inthe useof empirical rules ‘or the assumption of rigtd-body response have led several investigators propose the use of elastie response solutions for the determination of design Seismic coefficients. ‘The first dynamic response analysis of an earth dara, made in 1936 by Monobe, Taksta, and Matumura,® iavolved the following simplifying ascump- (1) The dain consists of an infinitely long symmetrical triangular section, resting ona rigid foundation; (2) the dam consists of homogeneous and linearly Slastie material having a uniform modulus and density; (3) the wicth to height Zatloofthedamis sufietetly large that bending deformations canbe noglectec and deformations resulting trom shear need anly be considered; (4) the shear stress over any horizontal plane is uniformly distributed; and 6) the influence ‘of stored vater ie negligible. In effect, the dam ls considered to consist of & Ttonsnabe, N., Takata, A.,andNatunra, i, Seismio Sabi of the Earta Dem) Procesdings, tad Congress on Large Dems, Washington, D. C., 1986, Val. 1V. “sms SEISMIC COEFFICIENT 31 ‘series of infinitely thin horizontal slices, the slices being connected by lin- ‘arly elastic shear springs and viscous damping devices. ‘The response at different levels resulting from 2 uniformly distributed base motion is etermined, ‘Tho gonersl formot the analysis may be illustrated by considering the free Vibration of such a dam for the case of zere damping. ‘The forces acting ona thin slice at depth y below the crest are shown in Fig. 1 and, for equilibrium, itis readily determined that 2 2, ou [ote a aul oe s[% | eae In which w= displacement at depth y in the x-direction (horizontal), t= time, = mass density of the dam material = y/g, y = unit weight of dam material, ‘and G=shear modulus of dam material. ‘The solution of this differential equa eee Sefer 5 FIG. 1.~ONE-DINENSIOWAL, SHEAR SLICE THEORY tion forthe appropriate boundary conditions: (1)u = 0 when y = h for all values || att and (2) au/oy = 0 when y = Ofor all values of tis given by. wo) « "Ea, ome, t +B, co, J ag(e, Z ® Jn which n= helght of damn =the number of the mode, fy» the zero vale of | the frequency equation JgloV/0 8) = 0 having atid Yalu for each mose | Ga. b1 = 2404, Bp = 5.82, Bg = 865, a4 = 11.0, ee), Joly 9/M) © mote | ape at vibration for the ath mode, corresponding to the Satara tregeaney on, q =the natural frequency of the dam in the nth mode = fy/h VG/e radians | Bet stood; JqDessel function of fart kind and order aerovand A aed Bee | contanta determined bythe intial conitons | From this solution the undanped natural frequencies of the dam, dn, in | erent modes can readily be determined, Xs interesting to note fet the | term G75 is the velocity ot shear wave propagation vg, inthe darn, te thet the natural frequencies in different modes may be written 32 -@ ‘The response of structure to a random ground motion can he studied by ‘considering the ground motion to be zero and the structure excited by effectiv2 forces equal to the mass at any point multiplied by the ground accelerations. Hence, if tho samo dam is subjected to a random horizontal ground motion ‘with displacements, ug(), velocities Ug(t) and accelerations, g(t), the equa | tion of motion (or zero damping) becomes eee) oe ‘m which u=the horizontal dlsplacement at depth y duringthe earthquake, rela- tive to the base of the dam. ‘Making use of the orthogonal properties of mode shapes and the principle of mode superposition G.e.,the over-all response of a structure is the sum of ‘the responses for individual modes considered to act separately), the general solution of this equation becomes. neo 24, (, » t 2 ana Seem [eat & esa oe welt od ses ee teas] a nel cos area eee aaa sue adc an ara Eo ers hope chy eager rereterter pate ee eee G0 +O) mee 23,(% 2) ot -rgigt?) ug . sin [wyyt-r)] ar. 8) 4, nt an Fn Yon) “0 # in which gq = the damped natural frequency in the allt mode = wn VI - Ay2 = top for small values of A,and An = the fraction of critical damping for the nth ‘mode, "The total or abeolute acceleration, tig(,t), acting on the dam at any time, 4, ds given by out 890+ HO. vm land can readily be determined from the values of w(y,t) expressed by Bq. 6 land the known values of the ground acceleration. In the cazeot zero damping, it may be shown from Bq, 4, after normalizing to uneouple the modes, that the modal contributions to the absolute accelera~ tons are given by 2 Wt) = 4,7 a.t) se @) In the damped ease, Eq.# is a good approximationto the absolute accelera~ tion for small values of X. Thus, from Eq. 6, assuming that waq ~ ty, the ab- SEISMIC CORFFICIENT % ge Fe # | HIG. 2.-ACCELEROGRAM, QUAKE, MAY Ie, 1840 908 EL CENTRO (CALIF, RARTH- COMPONENT} HIG, 3. VELOCITY SPECTRA, BL CENTRO CALIF.) EARTHQUAKE, MAY 18, 1940 (4-8 COMPONENT) r + L Al wy Lage eed tL FIG. 4. ACCELERATION SPECTRA, ELCENTRO (CALIF) EARTHQUAKE, MAY 18, 1840 0-8 COMPONENT) 34 May, 1966 Ms Seamed ly sinfo,-r]}ér. 02) Although this equation is apparently cumbersome, ite use can be considerably simplified if tt is rewritten as aya = (8b) in which a0) ) 25,(8, 2 +3) 70,07) and ‘sin [o, teri] er +s) t t+ fiwe ve J By 1 may be noted that: 1, The natural frequency for any mode, wy, can readily be computed from ‘knowledge of the dam height, the shear modulus, the mass density, and the fixed value of fq applicable to that mode, 2. Although the absolute accelerations in any mode (mode n)vary with time, | 45 shown in Eq, 10, the only term involving ime is the integral Va), Thue, the absolute acceleration in any mode will be a maximum when the value cf ‘his integral, for the particular Values of um and A, applicable for that mode, is a maximum, ‘Tae integral Vp )in Eq. 13 i called theDuhammel or Convolution integra: 1s value, which has the units of velocity, depends on four parameters, namely, the ground motion characteristics, tig, and their variation with time; the frac {ion of eritical damping, Ans the nalural frequency wn; and the time, t. For any given ground motion, it will havea maximum valve Vp (max match WL ‘depend only on the values of Xp and @. Such a maximar value 1S commonly called a spectral velocity, 8. For convenience, the relationship between the spectral velocity, Sy, the (raction of eritieal damping, Any and the frequency, ‘Gp (Or period Ty = 31/s%q) can be evaluated for any given ground motion an ‘ednelusively plotted for that motion, Such a plot is called a velocity spectrum; for example, tho N-S component of the ground accelerations for the El Ceniro (1940) earthquake is show in Fig. 2 and the velocity spectruia, for severe valuesof 2, corresponding to these motions, is given in Fig. 3. Using this type of plot, the maximum valueof the integral Va(t), corresponding to the particu lar values of wp and 2p, can be read off diecty. ‘3, Instead of computing the maximum absolute acceleration in any mode from the product ‘SEISMIC COEFFICIENT [ina Ui ie convenient to determine directly the value of the product ty 8. Such flues are called spectral accelerations, Sq, in which = 4, 4,0 8,» ‘| = 15) plotted directly for idiven ground motion in terms of Jo, (or Ta)and Ap, the resulling plot com- nly called an acceleration spectrum. Such a spectrum for the El Centro 40) earthquake is shown in Fig, 4, By reading off values from the acceler- 8 OS, eee eeeeee $165 cosa ot spectral velocities, rs of tny bo IG. 5.