You are on page 1of 52
Davin Moata View Seigsuic bES (GH Fou puclesé tL SEWER Olbwtr MIT Pram. Cotubridge , Ler FO AMEASURE OF EARTHQUAKE INTENSITY by Arturo Arias I, EMPIRICAL EARTHQUAKE INTENSITY SCALES Empirical scales are a first attempt to give a quantitative ex- pression to the notion of seismic intensity. All of these scales are based on the observed effects of earthquakes on human beings, on ani- mals, on certain objects, on the ground and landscape, and specially on structures made by man, Out of their own essence, empirical scales are descriptive and qualitative. They are not based on instrumental records, though the assignment of an intensity rating to an earthquake on a given empirical scale may, eventually, take advantage of the in- formation contained in instrumental records when these happen to exist. Starting with the ROSSI-FOREL scale (1883), and going through those proposed successively by MERCALLI (1902), SIEBERG (1923), WOOD and NEUMANN (1931), the several Japanese scales, and the recent MSK scale (1964) advanced by MEDVEDEV, SPONHEUER and KARNIK, a clear evolu- tionary trend may be detected:” 1. There has been an attempt to correct the limitations of some of the earlier versions, which grouped the strongest earthquakes in only a few degrees of the scale, through the addition of more degrees in the range of high intensities (some of the Japanese scales are an exception to this assertion); 2. Efforts have been made to insure that the estimates of inten- sity be objective and reproducible, through definitions each time more precise and univocal of the degrees of the intensity scales and of the elements that appear in such definitions. The use of questionnaires, the employment of observers with an appropriate We have mentioned only a few of the many proposed scales of seismic intensity. Charles Davison in a paper written in 1933 gives a list of 39 of them. See C, Davison, "Scales of Seismic Intensity: Supplemen- tary Paper's Bull. Seism. Soc. Amer., Vol. 23, pp. 158-166. ~438- -439- background, and the statistical analysis of tne collected data are some of the means used to this end; 3. There has been a tendency to assign more importance to the dam- age suffered by structures as indicative of earthquake severity; that is to say there has been a tendency to turn empirical scales into a measure of relative earthquake destructiveness. (This assertion is particularly applicable to the MSK scale). In spite of these efforts to improve the formulation and applica- tion of empirical scales, and recognizing their partial success, some criticism is still valid: 1, All empirical scales are qualitative, and consequently, do not establish a measure of earthquake intensity In fact, empirical scales are bound to give criteria for assian~ ing to a given earthquake at a given locality a certain place (or degree on the scale. Earthquakes may thus be ordered according to their as- signed intensities; but this ordering, although it may ne useful and usually is so, is essentially the result ot a convention. This is not a severe shortcoming, because, on ultimate analysis, all measurements on physical entities imply the use of one or more conventions. The sub- stance of our criticism 1s that earthquake intensity ratings on empici~ cal scales are not measures of intensity. Questions as the following “How much greater was the intensity of a given earthquake at locality A as compared with its intensity at locality 8?" cannot be given an intel- ligible answer; in fact the question itself has no sense at all, because the words “how mich" have no meaning when comparing degrees of any en~ pirical intensity scale. This objection appears outstandingly clear if someone tries to speak of “differences” of intensity: the jump froma particular degree of the scale to the one immediately following 15 not comparable, not even in a qualitative sense, with the jump between any other two consecutive degrees. 2. All empirical scales have subjective elements. Empirical scales do not only assign importance to the verifiable effects that may be observed after the earthquake, but take also into -440- consideration the effects on human beinys during the phenomenon it- self, at least for the evaluation of intensity at some of the degrees of the scale. The description of the latter effects 1s obviously strongly influenced by the psychological state of the observer. Subjective elements have less importance in what concems the observation of permanent effects of the earthquake. Nevertheless, even in this respect, subjectivity 18 not completely elimnated. De- scriptions of these effects depend on the possible interpretations that arise from the ambiguity or Jack of precision of the terms employed. These interpretations may change with time, or from one place to an- other, and of course they are biased by personal judgement. 3. Empirical scales are not reproducible, in the sense that they are not strictly univocal_and verifiable This 1s especially true regarding the effects of earthquakes on structures, these effects being some of the factors taken into account to define the higher intensities, and, incidentally, those which bear a higher importance from the standpoint of earthquake engineering. With respect to damage to buildings, :t should be remembered that the definitions of the types of structures used in earthquake intensity scales have been developed for structures built in a par- ticular country (or countries), Therefore, when the scale is used elsewhere, it is desirable to interpret the sense of each one of these definitions. Despite the efforts made to introduce greater precision, types of structures are described in relatively broad terms, so that structures of quite different behavior fall in one and the same type. Construction techniques, quality of workmanship, the extent and preciseness ot building inspection, and the quality of bu1lding mater- 1als vary from one place to another. These factors even differ among buildings im the same locality Architectural styles and fashions change and are different for different countries; new structural solu- tions are developed as design theories and provedures are improved and new materials,-or new uses for old materials, are introduced; hence -441- the epoch when the structure was designed and built has some influence on its earthquake response. The earthquake response of a structure depends on the parameters that describe it as a mechanical system; therefore, damage ratings at a given locality will be affected by the distribution of the pertinent structural parameters over the whole population of structures in that locality. Earthquake response depends also on soil-structure inter- action, and the relative importance of this interaction for a given