~RELATIONSHIPRETWEEN #09 AND yh BIG. 6.—SHISwIC COBFSICIENTS suG- GESTED FOR USE IN PSEUDO-STANIC ANALYSIS ation epectrum, the maximum acceleration in any mode may be determined from (Oa = ba”) 8, 4. The term gn{y) 15 a funetion of only two parameters, fy and y/h. Be~ cause, for any given mode, the value of fj, is fixed, the term gq may be con clusively plotted as a function of y/h for different values of 2 (Fig. 8), Thus, values of gy may be read off directly for computation purposes. ‘The function gy(y) actually expresses the modal participation in the re~ sponseof the dam and the distribution of accelerations inthe individual modes. ‘The distribution, in effect, represents the contributions of applied acceleration acting in each mode when 'a uniformly distributed time-dependent acceleration is applied to the dam. Hence, the modal distributions plotted in Fig. 5, when summed for an infinite number of modes, will give 2 uniform distribution of unity, L., = (18) 36 May, 1968 M3 FF go) a... am) mi It may be noted that higher modes contribute increasingly lees to the over-all acceleration. 5, The maximum value of accelerationin any mode is obtained by magnify- ‘ng the valuos of gy (Fig. 5),by the spectral acceleration corresponding to tie frequency of the mode, in accordanee with the equation a Tlgae "FH By ooo Generally, significant magnification of the values of guy) oceurs only tn the first few modes, because the frequencies of earthquake waves having the great est accelerations lie in the frequency range of these modes for most earth dams. As a result, the contribution to the over-all agealerations of modos higher than, say, the 4th can generally be neglected. 6. For a rigid structure, the spectral acceleration, Sq,is equal to the max ‘imum ground acceleration (llg)may for all modes. Thus, -{18) [gamer * #40 Caan 08) snd ocause inthis cate the maximum responses In al mades are developed atthe same lime, tho sisum acceleration fe te eructane | | | (49a 3 Cane ‘ as would be expected, Therefore, the response analysis is eapable of pre- dicting the performance of rigid as well as elastic structures, provided a Sif ficient number of modes are considered. ‘I. Ia general, for elastic structures, the maximum responses in different modes will occur at different times so that direct superposition of maximum values is invalid, : 2, Uy Mphnaed | | | | From the distribution of acceleration determined from euch an approach, the dynamic inevtia forces can be determined because, for any element of weight, AW, the inertia force will be ale Soo (20) * SEISMIC COEFFICIENT ar smnatively, if the dynamic inertia force is expressed in terms of a seismic thieient KG), AT = AW RG). = @1) “Hence, it follows that ky)= 2 Yo) = (22) aud appropriate values ean readily be obtained once the acceleration values have been computed. ‘Although the preceding Is limited tothe ease of an infinitely long das with a symmetrical and-homogencous cross section resting on rig foundation, ‘similar analyses have been developed for other conditions. Hatanaka® extended the analysis to include the case af a trienguay elastic wedge ina rectangular canyon and considered the veritionof horizontal response over oth th length and heightof the dams, He showed that when he length of the dam ts about fous times the height, the nfisonce of end restraint has negligible etfec onthe natural frequencies ot vibration and the magnitude of response inthe central resion;hence the use ofan analysis basedon the assumption of inte length ‘an be considered suificientiy accurate for all practical punposes ‘Subsequently, Ambraseys10 developed solutions for the case of a truncated vero of untform modulus and for 2 symmetrical wedge of taiform modslos resting on an elastic layer of finite thickness, More recently, Rashid! ana- lyzed the response of asymmetric wedge in which theshear modulus increased 25 the cube root ofthe dopth variation more appropriate for dams composed of cohestonless materials, Despite the availability ofthese solutions, the only specific suggestions for their ute in design have been made by Ambraseys,10 who preposed thatthe Interal forces acting on a dam during an earthquake’ be enpreseed by atte force with mageitude determines by a setemi coetticient evaluated by one of the following methods: @)The seismic coffiient at any depth should be taken 25 the square root of the sum of the squares of the selenteeocifctents for Peak roaponse in the fret four modes, he, na v2 Ro) =| ¥ 0} ae oF b) The setsanie coetticient at any depth should be taken as the maximum value at that depth for any one of the moeal distributions, i.e., *0) = [hy OMmax ** Both expressions, Bus. 28 and 24, represent an attempt to estimate the maximum seismic coefficient otribution acting at any Instant daring the 023) seer OA) —$ Tstanaia, Mi, “Fundamental Conslaeralions onthe Ravina Resietant Propertin of the Earth Dar,” Bulletin No. 11, Disaster Prevertion Research int, Kyoto Unive Kygtp, Japan, December, 1058. BW'Ambraseys, N. N., “The Seismic Stability of Hath Dems,” Proceedings, 2nd World Con, on Barthuake Brgrg., span, 1980, Vol, ‘LT Rashia, . ., “Dyemic Response of Earth Dame to Earthquakes," Graduate Stugent Research Report, Ualy, of Calorie, Berkeley, Calf, 2901. 38 May, 1966 earthquake, Similar recommendations were subsequently advocated by Krishna.12’ Application of these approaches using values from k(y) computed from Bqs.22 and 9a leads to adistribution of seismic coefficient varying from ‘a maximum value at the crest of the embankment to zero at the base. Used in this manner, the dynamic response analyses are only intended to determine the maximum inertia forces likely to be developed in a dam at any Instant during an earthquake and these values are incorporated in a pseudo~ ‘static analysis of stability. This approach suffers from the following limita> tions: sus 1. Use of the computed seismic coctticients in a static stability analysis leads to the conclusion that if the factor of safety becomes equal to unity the ‘dam will fail. Actually, the computed inertia forces are dynamic and the fat torof safety may be reduced below unity fora short perlodor a series of short periods without causing significantiy large deformations, Alternatively, this ‘method of analysis assumes that any permanent doformation, no matter how ‘Slight, constitutes failure of the embankment, Many engineers would disagree strongly with this,concept. 2, Any response analysis that considers the dam as a series of slices suffers {rom the following Umitations: (a) The analysis is based on the as- ‘sumption that the response of the dam to the ground motion is controlied only, by the shearing action developed between horizontal stress, However, the ‘work of Tsitzaki and Hatakeyamal? and Clovgh and Chopra! indicates that both horizontal and vertical compressive and tensile deformations within an ‘earth: dam, resulting from horizontal ground mations, ean contribute signifi- cantly tothe over-all dynamic stress pattern;and (b)the analysis is developed only for horizontal ground shaking and does not consider the effecte of ground motions in a vertical direction, 3. The approach assumes elastic deformation in the soil mass with energy being dissipated in terms of viscous damping. However, soil deformations under higher stresses are generally inelastic and Little is known regarding the nature of damping forces in soils under such strosses, as a function of strain amplitude and frequency of vibration. The development of small plastic or i elastic deformations in regions of the soil mass during a strong earthquake would result in a modification of the magnitude and distribution of accelera- tions given by the visco-elastic theory due to the additional energy absorted Sn these regions. Because a certain amount of plastic deformation in a dam ccan be tolerated during & major earthquake, consideration must be given 10 its effects. Although the useof increased viscous damping factors can rediee considerably the magnitude of forces within a dam, ina manner similar 10 plastic and inelastic deformations, it is difficult to establish a valid equiva Tence between the two mechanisms, 32 isina, J, ‘arthquake Resistant Desiga of Barth Dams, Prosondings, Fara ‘quake Sympostun:, Roorkee Untv., Roorkee, inaia, November, 2663. “5 Ishizaki, I, and Hetakeyatna, N., "Considerations ofthe Vibrational Behavior 2t Barth Dame," Bulletin No, 52, Disaster Prevention Research Inst, Kyoto Unies Kota, Japan, Pebrusty. 100%" 14 Clough, Ray W.,andChopre, Anil K., “arthquaie Stress Analysis in Barth Dams," ournal of the Bn ‘Meohantos Division, ASCE, Vol, 02, No. EM, Proc. Paper oar Aol eee pee Ms SEISMIC CORFFICENT 39 4, Incorporation of the seismic coefficients ina psoudo-static analysis does not take full advantage of the merits of a dynamic response analysis. ‘This point will be subsequently examined in greater detail. | Despite its imitations, an elastic response analysis does provide a quali- {ative pictureot the acceleration distribution in a nonrigid embankment; als6, st enables considerationof the characteristics of the embankment fin terms of its natural peviods of vibration) in assessing its responge to any ground mo- tion. In view of this, Hatanaka® proposed 2 simplified empirical approach in which the design seismic coefficient at the erest of adam is reduced as the | tabural period of the dam inereasos and the distribution throughout the height is assumed to be the same as that produced by vibration in the first mode. Appropriate reductions are made near the ends of the dam in cases where the canyon walls can restrain the movement of the embankment. Unfortmnately, Hatanaka does not indicate how his recommended values would vary with the intensity of ground motions. Comparison of Methods jor Determining Seismic Coeffictents.—In view of the different approaches currently (as of 1966) used for determining seismic | coefficients for use in pseudo-static analyses for design purposes, it Is tater= esting to compare the values obtained by the various methods. ‘This is best | Mustrated by 2 numerical example. Suppose values aredetermined for a long dam, with a homogengous section 300 ft high, construeted of compacted soil having a shear wave velocity of 1,000 fps (a vatue typical of modezn cohesive alluvial fills) and subjected to the N-S component of the El Centzo vartiguale of 1940, for which the maximum ground acceleration was about 0.58. Values determined by the following methods are plotted in Pig. 6: Pot ‘empirigal design coefficient of 0.10 to 0.18 (typteal United States practice); 2, a design coefficient in the range of 0,16 to 0.25 (Japanese code require- ment); 3. a design coefficient equal to the maximum ground acceleration, corres- ponding to the assumption of rigid-body response; 4. Ambraseys’ recommended values, determined by an elastic response analysis for 20% critical damping, using j | \ | 4 h/2 @) Rig) = [ z f.0}%ay] } oT | ©) XO) = Dk, nan | 5. the values given by the Russian code cited by Ambraseys, using n = 0.23, | Gor arelatively strong earthquake)and values for damping coellicien! of 0.2 and 0.3. Apparently; an engineer has a wide range of choice in selecting a seiemic cvetfictent for design use. The wide range of values and opinions currently in tse is a serious limitation of present practice and presumably reflects theun- Certainty of engineers concerning this aspect of embankiuent design. ‘The de velopment of rational method for determining which of the proposed values May, 1966 ‘represents an appropriate balance between safety and economy would be a sig- nifieant development in earthquake resistant design procedures DESIRABLE INNOVATIONS IN DESIGN PROCEDURES Apparently, attempts todotermine astatie solemic coefficient to incorporate 1m a pseudo-slatic analysis of the seismic stability of a dam leave much to be desired. Empirical methodsof selecting a coefficient have little rational basis, are based on past precedents rather than past experience and, although they ‘may sometimes be adequate for thelr intended purpose, they aleo mislead the engineer into a false sense of security. Seismic coolficients representing the ‘maximum inertiaforee to which a dam might be subjected, whether determined by nigid-body or visco-elastic response analyses, can only be justified if the Aesigner accepts the concept that any deformation, no matter how small, con- stitutes failoreof the dam (@ concept whieh is both unnecessary andunoconoz— ical). Also, attempts to determine a static foree, equivalent in ite deforming effects to those of the earthquake ground motions, are both unnecessary ard tmnattainable because they must necessarily be based on a prior evaluation of the dynamfc response and deformation of the embankment whieh, in iteel, i8 the objective of the analysis. ‘In view of these facts, it appears that the complexity of the analytical prot lem is, in fact, increased by attempts to represent the carthqualte effects by & static Seismic coefficient and that a simpler, more rational approach could ke developed by determining dynamic seismic’ coefficients which represent tke ime history of the inertia forces to which the embanitment will be subjected by any given ground motion, In fact, evaluation of such forces is a necessary prerequisite to any rational approach to the problem predicting embankmest ‘deformations during earthquakes. Ih addition to the foregoing arguments, there is growing evidence to show that soll strength mobilized during earthquakes is quite diferent from that determined under static or transient loading conditions and ie a function of the entire time history of the stresses developed during the earthquake, Typical ‘examples of thie evidence are as follows: 1, The failureot saturated sands caused large slidesin the recent Alaskin ‘earthquake and foundation failures in the subsequent Niigata earthquake. No failures were produced in these sands by relatively high stresses applied statically by footings or by steep slopes, but complete failure by liguificatien ‘occurred under relatively flat and gently sloping areas as a result of modest stress increases produced by the earthquakes. 