structure depends on the type of structure; therefore damage at a place will depend on how important the soil-structure interaction is for all the structures at that place taken as a whole, that is to say, it will depend on what is the predominant type of structure considered in re- lation to the foundation soil. If structures that are sensitive to foundation settlement are relatively predominant at a certain locality, estimates of intensity at that locality will tend to be exaggerated, as compared to other placer with similar soil conditions where settlement-sensitive structures are relatively fewer in number Finally, a subjective element is at once introduced when the ex- tent of damage is qualified by using such terms as "important," “minor,” etc., without adequately defining when and in what sense to use them. The MSK scale is a good example of an intent to introduce some precision in this respect, by qualifying the use of the above said terms and, es- pecially, by classifying the structures more precisely. 4, Empirical scales are not fine enough, in the higher degrees of intensity, to discriminate between the severest earthquakes. Discrimination between very strong earthquakes 1s of paramount importance from the standpoint of earthquake engineering, so this lim~ tation of empirical scales 1s of consequence. Some efforts have beer exercised, in the more recent versions of intensity scales to overcome this shortconing, but we cannot as yet judge their success. 402 Empirical scales tend to give too much importance to the dam- age observed on structures that have poor earthquake-resistant propert: Wooden houses, brick masonry construction without reinforcement and other similar structures are given their proper place when judging the limits that separate the intermediate degrees of intensity scales, but no systematic consideration is given to damage induced by earth- quakes on structures specifically designed to withstand earthquakes. In fact, the MSK intensity scale expressly excludes structures designed to be earthquake-resistant in the rating of damage for a given earth- quake. Some of the above comments are quite obvious and may seem un- necessarily elaborate. It was felt that it was convenient to set them forth systematically as a motivation and justification of the need for a more scientific approach to the definition of the concept of earth- quake intensity. It is not implied, and it should not be construed, that empiri- cal scales are useless or that they should be abandoned. On the con- trary, these scales will continue to be useful for a long time to come, for several reasons. We enumerate some very briefly. Any definition of earthquake intensity based on instrumental records will be affected by one very important limitation; in many areas of high seismicity there is a shortage or even a complete lack of adequate instrumentation for the recording of strong ground motions. Empirical scales will continue to be useful for the comparison of the destructiveness of future earthquakes with those which occurred in the past. In those applications where an estimate of intensity at a great number of places is needed, as, for example, in the description of ef- fects of earthquakes over large areas (intensity maps), or in programs of seismic zoning or seismic regionalization, there is no possibility of substitution: empirical scales, as far as we can see now, are in- ~443- dispensable. The influence of soil conditions on the earthquake-res ist- ant behavior of structures can vary very sharply from one place to a nearby one, even in the same citys it would not be economical to install and operate such a large number of instruments as would be needed for detailed studies of local variations of intensity in such an instance. Finally, a tremendous amount of work has been spent in correlating in- tensity, as judged by empirical scales, and other pertinent variables (as magnitude, epicentral distance, etc.); the results of these corre- lations are very useful, and will be more so in the future, for the for- mulation of local and national design codes, or for decisions about design forces to be used in the case of specially important structures. IL, HOUSNER'S SPECTRAL INTENSITY. A first attempt to overcome the limitations of empirical scales and to establish a measure of earthquake intensity based on instrumen- tal records was set forth by OMORI (1900) and CANCANI (1904). These researchers tried to relate intensity with maximum ground acceleration, assfgning to each one of the degrees of intensity scales proposed by them a certain interval of maximum ground accelerations; thereby they were setting forth maximum ground acceleration as a measure of intensity. As a conventional measure of intensity this is unassailable; but it re- mains to be shown that earthquake damage is directly related to maximum ground acceleration. Later research work has disclosed that the damage rating of earthquakes cannot be described adequately, not to say com- pletely, by maximum ground acceleration alon tant, as total duration of the earthquake, amplitude of motion, espec- ially of the wave of longer period 1n the strongest phase of the phenom- other factors are impor- enon, etc. GALITZIN, in 1913, discussed the use of parallelopipeds of uniform cross section but of varying heights for measuring horizontal accelera- tions in severe earthquakes, and thereby establishing what he called a “rational dynamic scale" of intensity In a certain sense, the idea of Galitzin was an anticipation of the spectral intensity introduced later by HOUSNER, -444- WESTERGAARD, 1n 1933, 1n a paper entitled "Measuring Earthquake Intensity in Pounds per Square Foot" (Engineering News Record, Vol. 110), suggested the idea of taking the kinetic energy density of around mo- tion (kinetic energy per unit volume) as a measure of earthquake inten- sity. The important point in Westergaard's paper is that, in general, the higher earthquake accelerations are associated with the short period waves, and that it may be reasonable to consider the accelerations (in each component wave) to be inversely proportional to the period. Then the maximum velocity becones constant for all periods. Inasmuch as the kinetic energy is proportional to the ground velocity squared, the maximum velocity would also be a measure of ground kinetic energy. The units of kinetic energy per unit volume are ft. Ib. per cu. ft. or Ib. per sq. ft., hence the title of Westergaard's paper. Even though vague and imprecise, as the language employed by Westergaard in this short paper is, the idea behind his proposal is a very sound one, and it can be exploited to give a very interesting definition of earthquake intensity In 1934, BENIOFF proposed to define seismic destructiveness