2, Failure of saturated sands causing sliding in the Alaskan earthquake oc- curred after 1 min or 2 min of severe ground shaking. No significant defor rations would have developed in many of the slide areas if the duration of the ceartixuake had been only 20 sec to 60 sec. ‘Thus, the duration of pulsatirg stress applications or the number of pulses developed is a significant factor {hat must be considered, as wellas the appliod stresses, in determining whether failure will occur. 3, Sliding in sands in the Alaskan earthquake ceased as soon as the ground motions ceased and a static stress condition was restored. sm | SEISMIC CORFFICENT a | 4. Studies of slope failures in tho Alaskan earthquake indicated that satu- tedeensitive clays may also lose strengthunder pulsating loading conditions. 16 5. Laboratory tests show that soil strengthunder eyclie loading conditions {8 quite different from that mobilized under static loading and is a function of the number of stress cycles as well ag the stress intensity. Becauseboth fieldand laboratory evidence demonstrate the major influence _of the time history of stress applications on soil behavior, failure to consider | this factor would be a serious deficiency of any design procedure. ‘Thus, from considerations of both soil behavior and deformation analysis, the development of some method of predicting the dynamic inertia forces (or the corresponding dynamic seismic coeificionts)and their variation with time, | would appear to be essential for design purposes. It might well be questioned way a pseudo-statfe approach might be satis- | factory for structural design purposes but not for earth embankments. Two _ major factors contribute to this situation: (1) Dynamie response theories for buildings are relatively well-developed and thelr applicability hasbeen checked by field tests. Thus, some basis for determining an equivalent statie seismic coefficient ss available. (2) Observations of the performance of structures in _ previous earthquakes have provided literally thousands of full-scale tests on which tobase the selection of appropriate design seismic coefficients. Neither of these conditions exists with regard to earth structures, Dynamiie response ‘theories are not well-understood and there is @ severely limited number of luseful case histories of embankment performance during earthquakes. Even ‘when failures of embankments or earth slopes have occurred, quantitative data ‘nthe soft conditions.and earthquake ground motions producing hem has rarely | been obtained, Thus, it is virtually impossible to benefit, in a quantitative sense, from past experience. In view of thie, it would seem especially desir able that dnamicresponse theories be developed toprediet embankment per formance and also check the appropriateness of pseudo-stalie approaches, DETERMINATION OF DYNAMIC SEISMIC COEFFICIENTS Acceleration Distribution in Embanbmente,—The simplest approach for de- torminingthe time historyof dynamic stress induced in anembankment by any given sories of ground motions is to consider the embankment as a rigid body, firmly attached to the ground surface, In this case, all elements of the em” bankment are subjected to the same accelerations as the ground surface And the time history of inertia forces on any potential sliding mass is identical in form to the time history of the ground motions. For small, tiff embankments or for embankments in narrow canyons where the embankment is constrained tohave the same motions as the canyon walls and base, this approach would be entirely appropriate, ‘There is, however, a growing body of information to show that many em= ‘bankments do not behave in this fashion, but respond as deformable bodies to the ground motions. Hence, analyses based on the concept of viseo-elastic response are likely fo provide a better indication of the actual stresses during ‘an earthquake. The following evidence supports this concept: 2 May, 1966 sms | smo SEISMIC COEFFICIENT a 1. In recent years, several dams, about 100 ft high, have been subjected to = forced vibrations using large shaking machines.15,18 ‘The response of tte dams, a typical example of which 1s shown in Fig. 7, has clearly shown that peak accelerations are developed at certain characteristic frequencies and, furthermore, the response is in good agreement with that predicted by visco~ clasticresponse analyses. (Arigid body would exhibit the same peak response independent of the frequency of vibration.) 2, Viseo-elastic response analyses show that the accelerations induced in an embankment inerease with height above the base, This type of behavior is in good accord with that observed (a) in model tests of small-scale embank- ‘ments, (&) by seismographs installed at the base and crest of the Cachuma Damn in California during a small earthquake in 19677 (the maximum grou ‘acceleration was 0,01gbut the maximum acceleration at the crest was 0.025¢), ‘and (c) in measurements of tho response of the Sannokal Dam (pan) to feral small earthquakes.!8 Theaccelerations recorded at the base, midheigit, IG. 1. RESPONSE oF land crest of this dam during an earthquake are shown in Fig. 8; apparently, Eceheareteaiacuiceee characte the accelerations arenot uniform throughout the height as assumed in a rigit- | body response approach. | aaa uaa Furthermore, because viseo-elastic response snaivseoare widely wed ®t | ss Hlnl Hi ! assessing tne eaponge of buildings andother strvetares to earthquake ground i i cm arene Te to oarth structures el “e ‘Tus, t appears that a viscovelastic response analysio currently provides b eae the meni reneonabie approach for assessing the dynamic forces induced in an . fmbankmont by an earthquake, and the shear slice analysis deseribed previ~ Susly may be Used for this purpose, However, il fs not sulficert to assess Stet the lasimum stresses; the analysia must He wilted to provide the time 2G. #.