as "the integral with respect to pendulum frequency of the maximum dis- placement of an infinite series of undamped pendulums extending over a significant range of frequencies” (Ref. 1, p. 399). In the language that is now customary in earthquake engineering, Bentoff is here pro- posing to define sersmic destructiveness by the integral 1 {S44 a) “o where Sy is the maximum relative displacement of an undamped one- degree-of-freedom oscillator of circular frequency ., and the integral is taken over a significant range of frequencies (u,, «1)- Benioff went so far as to consider the response of complex structures by introducing what 1s now known as the equivalent moda? -445- oscillators. He says: "Complex structures having more than one mode of vibration may be considered as made up of a group of pendulums with the frequencies of the individual modes," and on page 399: "...suppose that we substitute for the engineering structures a series of undamped pendulum seismometers having frequencies ranging from the lowest fun- damental frequency of engineering structures to the highest signifi- cant overtones." He tries then to justify the election of zero damp- ing for the pendulums, and to give some theoretical arguments in sup- port of the proposed definition of seismic destructiveness. Following this idea of Benioff, G. W. HOUSNER, in 1952, intro- duced a measure of earthquake intensity based on the pseudo-velocity spectrum of a linear oscillator with viscous damping. Housner defined spectral intensity as the area enclosed by the pseudo-velocity spec- trum curve, considered as a function of the undamped natural period, T, of the oscillator, the T-axis, and two extreme ordinates that corre- spond to T= 0.1 sec. and T = 2.5 sec., respectively. (See Fig. 1). Housner's spectral intensity is then given by the integral 2.5 sec. J Sy(Tandar (2) 0.1 sec. where Sy(Tn) ‘is the pseudo-velocity spectrum for a damping ratio equal to n, expressed as a function of the undamped natural period, T. Spectral intensity has, therefore, the physical dimensions of a length. Let us remark, that Housner's spectral intensity would measure the intensity of an earthquake at a given point in a given direction; i.e., at the point the accelerogram was recorded and in the direction of the recorded ground accelerations. It is interesting to quote how Housner arrives at the defini- tion of spectral intensity (Ref. 2): -446- “The basic problem of engineering seismoloay concerns the design of structures to resist earthauakes and there are three significant aspects of this problem. First, there is the problem of designing a structure so that all of its parts have equal strengths to resist the stresses produced by earthquakes; that is, there should be a uniform factor of safety against failure. Second, there is the problem of designing different structures to have the same factors of safety; that is, to insure that a tall buildina will have the same degree of strength as a low building, that a flexible building has the same degree of resistance against earth- quakes as a rigid building, etc. Third, there is the problem of determining the factor of safety, or the required strength to re- sist earthquakes, so that structures will be able to withstand the most severe ground motion to which they are likely to be subjected, without experiencing serious damage." “The spectrum analysis of strong-motion earthquake records is pertinent to the solution of all three of the aforementioned prob- lems...the third of the above problems, namely, what is the most severe ground motion to which structures are likely to be subjected requires a method of measuring the intensity of the ground mo- tion that is related to the maximum stresses produced in structures, which means that the measure of intensity should be such that if it is said that the intensity of one earthquake is twice the intensity ! of another it implies that the maximum stresses produced by the first earthquake are twice those produced by the second. There is Now no measure of intensity which has this property so that the spectrum intensity defined in this report, which does have this property, is very meaningful for engineering purposes."(p. 4). Now, if a linearly elastic one-degree-of-freedom system, with viscous damping, like the one shown schematically in Fig. 2, is cho- sen as a mechanical model of a structure, the maximum lateral force experienced by it is given by Frax = KSq (3) where Sy is the ordinate of the maximum relative displacement response -447- Spectrum corresponding to the undamped natural period of vibration of the structure. For small values of the damping ratio, Ea. (3) can be written as Fan k max" % = fi Sy (4) where Sy is the ordinate of the pseudo-velocity spectrum. "The spectrum curve [Sy]" is thus a measure of the intensity of the earthquake” in the sense that a given structure subjected to different earthquakes will experience maximum stresses directly proportional to the values of Sy for the different shocks..... In an actual city there will be a large number of structures with periods of vibration ranging from 0.1 seconds to perhaps 2.5 sec- onds or more and this range of periods must be taken into account when defining the intensity of the shock. If a city were com- posed of structures with periods of vibration ranging uniformly from 0.1 to 2.5 seconds, then the average value of the spectrum between 0.1 and 2.5 seconds would be a measure of the intensity of an earthquake in the sense that on the average the maximum stresses experienced by the structures would be in the ratios of the average values of the spectra for different earthauakes. Since the inten- sity is @ comparative measure for expressing the relative severity of earthquakes, the area under the [pseudo-velocity] spectrum curve can be used instead of the average ordinate." (pp. 5-6). Accordingly the definition of Eq. (2) is then proposed by Housner. the spectrum intensity..., is a function of the fraction of critical damping, n, and it will be necessary to specify whether the undamped spectruin intensity is being referred to, the 0.2 We include explanatory interpolations in brackets. Notation and Eq. numbers have been changed. “This is along the lines of a proposal made by Hl. Renioff, "The flyst- cal Evaluation of Seismic Destructiveness,” Bull, Seis. Soc. Aner. (1934) Vol. 24, pp. 398-403 (Housner's footnote. -448- damped intensity, etc. This must be done since 1t 1s not possible to separate the effect of damping from the effect of ground motion; that is, it is not possible to write Ea. (4) 1n the form Fax 7 Yuk t(n)S (5) max where f(n) is independent of the base acceleration and S 1s the un- damped spectrum."