~OBSERVED RESPONSE OF SANNOKAT DAM History of stresses throughout the easthgako, Fortunately, with the aid ot Jnodarn computore, this presents no major obstacle, ‘The distribution of ae~ Ecloration wth helgt inanembarkment, having a length greater than three OF | feu times lis height, at any instant of time, 38 expreesed by Eq. 9. “ or any specified variation of the ground sceeteration, Uy, with time, Fa. ta may Teady be programmed for a computer to detersin€ the distribution ot acceleration at any given series of instant of time. For iustration, Fig. — 410 presents the results of such computations for a 100-ft dam having a con- | pA ee stant shear modulus G~4% J0® pe, damping factor, A,of 0.2 for exch mote, nda deneityof 190 I per eat, when subjected to the ground accelerations iFis.2) recorded in the £1Centro earthquake of 1940, Using Bq. 8 and cuper- Imposing the response for the fist six modes of vibration (sulficient to give. an aceurato representation ofthe actsal acceleration distribution), the distei= buttons of accelorations wore evaluated a time intervals of 0.02 sece 1B Reiptey, W. Ox "Vibration Tesi of Sructare,?Earthake Engrs, Roveaeoh | tah Calle Insts of Tech Faandenns Cay ly, oe Keipuley, Wr Ou 2a ‘bynamie Rvertiton 0 Boigat Canyon Daz” ant- quate Soere Tooter Lats: Galore bet of Tech, Paonten Gal, Sptomber, te HH petit) 1 Ambraseys, Ne Noy “On the Seleraio Behaviour of Barth Dams,” Proceedings, 2nd World Cont. Gn Barthquake Bagrg., Japan, 1960, Val 18 Oeamnoto, & Motohiko, K., Katsuyuki, K., and Fusayoshl, K,, “On the Dynamat Behaviour of en Earth Dams During Earthquake,” Progeadiags,” 3rd World Conlon Earthquake Engrg., New Zoaland, 1986. FIG, 9.-COMPOTED RESPONSE OF 109-PT- CENTRO EARTHQUAKE baterneicted SEISMIC COEFFICIENT Aervals of 0.1 vec, for the first $ sec of the ground motion, are shown in Fig. 4s apparent that the distribution of lateral acceleration vari ‘erably, ranging from 2 fairly uniform distribution to an almost triangular | distribution. For this particular dam, the first mode of vibration is predom- inant (T} = 0.262 sec) and, consequently, nodal points in the acceleration di tributions rarely occur. However, for higher dams, where the higher modes | of vibration play a greater role, it ispossible for the accelerations to reverse directions several times along the height of the dam at any given instant of time. By plotting the accelerations at different heights in the dam as a function of Lime, the entire acceleration response throughout an earthquake can be ob- tained. ‘Such plots or the base, midheight, and crest are shown in Fig. 9. The amplification of the groundaeeelerations with increasing height of the embank- ‘ment is readily apparent, as isthe agrooment botwoen thoform of these results and those observed on an actual dam (Fig, 8) Feit) oe pay FIG, 11.~CONCEPY OF AVERAGE SEISMIC COEFFICIENT EO OS FIG, 12. WEDGE IDEALIZATION OF POTENTIAL SLIDING MASS The Concept of Average Seismic Coefficients.—Raving obtained the time history of acceleration distributions in an embankment as described above, it is possible to evaluate the lateral force acting on any potential sliding mass at any Instant during the earthquake. Referring to Fig. 11, the total lateral force acting on the sliding mass bounded by the arc AB at aay particular in- stant is given by F- Dmg) ¥0) 0 cree lQ5) | which m= the mass of an incremental slice of the mass and lig = the cor | esponding absolute acceleration of the sliceat the instant under consideration. |The lateral force acting at this instant may alternatively be expressed by an average seismic coefficient, Kay, in which & 2 & : g 3 & i i i : é gi i £ g E 8 May, 1968 F sus (23) Kgy Woes and W = the weight of the sliding mass, ‘Thus, at this particular instant under consideration Bat ww Similarly, the value of kay may be obtained a othr instants of time during tho carinjuate sad the obulting values ptt agtnst time to give the come piste ime history of the average seismic coefficient acting on the elldiig ‘The evalustionof such average values of seismic eoeficlent eliminates the nocousty of eoneiaesing the couplex aatureot acceleration distibaions over the height of the damn. Oving tothe irregular satare of he acceleration ise tefoulens, it ie inecorate mevely 10 average the acceleration datributoe Toclf; considering the paticnay dhetsibation shown in Pig. Hay it ean be seen tint ereas tne average value of kg) is near zero, the lista force, May ¥, sutng onthe potential slding mass, could be conaerable because the lower stebution of kG) acta on a greater proportion of the sliding mass than he Cpper dlatribution, Hence, to compute average valuse of the seismic coe! Clout ts necessary to apeeay tho boundaries of tbe potential Siding mas6 Smvotied. ‘Theorettcaly this wosld lead to 9 aerent vue of the average Sclome coelfcieat for cach potential alidiog ase at any intant of irae. | ‘Tals complexity con be eliminated, however, i tia assumed that the fo~ tential stain mass can be represented with satstactory accuracy y a tic ngulae wedge as shown ia Fig. 12. This approximation algo permite Ne trerage seismic cosficient acting over any partcvlar depth to be computed alrocly and avosas the necessity of Using & sumination procedure a8 sbore- ‘Tuc is accomplished by evaluating the shear force eaused by the ground mo tions acting onthe base of the wedge t depth y, and dividing this foree by ho Imus of the wedge sbove to dain te equivalent average atral aeceleration prodcing the same free. Secauoothe acceleration are assumed constant along aay horizontal ple, tho average seiomle cuffctent mil be the sate for al wedges having Pht tse a thesame elevation, regardless ofthe slope ofthe Aneined boundary of the wedge, Thus, the average slam coefficient forall wedges with ip same babe elevcion may bedatormined by evaluating tte average vale for & welge ‘htenng tothe centerline of the embankment and having tvertcal Rowdy Cotneidng with the center line. Besnuse the abeas force on the ase for such nee te goueratod by the accelerations tn the mass above, a determination fc"teeatesage setsmte coefficient irom the shear force cting onthe bise Seceasaily leads tothe samme results a the averaging procedure previously deserted. “fh ealcalae the shear force onthe base of any wedge, ti int necessary to determine the dlstribution of shear strain and shear atress throughout the height of the embazkinent. Prom Eqs © and 13, the dlplacement of th ex Senkmeat at any doth, 5, below the est in expressed by 2 23,(3,F yy Tmo) a, o).- 2) VQ eee 28) ‘SEISMIC COEFFICIENT a ‘The distribution of shear strain is obtained by differentiating this expression 29) | and thus the distribution of shear stress, r(y,t) is expressed by : 2 2G4,(6,2) ron=o Bon FER va... 00 yy ey FG Vn! » onc, th shear free acting on he bas potential siding Wedge extend- ing to depth y below the crest and with base length b(y) becomes: ie FQ) © r,t) 9”. en) ‘The average lateral accolerstion, tg tay, which would have to act on the ‘wedge to produce this same force, is detevinined by the relationship F Gt) = (mass of wedge) ti, () i, qy= (FEROy) UW, ©, 02) _ EWA) Ts, 4 04, EM ee 09) ‘and substituting from Eqs. $0 and 31 leads to 405,(65) 8 Oye 8 4%. lak) met &¥ 8, 5,@) v,e) = (98) OSes because oy = Aa/h (vg) and ve = VG7B. Alternatively, expressing Bq, 94 in terms of average seismic coefficients leads to sid re @ 4v. P wf tte AD net &Y By 4,8, ot given Vues of the volocty of sue” wavepropagaloh thcouh the em Tanicnot states vy the damping facor, Nan the helt of enbanknent, Ie values of KOyy a fern tantante of Gee ean sony bo cease ot er ering peo eerie nererrere enim tt Ecclestone induced by aneasingute. Die veows eather be puted seat, fie to show tho tine neary of fig sverage ston cooiiontfor te ae ee ee eee ‘mass extending foany given depth is independentot thease with ofthe wedge KW gy = Bi, Ogy vat) + (85) 8 May, 1968 sus ‘To ilustrate tho results of such anaveraging procedure, computations were rade of the time histories of average seleme coefficients for the same con- ditions as those shown in Fig. 10,1.e., a 100-ft-high embankment subjected to the N-S component of the Bl Centro earthquake of 1940. Average seismic co- ‘etficients were determined by euperimpocing the effect of the first eix modes of vibration for wedges located over tho top quarter, top half, top three- quarters, and the full height of the embaakment, ‘The results are plotted, for 8 80-sec periodof ground shaking, in Fig. 19. For purposes of comparison, i ‘may be noted that the ground motion induelng the computed seismic coefli- cients are shown in Fig. 2. TE may be seen that the average seismic coefficients vary considerably for different positions of the soil wedges considered, with higher values being de- ‘eloped with increasing height of the wedge within the embankment. Further- more,the magnitudes of the average accelerations represented by the seismic cociticientsare quite different from thoseot the ground accelerations inducing them, ‘Analyses of this typo can provide a valuable guide to the seismic forces, induced in earth embankments, Once these forces are known, the resulting deformations may he assessed, However, in many cases procedures for con sidering the complex variations of solsmi¢-force history (Fig. 18) have not yet been developed and it is necessary to approximate theforee eyeles of varying amplitudes by an equivalent number of force eyeles of constant amplitude, Th ‘effect, this involves an assessment of the significant number of force eycles land ah approximation of their maximum amplitudes and frequencies. Thus, tor example, for theplots shown in Fig, 18,the equivalent seismie-force var:~ ations might be considered to consist of fifteen significant cycles of force, with a predominant frequency of 8.9 cycles per second and with the following av~ erage amplitudes: Equivalent maxioum Extent of potential slide mass ‘seismic coefficient ‘Upper quarter of embankment 0.40 ‘Upper half of embankment 0.35 ‘Upper three quarters of ‘embankment 0.30 ‘Full height of embankment 0.25 Expressing the dynamic forces in this manner, by an Yequivalent setemic foree series,” can often provide the design engineer with a better picture of thelr potential effects on any given embankment, Material Properties Jor Computation of Seismic Response.—To make use of theone-dimensional response theory described above, t is necessary to be fable to evaluate the pertinent characteristics of the material involved, i. ‘the shear modulus, G, or the velocity of shear wave propogation, vg, and tie ‘damping factor, A,’ Although accurate data related to these characteristi2s are scarce, they do provide 2 gulde to the rangeof values likely to be encou- tered in practice. ‘The theory presented previously is based on the assumption of a uniform shear modulus inthe embankment, Embankments constructedot homogeneoas SEISMIC CORFFICIENT | mor iiier| Ren tr eon i = ey ea Pay mh Tae, cra “NN A vn IE Peer ar poe 1.000 PPS; 20% CRITICAL DAMPING FIG. 12.—VALUES OF AVERAGE SEISMIC COEFFICIENT FOR 100-FT-HIGH EMBANKMENT SUBJECTED TO EL. CENTRO EARTHQUAKE: 49 50 May, 1968 Ms ‘compacted earth fll consisting of a material that fs cohesive in nature (0 hydraotie fll alluvial material with a clay Dinder, etc.) can, a8 a first ap- proximation, be assumed to conform to the conditions postulated in this ap- roach, Information on the magnitude of the shear modulus or the velocity ct hear wave propagation in euch materials is primarily the result of seismic, ‘or forced vibration tests on existing dame. ‘Mononobe, Takata, and Matumura8 report seismic test results for four Japanese casth dams, hich ape deseribedas consisting of sandy clay and Toamny Jterial, Compressive wave velocities were messuredon the rest and slopes at the dam, computations indicating that, iy ell cases, the value of va. Was Somewhat less than 400 fps. Keightly18,16 conducted forced vibration ard Selemie tests on two earth dams, ‘The former tests result in series of ree~ ‘nant frequencies, and by vsing the “shear slice” theory for an elastic wedge in a rectangular canyon,10 a value of Ys can be compated from the observed fundamental frequency, ‘The first dam (Dry Canyon Dam) primarily consisted tha large contrat hydraulie fil core; the value of vp, as determined by boh Seismic and forced vibration tests, was about 303 fps. The second dam (ov- quet Canyon Dam) was constructed from compacted earth fil, for which the Value of tg was similarly determined tobe about 1,270 fps. The University of Californicrecentiy contucted similar tests on atomogensous compacted earth {ill dam (Bon TempeDam),in which the value of Ye was estimated as 1,180 fps. ‘The material in the latter wo dams consisted mainly of gravelly, sandy, siBy ‘ny, typical of @ Well-graved allovial material with some binder, both dam being constructed using modern compaction methods and control Hence, it would appear that shear wave velocities for homogeneous com pacted fll dams of the types deseribed, can vary over a wide range depending Daithe nature of the material, For hydraulic fi dams and fordams construc feof silty clays, the value of vp on te order of 300 fps seems appropriate, ‘whereas for dams constructed of alluvial matesial with some binder, value Of vp equal lo 1,000 fps appears to be of the right order ‘Sor larger dynamic strains, such as would occur in earth dams subjected to large earthquakes, the effective shear modulus would be less than that for | low amplitude elasti¢ vibcalions, because of the nonlinear hysteretic benavior of buch soils. Forexample, dnaimic tests conductedat the University of Calle fornia on samples trom the Bon Tetnpe Dam gave modult 50% less under such Sains. However, because vs = (G/p)1/2, the influence on the effective shear ‘wave velocity would be less. | “analysesof forced vibration teste oneasth dams have indicated that equy= alent vibeous damping factors on the order of 8% to 10% are operative durang qow amplitude elastic vibrations, However, dytamic laboratory tests on Lon ‘Tempe semples indicate that for dmamic straina such as wouldoccar sn eneth dams subjected tolange earthquakes, average values onthe order of 20% wosld Dermore appropriate. Influence ot Materia! Properties and Dam Heighton Seismic Response.