(p. 6) "Since all buildings have dampina, and ordinary buildings not specially designed to resist lateral torces can be expected to have relatively large amounts ot damping, the spectrum intensities for 0.2 damping are perhaps more indicative of the destructiveness of an earthquake than are the undamped intensities." (p. 32 In one 0, his conclusions Housner states: “For each earthauake the spectrum intensities corresponding to several values of damping should be computed with the intensity for 0.2 critical damping being a reasonable measure for ordinary struc- tures." (p. 58) IIT. COMMENTS ON HOUSNER’S SPECTRAL INTENSITY. The measure of earthquake intensity proposed by Housner 15 un- doubtedly a very significant step forward. Nevertheless, an examination of the hypothesis on which it rests and some comments on its limitations seem to be pertinent. 1. Housner's justification of spectral intensity 15 based on the consideration of the maximum stresses induced in structures by the earthquake, averaged over a collection of structures whose period are uniformly distributed in tne interva) 0? to Z.5 secemls On the other hand, the criteria implicit in all empirical scales 1s te rate the severity of an earthauake by 1ts destructiveness This fs pointed out by Housner when he says that: -449- "It should be noted that the Modified Mercalli Intensity doe: not actually measure the same thing as the spectrum intensity. The latter is a measure of the maximum stresses produced in undamaged buildings, whereas the former is a measure of the actual damage incurred." (p. 32). 2, It is not clear why the pseudo-velocity spectrum Sy is chosen to define spectral intensity instead of some other spectrum, as, for exam- ple, the maximum relative displacement response spectrum, Sy S pro- posed by Benioff), or the maximum absolute acceleration response spec trum, S,. The justification given by Housner fs that the maximum stresses in the elastic element of the one-degree-of-freedom mechani - cal model are proportional to S,. But we also have Fax 7k Sy =m Sq (6) That is to say, the maximum stresses are proportional to Sy and also to S,- So the areas under these spectra could have been taken as a measure of intensity on the same grounds that the pseudo-velocity spec- trum was selected. 3. The hypothesis that the natural periods of the structures in a given city are uniformly distributed over the interval 0.1 to 2.5 seconds could perhaps be justified on account of its simplicity. But it would be as simple to suppose that the natural frequencies are uni- formly distributed in some range of frequencies, as Benioff does in his definition of seismic destructiveness. If one takes into account that for a very large class of mechani- cal systems with a large number of degrees of freedom, the frequencies of the higher modes are very approximately evenly spaced, it seems that there are reasons to expect that in cities with tall burldings the natural frequencies are more likely to be evenly distributed, when due consideration is given to all significant modes of vibration, and con- sequently the shorter pertods are relatively more abundant. If there are few tall buildings, or none at all, this reasoning 1s still move pertinent. This means that it would be better to integrate the spec- ~450- trum over the frequencies if what is desired is a measure of the mean value of the maximum stresses to be expected over a population of structures. 4, The damping ratio has a very pronounced influence on the ordi- nates of the response spectra, especially for small amounts of damping. Therefore the choice of some particular damping ratio is a must in Housner's definition, From a practical point of view, perhaps this is not a very serious objection, so long as we compare only spectral in- tensities computed for one and the same dampina ratio. 5, The selection of the limits of integration deserves some comment. The lower limit (7 = 0.1 sec.) is practically dictated by the limita- tions of recording instruments. Fortunately the value of Sy dT (7) 0 would be very small because it can be easily shown that Sy +Oas T +0. This might be one of the reasons that led Housner to choose the pseudo-velocity spectrum. fy sec. A different situation arises at the upper limit of integration. Many earthquakes have pseudo-velocity spectra that are quite flat at T = 2.5 sec., so, that choice of that value of the period as the upper limit of integration is rather critical: a displacement of the limit into the range of longer periods would increase the estimates of the spectral intensities substantially; conversely, a displacement of the upper limit into the range of periods shorter than 2.5 secs. (let us say, choosing it equal to 2.0 sec.) would decrease the values of the spectral intensities substantially. The variations in spectral intensities that arise by changing the upper limit of the integral would have little importance only if the ground acceleration contains a small proportion of low frequency com- porents. 6. Finally, there is an important limitation in the concept of spec- tral intensity. As was pointed out before, spectral intensity 1s 4 ~ 451- measure of intensity in a given direction. Strong motion accelero- graphs will usually record the ground accelerations along three orthog- onal directions (one vertical and two horizontal components), so spec- trum intensities could be computed for those three directions. But there is no way of combining the three numbers thus obtained to give a number that has a physical meaning. For example, given the spectral intensities for the two horizontal components, it is not possible to get from them the spectral intensity of the earthquake alona a third horizontal direction. The average of the spectral intensities along two orthogonal hori- zontal directions may only be given a conventional meaning, as Housner does when comparing spectral intensities with Modifved-Mercall1 inten- sities, for example. All these difficulties arise from tne undesirable analytical proper- ties of maximum response spectra, that do not lend themselves easily to analytical manipulation. In spite of all these adverse comments, the basic 1dea introduced by Benioff and Housner, namely the evaluation of earthquake intensity (or seismic destructiveness, as Benioff prefers to say) through an integral that bears a direct relation to response spectra, represents a substantial improvement, and, as we shall see, points in a direction that may be con- ducive to a more satisfactory definition IV. A_MEASURE OF EARTHQUAKE INTENSITY BASED ON CUMULATIVE DAMAGE. We shall try to give a definition of earthquake intensity which, being simple enough, is on the one hand as little as possible subject to the above-raised objections, and on the other hand takes into account the intent implicit in all empirical scales, namely tnat intensity be some- how related to seismic destructiveness. We shall accept that the ultimate purpose of intensity scale is to give an indication of the potential damage that an earthquake can ~452- produce in the locality at which the intensity is being ascertained, independently of whether in that locality there are or there are no structures, and independently of the quality and nature of the struc- tures that might exist. We shall then accept that a reasonable defi- nition of a measure of intensity should comply with the following re- quirements : A. It should be a measure of seismic destructiveness; B. It should make as little hypothesis as possible on the nature and dynamical properties of structures; . It should contain a minimum of arbitrary elements; D. It should be simple. At first sight, it seems quite an impossible task to frame a defi- nition cf intensity that will meet all of these requirements. As we shall see, it is at the least possible to satisfy most of them reasonably well. We shall assume that the amount of damage experienced by a struc- ture is proportional to the energy dissipated by the structure per unit weight during the overall duration of the motion induced on it by the earthquake. Strictly speaking, only a part of the dissipated energy is reflected as deterioration of the structure, permanent set and changes in its mechanical properties. A fraction of the dissipated energy is converted into heat without associated damage. To compute the amount of energy that goes into the production of damage it is necessary to choose some mathematical model for the struc- tures considered. Because of the reasons just stated above, we should necessarily choose a non-linear model, with the consequent analytical complications. We shall return to this point later. For the moment, let us point out that by assuming that the energy dissipated per unit of weight is a measure of the extent of the damage experienced by the structure, we are thereby eliminating one parameter, viz. the mass of the structure; so the distribution of the structure in a given collec- tion of structures, as far as their masses are concerned, is by the same ~453- token eliminated, and we are by the way putting light and massive struc- tures on the same footing. This seems reasonable, as what we are really ‘interested in is not the absolute value of the damage, but the extent to which it affects the structure. Let w be the undamped circular frequency for small oscillations of a one-degree-of-freedom structure, Let us consider a population or col- lection of structures whose frequencies are uniformly distributed in the interval (0, +=); more precisely, let us assume that the ratio of the number of structures whose frequencies lie in the interval (w, utau) to the total number of structures in the population does not depend on w and is proportional to au. Let us further assume that the structures in this population are initially at rest in their equilibrium configurations and that they are so orientated that they can perform oscillations only in one given direction, the same one for all of them, Let E be the energy dissipated per unit of weight by a structure of frequency w as a consequence of the motion induced on it by the earth- quake. We shall define the intensity, I, of the earthquake along the above-mentioned direction as the sum of the energies dissipated (per unit weight) by all the structures belonging to the population. Therefore we may write le {ee (8) According to this definition, intensity has the dimenstons of a velocity. The computation of I requires the evaluation ot £, and this depends on the mechanical model chosen to represent the structures in the popula- tion. As a first approximation to the problem of defining a measure of intensity, we shall adopt as a model for the structures belonging to the population a simple linear oscillator with viscous damping. With this choice, the restriction on amplitudes is not necessary for the defini- tion of w. From the standpoint of analytical simplicity the model selec- -454- ted is highly convenient; vt allows us to carry theoretical analyse much farther, and 1n a relatively elementary level; we shall see that the final results assume a very simple and meaningful form, V. INTENSITY INA GIVEN DIRECTION THROUGH A POLNT We shall for the present assume simple linear osci\tators with viscous damping as appropriate models of the structures belonging to the population considered. The frequencies ot these oscillators, ac- cording to the hypothesis already made, shall be unitormly distributed in (0,*). For the time being we shal} admit that all the oscillators have the same damping ratio n. The intensity Ty (n) at 0 and in the direction of the Ox-axis will be a function of n. It can be shown, after @ lengthy calculation, that fo -arccosn | 2 Lyn) = aa Ja%nret (9) g ‘I-n ° where I (n) = intensity at 0 along the x-ax1s n= damping ratio of osci! lator: a,(t) + ground acceleration along x-axis at instant t t, = total duration of eartnquake = acceleration due to gravity. Let us point out that what was not possible tor Housner's spectral intensity turns out to be possrble with the new definition: the intensity 18 expressed as a function fin) - $2 ces (10) ohn? that does not depend on ground accelerations, multiplied by functional of the yround accelerations -455- Furthermore, this function (n) varies very slowly with n, in the range of damping ratios interesting for earthauake engineering. For those values of n that are to be expected in real structures, f(n) is practically a constant, as the following table shows. Table 1 n are cos n 0 F = 1.5708 0.02 1.5511 0.05 1.5227 0.10 1.4780 0.20 1.3977 0.30 1.3272 0.40 1.2649 0.50 1.2092 1.00 1.0000 Therefore the value of n chosen for the oscillators that represent the structures does not materially affect the value of intensity derived in Eq. (9). This conclusion has a very important consequence. By assuming that all the structures belonging to the population had the same damping ratio we were allowing an element of arbitrariness to creep into our def- inition. It is now evident that this has no practical consequence on the values computed for I,,(n), as long as the damping ratios of real struc~ tures are small. For values of n in the range n = 0 ton = 0.2 the total oscillation of f(n) does not exceed :7% of the mean of its values at the two extreme points of the range, and we can thus choose n = 0 as the standard value of n, without departing too much from the values of 1,(n) that are obtained for moderate damping ratios. Let us then standardize our measure of intensity by setting t “2 xx) * i ay (td (1) -456- Our definition has thereby become independent of the distribution of the structures belonging