~To | stiustrato the influence of material properties and dam height on the average | Seismic coetficlents, computations eimilas to that deseribed above to detor~ | Shine the average coctficients operative over different parts of a dara wore | tnade for three heights of dams (100 ft, 300 ft and 600 ft) and for materials | Fopresenting the extreme range of cohesive soils examined previously (shear | wave velocities equal to $00 fps and 1,000 fps), SEISMIC CORFFICIENT ] | | | | Biel | | L | Ve iT i H vei | | E 4 BALE HI 1 | i : uy AA F H Hl STH || | | i | Eg i et] | fe THY | | H | EPI | | PLU i ae Or OR 100-FT-HIGH EMBANKMENT SUBJECTED TO EL CENTRO ‘208 CRITICAL DAMPING FIG. 24.—VALUES OP AVERAGE SEISMIC COEFFICIENT i ' 51 wa May, 1968 sus Using Eq. 85 and superimposing the first six modes of vibration with the damping factor equi to 20% of critical damping for each mode, the time: Iistories of averagesoismnic coefficients were determined for grotind motions corresponding to the El Centro, 1940 (N.S) earthquake. Values of Klay were computed for potential sliding ‘wedges located over the lop quarter, © hal, top three-quarters, and the full height of each dam and the resuils ave plotted ‘over a lime duration of 30 sec for four of the six examples selected in Figs. 13 to 16, For eachof the computed sequences of seismic coefticient variations shown in Figs. 19 to 16 together with the two additional cages, an equivalent seismic force series wae assessed and the results are summarized in Table 1. For each embankment height and material characteristics, an assessment is shown of the number of significant force cycles, the predominant frequency of these force cycles, and the equlvaient maximim seismic coefficients operative on potential sliding masses over different parts of the embankment, Pestinenl re- sults from the table are shown graphically in Figs. 17 and 18, ‘TABLE 1. EQUIVALENT SEISMIC FORCE SERIES ASSESSMENTS ae] ess foe |rsie | besarte ee | ne eo e-(saer|_ amare | Beart |Ska | OP omre | ER | alet | atte | Rs a(la [TPS Paap psy eli fe fe fee fee fes pep san can oasTalent masimam seit couticint, Rey opentive over different sections of an embankment are plotted against height of embankment in Fig, 17(a) for embankment materials having 2 velocity of shear wave prope ‘gation vg = 200 fps and in Fig. 17¢b) for materials with ¥g = 1,000 fps. In all eases, the valuoof keq increases considerably with increasing elevation of the potential sliding mass within the embankment section, ‘This is a particularly ‘mportant result because it will often cause a marked change in the position of the most eritical siding surface. If the same seismic coefficiont for sliding ‘masses is used in 211 sections of the dam, the most critical surface will uous ally extend to thebase of the embankment; however, if higher values of the eos efficient aze used with increasing elevation of the potential sliding mass, the critical surface will ofton be well above the base. ‘Some confirmation of this type of behavior is provided by the performance ‘of the Otani Dam in Japan uring the 1946 Nankai earthquake.” ‘This 89-f1-igh Ms | dam was constructed with compacted shellsand a central lay core on a sand- | stone foundation. A eross section through the dam is shown in Fig. 18. As a SEISMIC CORFFICIENT 83 [ff result of the earthquake, longitudinal erack 250 ft long appeared on the crest "with additional cracks varying inlength between 80 ft and 180 ft developing on the higher regions of the upstream slope and running parallel to the axis. An incipient slide involving the upper third of the central part of the upstream "G. 18. VALUES OF AVERAGE SEISMIC CORPFICIENT TOR 900-F7-InGH EX- Eninhinnt SUBscCEed 6 Se ENR RANT MOON ee AE, BOY AE TNCAL DAMPING FIG. 16 “VALUES OF AVERAGE SEISMIC CORFPICIENT FOR 000-FT-10GH EM- BANKMENT SUBJECTED TO ELCENTRO EARTHQUAKE: , = 1,000 FDS, 20% ORI- ‘TICAL DANPING slope developed as indicated in Fig, 19, resulting in appreciable movements {in the upstream direction, Reference to Figs, 17a) and 17(b) also shows that, in general, the magni {ade of the maximum equivalent seismic evelficient, for potential sliding masses {in any given section of a dam, decreases withincréasing heightof the embank- ment. Thus, for example, for potential sliding masses extending through the full height of an embankment constructed of soil having & shear wave velocity 54 May, 1966 sM3 SEISMIC COEFFICIENT Gt 4000 fp, he maximum ease! selon costiciet is approsinatety es 0.25 for an embankment 150 ft high but ‘only about 0.1 for an embankment 450 ba, g i i 1 Exbaninenyencrac @ mrwins ye i i. =e peas tL aS / | as 7 Raman Pact Enteric 2 bad T }. 18.—RELATIONSHIP BETWEEN EQUIVALENT SIMPLIFIED i I SEta von Soars ahs ene BOUTIN SINTED I cceoah Guns cp oe ROMENTAL FEROD TOR 1 a a BARTHQUARE Moghet Enbortmer eet HIG. 17 VALUES OF EQUIVALENT MAXIMUM SEISMIC CORFFICIENT FOR HOMOGENEOUS EMBANKMENTS SUBJECTED 10 receiny EARTHQUAKE {Mah This result is of considerable importance in the design of high em= bankment dams. Comparison of the resulte in Figs. 17(a) and 17(0) shows that the material Propertiesalso have 2 large nfiuence on the selsmic forces induced is anne IG. 18, DAMAGE 70 OTANI DAM IN NANKAT BARTHQUAKE, 1946 56 May, 1968 a buniment. Yor ease of comparison, the va son, the values of equivalent maxi tole coefficient operative on potential slidi through the ets Hal sliding maseco extenang 4 shear nite sobmmauimcle are repeted In Tig ie or aE Se ites f 800 ps and 1,000. Apparently cone erage Shoe’ fet of even heigt witha mere feue ol barrage ee for ety gant) tases the devetopmen' of smaller seeie eres ea iO! & 00-1 damn contracted of materi ig nee eo only 0,08 com X may be conchided trom these figures that, in desigat See Gesiem purposes, is important to difenestae pee a cost salerent helgte and diferent material chaructertane eee a of eitions ofthe potential side mass wate te oe 1h this connection, it shoull be ns height and ‘uaterial Character sms SEISMIC CORFFICIENT 5 ‘and for matoriale exhibiting 20% critical damping, similar computations for " other ground motions and material characteristics could readily be made, Ut should be noted that the viscoelastic response