to the population, as far as their damping ratios are concemed. Let us remark, by the way, that 1,,(0) 1s defined as the limit of I,,(n) as n+ 0, This limt is ditferent from zero and has a simple physical meaning: 1t is the sum of the total energies per unit weight stored in the oscillators of a population of undamped linear oscillators uniformly distributed as to their frequencies, at the moment the earth- quake ends (or for that matter, at any instant after the end of ground motion). VI, INTENSITY TENSOR AT A POINT. If a(t), ay(t), a(t) are the components of ground acceleration along three mutually perpendicular axes Oxyz, the acceleration along any straight line through 0, whose direction cosines are (2,u,+) will be given by a(t) = ra,(t) + nayit) + a(t) (12) and therefore the intensity 1(, ||) along the straight Tine will be 2 2 2 Tas)” x lt ry Lyt v Ut arul yy? 2uv Lt 2 ly (13) here ; a 'o { Lyx * 2g J a,( tat Ty © yx * Oy { ayltiay(tidt | ° oO Z 1 \ ty + ty i ror J 2 Ale 4) wit J Sire tye fata tne | hy ty i . [2 Lat ; Lt al ae (t)dt Lace a(t), (tidt oO ° ! -457- We can then construct a symmetrical tensor , whose components re- ferred to the Oxyz axes are the nine quantities defined above. We shall call it the intensity tensor at 0, Its matrix referred to the Oxyz axes is | 9 ite ly Le CATs bly Wy Tye (1s) ly Ly daz The trace of this matrix (or first invariant of the tensor) is in- variant for a rotation of the coordinate axes that leaves the origin fixed, This invariant we shall denote by 9, and give it the name of scalar intensity at 0. Clearly we have i ty : -t 2 ate aide = ot | wo Iie 1yt lyt Lz * Oy [keds dae eg Sat (16) ° ° Let us choose 0z to coincide with the vertical through 0. If we Now rotate the reference frame around the z-axis, the sum 1,,+ I, does not change; that is to say, it is the same for all pairs of mutually perpendicular horizontal axes through 0. We shall call this sum the intensity on the horizontal plane through 0, and will denote it by J,. Therefore to Tn * Text lyy 2 oe las + a2)dt (Oz, vertical) (17) he xx * yy © 2g jx * 4y , ° It follows immediately that 1, 1s an upper bound for both I, and Iyys and for all intensities along horizontal straight lines through 0. This upper bound is attained only when the ground motion at 0 1s polar- ized on a vertical plane. It is obvious that I, will be specially important because man- made structures are usually more sensitive to horizontal wotions of their foundations than they are to vertical motions. -458- Applying well known results on quadratic forms, the eigenvectors and eigenvalues of the tensor § may be determined; that 1s, we may determine the principal directions of the intensity at 0 and the princt- pal values of the intensity at 0, much in the same way as the principal stress planes and the principal stresses at a point are found in problems of elasticity or strength of materials. To determine & it suffices to have records of ground accelera~ tions at 0 along three mutually perpendicular directions. These data are provided by standard three-component strong motion accelerographs. It is pertinent to remark here that for the computation of the crossed components I,,y ly,» 1, it 1s important that the records for the x-, y-, and z-axes be referred to the Same origin of time. The same does not apply to J; nor to 1, because in the computation of these scalars only squares of the accelerations are used, so a small shift in the time origin for different components of the acceleration has practically no effect upon the end result VII, RELATIONS BETWEEN {INTENSITY AND RESPONSE SPECTRA. The intensity along a given direction may be related to the re~ sponse spectra of simple oscillators so orientated that they can only respond to base motion in the given direction. There are two kinds of spectra that bear a strict relation to intensity as defined 1nP5: these are the maximum residual motion response spectra and the spectrum of dissipated energy. (A third spectrum that bears a direct relation to intensity will be discussed in #7). Let us call E(u,n) the eneray dissipated per unit werght by a viscously damped simple linear oscillator of damping ratio n and un- damped circular frequency «. Then E(u yn) = E (18) where E has the same meaning as in Eq. (8) where the basic definition of intensity was introduced, Therefore the special choice we have made ~459- for the model of the population structures simply means that intensity is equal to the area under the spectrum of dissipated eneray, when the ordinates of this spectrum are plotted as a function of the undamped circular frequency. If we standardize the definition of intensity by letting n + 0, the eneray dissipated by each oscillator is precisely equal to the energy stored in the oscillator when the earthquake ends, so I,,(0) may be in- terpreted as the area under a spectrum of the total energy absorbed by an undamped (simple harmonic) oscillator when plotted as a function of circular frequency. If we denote by R, the displacement amplitude of the residual mo- tion, and by V, the amplitude of the velocity of the residual motion, we have 1k Re 12 Re = 1,2 = Bag eo ag Re * By x (19) Therefore, ening (menzt2 aire vo) aI, oPatdas # 3g | Vlad 4 (20) So the standardized intensity 1,,(0) is proportional to the area under the curve whose ordinates are o f jw), and also to the area under the curve whose ordinates are V5 bw) » both functions plotted with « as inde~ pendent variable. We might call R,(w) and V,(w) the spectra for residual maximum displacement and for residual maximum velocity, respectively. Turning to the more commonly used spectra, it 1s easy to show that the following inequalities hold @ 2 So (oso) Il) « 2b fs Pefloroldy « ptf Slerolde = xt f Mae (21) ° ° ° ~460- where Sy,(us0)+ Sy,(a,0) and S,,(u40) stand respectively for the maximum relative displacement response spectrum, the pseudo-velocity spectrum, and the maxinum absolute acceleration response spectrum, in the Ox direc- tion. VIII, RELATION BETWEEN [NTENSITY THE FOURLER SPECTRUM OF GROUND MOTION By Parseval's Relation we have Fata . xf a(t )dt (22) ° where Faw) = | a,(2)e"'*" (23) is the Fourier transform of a,(t). Therefore T(0) = gb 1, (a) Pau (24) and = arc cos f I, (n) = Sr 695-0 (28) gy 1-n We see then that there 1s a close relation between intensity and the area under the square of the modulus of the Fourier spectrum of ground accel- erations plotted as a function of w Recalling that ’ y a 24 {IF (od lau = { IFy (6) )a2du a [Fo (w)Iur de (27) ok o ‘x o §x we can also relate intensity to the Fourier spectra of ground velocity, Yg2 and of ground displacement, s, ~ABI- IX. SOME NUMERICAL RESULTS AND CORRELATIONS. Intensities for the components of earthauakes recorded at 27 sites have been computed by G. LANGE (Ref. 3). The total number of components analyzed by Lange is 59. The main results are given in Table 2 at the end of this paper. Intensities range from 1.93 m/sec. for the strongest (N.S.) component of the May 18, 1940 El Centro earthquake, down to 0.016 m/sec. for the weakest component recorded at Los Angeles Subway Terminal (Oct. 2, 1933). The intensities on the horizontal plane range from 3.25 m/sec. for the aformentioned E1 Centro earthquake, down to 0.037 m/sec. at Los Angeles Subway Terminal. IXA, Correlation between Intensity and Maximum Response Spectra. Response spectra were available for 53 components out of the 59 for which the intensity was computed. A correlation was established be- tween the values of I(o) and those of the integral 62 ‘$5 (w,0) = ‘a i 2 z | a tg [Sirover é for the abovesaid 53 components. See Fig. 3. The coefficient of correlation was found to be r= 0.953 which is highly significant. The equations of the lines of regression are 1 = 0.572 + [ser (0.536 0.607] and 1L (20 | ag fs = 1.671 (1.56, 1,77] The confidence intervals for the coefficients of rearession are (for a 95% level of confidence) given in brackets at the right of each equa- tion. -462- 1X8. Correlation between Intensity on the Horizontal Piane and Maximum Response Spectra Though the sum 1 J 2 1 j 2 st t+ s? at rg ox Tra } ay has no physical meaning, a correlation between this sum and the intensity on a horizontal plane was established. See Fig. 4. The total number of points was 26 (at one site one of the hortzon- tal components was not recorded). The coefficient of correlation was found to be = 0.978 and the equations for the regression line 1 2 2 = a! (Sa,7 Sgy)@T = 1.77 Ty (1 65, 1.90] ° I, = 0,547 + (0.518, 0.586) IXC. Correlation between Intensity and Housner's Spectral Ju1 The total number of components considered was 30. Housne:'s in- tensities are in ft. and correspond ton + 0. See Fig. 5. The results obtained are as follows: r= 0.951 Bs 42.01 (37.9, 46.0) Iy in ft., I in m/sec -463- IXD, Correlation between Intensity on the Horizontal Plane and M.M.I. The Modified-Mercalli Intensity (MMI) is a psychophysical variable, so a linear regression of MMI on log,y Jy, was tried. See Fig. 6. The total number of points was 15. The results are as follows: yr = 0.573 MMI 7.25 + 0.89 logig 1, I, in m/sec., MMI in units of this empirical intensity scale. The confidence interval (95% level) for the free term is [6.71,7,80], and for the slope [0.13, 1.65] It is seen that, although the correlation is significant, the con- fidence limits for the regression line are very wide. It may be pointed out that for Housner's spectral intensity the coefficient of correlation with MMI, for the same data, turns out to be poorer (but not much so): r = 0.50 , and the regression line is: MMI = 6.23 + 1.41 lomo Ty IXE, Correlation between Intensity and Maximum Ground Acceleration. The total number of points considered was 53. The coefficient of correlation between ora )® and I was found to be (see Fig. 7). r= 0,734 and the regression line = 0.107 1 (0.087, 0.127] ~464- X, DISCUSSION AND CONCLUSIONS . We are now 1n a position to discuss the results obtained and to draw some conclusions. Let us examine how far we have succeeded in satisfying the require- ments we had put forward as desiderata for an ideal measure of intensity. A. We stated as our first requirement that intensity should be a measure of earthquake destructiveness. There is no doubt that damage 1s related to the energy dissipated by the structure, though we may be sure that the relation 1s a complica- ted one. Part of the energy is dissipated as heat; the rest remains "trapped" in the structure and is reflected as permanent deformation and permanent changes in the mechanical properties (change in stiffness, for example, and more generally, changes 1n the stress-strain relatvonship). As long as we do not choose a particular mechanical model to repre- sent the structures, the defimtton given in @4, Eq. (8) is not liable to criticism in this respect. But, of course, 1f we do not choose a par- ticular model we cannot give a numerical value to intensity. So we are bound to choose some model. The viscously damped one-degree-of-freedom oscillator is perhaps the simplest one that can be chosen, although this choice 1s quite open to criticism. First of all, the total amount of the energy dissipated at the dash- pot of the linear model 1s transformed into heat, so no fraction of it goes into producing a deterioration of the structure. Secondly, linear models are a poor representation of what happens to real structures ex- posed to very strong earthquakes Let us examine these objections more closely. The first objection is a very strong one, because it goes at the root af the matter: remember we are trying to establish a measure of intensity on a cumulative damage basis. In a viscously damped oscillator ~465- there is dissipation of energy but there is no cumulative damage. This objection, on the other hand, does not affect Housner's definition, be- cause he does not intend to base it on a criterion of accumulated dam- age. From an empirical point of view, if we take Modified-Mercalli In- tensities as a point of reference (and we must recognize they are a very poor point of reference), the measure of intensity proposed here correlates with MMI better than spectral intensities, though the im- provement is not very significant: r = 0.573 as compared with r = 0.50. What really counts is whether the amount of energy dissipated by viscously damped oscillators can be taken as indicative of that part of the energy dissipated by real structures which goes into the production of damage. There is some ground to believe that this might be so. Let us suppose that one-degree-of-freedom systems with non-linear time-invariant characteristics are representative of the behavior of real structures. There is no doubt that this kind of system constitutes amore realistic representation, Now, non-Vinearity may arise in the differential equation of motion either in the velocity terms or in the restoring force terms. If only the velocity terms are non-linear, there is some justifice- tion for representing energy dissipation by an equivalent viscous damp- ing (see Refs. 4, 5). Sut one of the main results we have obtained is that the measure of intensity proposed in the present paper turns out to be practically independent of the damping ratio. This means that the sum of the energies dissipated per unit weight by all the oscillators belonging to the ensemble considered is practically independent of the damping ratio for each one of them and, therefore, is also independent of the distribution of the structures as to damping ratios. It follows then that if