theory used for the anal ses has certain limitations: 1, The method assumes that the response is controlled by shearing between horizontal slices and that the stress distribution along any horizontal surface is uniform, More recent analyses in which these assumptions need no! be made, show that the stress determined at aay olovation by the shear slice method is @ reasonable average of the streas distribution along a horizontal plane at that elevation, but that the actual stzesses are smaller than average ear the cuter slopes and higher than the average near the center part of the embankment. Thus, the values of average seismic force determined by the shear slice method are reasonable approximations for potential sliding masses extending inside the embankment to the center line, but are somewhat conser vative for potential sliding masses near the outer surfaces of the embankment, Preliminary studies indicate that the average values might be reduced by 25° ‘to 30% for potential sliding masses extending only about half-way towards the enter line of the embankment, Howover, further studies are required to confirm these preliminary conclusions, 2, ‘The methodonly considers the response of an embankment to horizontal ground shaking, although some allowance for the vertical components of a0 celeration can be made, if desired, by inclining the resulting forces on the po ‘ential sliding mass in 2 somewhat loss favorable direction than the horizontal, 3, The method does not consider the energy absorbed if the embankment Gevelops small plastic or inelastic deformalions during the earthquake, Bo- cause some plastic deformation ean be tolerated during a major earthquake, allowance for this effect should be made, Thie might be accomplished by using ‘| damping factor somewhat higher than that developed uring symmetrics cyclic stress or strain behavior of the construction material. A damping fac~ tor of 20% critical damping was used in the computations for this reason, ‘Thus, for these and other reasons, some judgment is necessary in select= ing designparametorstor any project. However, the type of results presented in Fig, 18 might be considered a useful basis for guiding this judgment, Fur~ thermore, it would seem reasonable to conclude that, in desigualing seismic coefficients for design purposes, \t 1s Important to distinguish between em- Dankments of different heights and material characteristics as well ae differ ent positions of the potential slide mags within the embankment section, ACKNOWLEDGMENTS ‘This roportis partof a study of embankment responseto earthquakes being conducted at the University of California, Berkeley, under the sponsorship of the Stateof California Department ot Water Reeources. ‘The support provided by the Department of Water Resources is gratefully acknowledges, A. K. Chopra and 1, M, Idries, members of the Soil Mechanics Research staff. pro~ vided valuable assistance in writing the computer programs. 5a May, 1966 APPENDIX.—NOTATION ‘The following symbols axe used in this paper: An = constant; wy = gota y) = hase length of potential sliding we Ff literal fore atumeg ee = shear modulus of embankment materia @ = acceleration of gravity; ne 1 = height of embsaitment T= inertia tore; 3 43g = Bessel function of first kind and orde; of first kind and order zero; sek | selec cetciat " xy = average seismic coelficient for potential slidin lay + erage sesmi cotictn for polenta ang mass seremental slice of potential sliding zones ‘n> number of mode; acreage! Sq = spectral acceleration; Sy = spectral velocity; T= period of vibratan inthe n€ modes tS me; : ¥ = displacement relative to base of {ig = eround displacement; ea 8 = ground velocity, Ug = ground accolertin; 'iy = absolute acceleration, Walthay = average absolute acccle wedge; afl 7 Sereue absolute acoteration on otra shiing wedge; Yq = velocity of shear wave propaga W = welght; eae A = vero value ofthe trequ: are eney equation Jot > 7 = unit weight of embankinent material, O°”? Ay = faction of critical damping in the n2 mode; 2 s mass dosity of embankment material; shear stress or time; fanetion of 37 and ys atural frequency of Wibralfon tn the nt modes an an = damped natural frequoney in he sl) modes SOIL MECHANICS AND FOUNDATIONS: May, 1966 Journal of the DIVISION: Proceedings of the American Society of Civil Engineers ———— BEHAVIOR OF A SOIL MASS UNDER DYNAMIC LOADING By Robert W. Sparrow! and A. Colin Tory? INTRODUCTION In recent years, inereased attention has been directed toward the develop ment of rational methods of flexible pavement design based on theoretical ‘Considerations of the distribution of stress and deftectionin idealized layered ‘ystoms (e. €.» Peattie,S Dorman 4ana Whiffen and Lister®). The multilayered. gystem forming a typical road construction is an assembly of material with complex physical properties subjected to complex loading patterns, and in ‘Seder to obtain such a theoretical solution, it is necessary to make a number of symplifying assumptions. ‘These assumptions are usually that the materials exhibitlinear, isotropic, ‘elastic, time-independent behavior, and that theloading system is quasi-static ‘and consists of a uniform normal surface contact stress over circular area. fn addition, the layers are assumed to be of infinite horizontal extent, and tho interfacial conditions are either perfectly rough or perfectly smooth. ‘Even within these limitations, solutions have only been obtained for stresses land strains on the axis of the load for a maximum of three layers.® For a Toie —biscussion ofen will October 1, 1565, To extend the closing dite one month, a written request must be ‘led with tne Fxeoutive Sooretary, ASCE. This paper 18 pert Sythe Sopyelgneod Joureal of tho Soil Mechanies and Foundations Division, Proceedirgs Gf the Ainerioan Society of Civil Enginoore, Vol. 82, No. SMS, May, 1968, Manuscript ‘as submitted for review for possible publication on September 23, 1965, i RSot, Duns of Research im Civ. Engrg, British Rallways Research Dept Derby, nglandy formeely, Lecturer, Dopt. of Civ. Rngrg., Univ. of Nottingham, Notinghars, England. ‘ostarer in Engrg, Univ. of Lelcester, Laloester, Bngland; formerly, Senior Research Asst, Univ, of Nottingham, Nottingham, England. AA. Fundementsl Approach to the Design of Flexible Pavements,” Tpeatiey rooecdingy ofthe intomational Conference on Structural Designof Asphalt Pavements, ‘Kon Arbor, Wigh., August, 1967, pp. Asma. "Lid, Dorman, GoM, “Tao Eyiension to Practice of @ Fundamental Procedure for to Design of Flexible Pavements." pp. 785-708, ‘Side, Wnitfin, A. Go aid Lister, N. Woy “The Appllostion of Elastic Theory to legible Pavements,” pp. 499-521. Troness hr “Tables of Stroneca ip Throo Layer Elastio Systems,” Bulletin No, 942, Highway Researoh Bd, Washington, D. C. 1962, pp. 176-214. 58

You might also like