non-linearities arise only 1 the velocity terms, and an equivalent viscous damping can be defined, the total amount of energy dissipated will be practically independent of the form assumed by such ~466- non-linearities, as long as the damping ratios necessary to reproduce the non-linear effects is not very large. So, in this case, viscously damped linear oscillators seem to be acceptable models. If the restoring force terms are involved 1n the non-linearity, an equivalent viscous damping may not be very meaningful. This depends on the form assumed by this kind of non-linearity. If there are no hys- teretic effects, the situation is not very critical. MURPHY and BYCROFT (Ref. 6) have found that the energy input into a non-linear oscillator, where the non-linearity arises only in the restoring terms, is practical- ly independent of the degree of non-linearity. This would mean that the total energy input into a collection of these oscillators would be sub- stantially the same as for a collection of simple harmonic oscillators with no damping, having their natural frequencies distributed in the same way as the frequencies for small oscillations of the non-linear oscillators. The conclusion 1s that, if there are no hysteretic effects, the measure of intensity herein proposed will not be very sensitive to non-linearities arising in the restoring force terms. The above discussion shows that the critical situation seems to lie in the hysteretic effects. To evaluate the influence of these hys- teretic effects, an elastoplastic model was chosen with a yield accel- eration (ratio of yield load to mass) of 0.1 g. The 22 horizontal com- ponents of 11 earthquake accelerograms were used aS input and the intensity was computed through the equation Te, * Edu where E was taken as the energy dissipated by hysteresis per unit weicht, and w is the circular frequency for oscillations in the elastic range (Ref, 7). Intensities computed in this way were compared with those ob- tained from a viscously damped linear model. A good correlation was ob- tained, The results are as follows 0.98 (N = 22) -467- A parabola was fitted to the points. The equation of this regression Vine was found to be : 2 . T= 0.028 + 1.10 1 5 #0.16 Ip 5 With such a good correlation, there 1s some ground to assume that the values of I obtained using a viscously damped linear model are cor- related with structural damage. A bi-linear hysteretic model could have been used, but taking into account the results obtained by THOMALDES (doctoral Thesis, University of Michiaan’ that show that the energy absorption characteristics of bi-linear andelastoplastic models are not substantially different, it was not deemed necessary to consider the bilinear case. We should caution the reader not to give the above results on the correlation of I with Ip p more value than what 1s here stated. Results might be different if sone other model of hysteretic behavior 15 chosen, or if changes in the slope of the elastic part of the elasto-plastic load-deflection characteristic are considered to take place as the struc- ture deteriorates by successive incursions in the plastic range XB_and XC. As our second and third requirements which an idea) defini- tion of a measure of earthquake intensity should meet, we stated that it should assume as little as possible of the nature and properties of structures and that it should contain a minimum of arbitrary elements The assumption that the structures can be modeled as viscously damped one-degree-of-freedom oscillators looks very strong and appears at first sight to impose severe limitations. But atter we have shown that the amount of damping assumed 1s practically immaterial, and when due consideration is given to the results obtained by MURPHY and BYCROFT (that the energy input 1s not substantyally affected by non-linearities in the restoring terms for elastic structures), by THOMAIDES (that the energy input is similar for bi-linear and for elasto-plastic structures", and by ARNOLD (that the energy dissipated by a population of elasto- ~468- plastic structures has a very close correlation with the intensity calcu- lated by supposing a linear model), one comes to the conclusion that the selection of the model 1s after all not so critical for the evaluation of intensities. Ke are aware that some elements of arbitrariness still remain. But they seem to us to have been reduced to a bare minimum XD, The final result attained 1s a very simple one (Eq. 9 ), so the fourth requirement has been fully achieved. And on achieving it we have been able to define some very interesting concepts After having analyzed the shortcomings of the proposed definition on intensity, we think 1t appropriate to summarize its relative advantages: 1, It 18 simple; 2. It relates directly to clearly defined physical concepts; 3. None of the objections applicable to empirical scales apply to it; 4, The final expression for earthquake intensity relates 1t directly to the ground acceleration; that is to say, directly to the raw data supplied by accelerographs; 5. It has a sense to combine intensities at a given point in differ- ent directions, defining an intensity tensor at that poin 6. Once the accelerograms for three mutually perpendicular direc- tions through a point are known, we can compute the intensity in any direction through that point; 7. A-scalar intensity at a point and the intensity on a horizontal plane can be defined; 8. Intensities are practically independent of the damping ratio of structures, and (though this deserves further research) seem not to depend very critically on assumptions concerning the struc- tural model adopted; 9. Intensity 1s directly related to the Fourier spectrum of ground motion; a property that opens the possibility of interesting theoretical developments; and 10. Intensities (especially intensities on a horizontal plane) cor- relate very nicely with som integrals related with response spectra 4. -469- REFERENCES. H. BENIOFF, “The Physical Evaluation of Seismic Destructiveness," Bull, Seism, Soc. Aner., (1934) Vol. 24, pp. 398-403, G. W, HOUSNER, "Intensity of Ground Motion Qucing Strong Earth- quake," Calif. Inst. of Technology, Earthquake Res. Lab., Aug. 1952. G. LANGE, "Una medida de intensidad sismica," Thesis, University of Chile, 1968, L. S. JACOBSEN, "Steady Forced Vibrations as Influenced by Damping," Transactions Am. Soc, Mech. Eng., 1930. L. S. JACOBSEN, "Frictional Effects in Composite Structures Sub- jected to Earthquake Vibrations," Stanford University, March 1959 M, J. MURPHY and G, N. BYCROFT, "The Response of a Non-linear Oscilla- tor to an Earthquake," Bull. Seism- Soc Amer., Yol 46 (1956) pp. 56 - 65. P. ARNOLD, Thesis, University of Chile, (1969). Unpublished.

You might also like