Вы находитесь на странице: 1из 101
DEPARTMENT OF CIVIL ENGINEERING MASSACHUSETTS INSTITUTE OF TECHNOLOGY SIMQKE ‘A PROGRAM FOR ARTIFICIAL MOTION GENERATION USERS'S MANUAL AND DOCUMENTATION A Program Distributed by NISEE/Computer Applications November 1976 Massachusetts Institute of Technology Department of Civil Engineering Constructed Facilities Division Cambridge, Massachusetts 02139 Evaluation of Seismic Safety of Buildings Report No. 2 SIMULATED EARTHQUAKE MOTIONS COMPATIBLE WITH PRESCRIBED RESPONSE SPECTRA by Dario A. Gasparini and Erik H. Vanmarcke January, 1976 Sponsored by National Science Foundation Division of Advanced Environmental Research and Technology Grant ATA 74-06935 Publication No. R76-4 Order No. 527 ABSTRACT Alternate methodologies for generation of simulated earthquakes are briefly reviewed. The method of superposition of sine waves is discussed in detail. Theoretical relationships existing between the ground acceleration spectral density function and the response spectrum are derived. The program SIMQKE, which can generate response spectrum - compatible artificial motions, is listed and explained. Properties of resultant simulated motions are presented and their use in seismic design is discussed. PREFACE This is one of a series of reports prepared under Grant ATA 74-06935 from the National Science Foundation on the “Evaluation of Seismic Safety of Buildings." The purpose of this project is to evaluate the effective- ness of the total seismic design process, which consists of steps beginning with seismic risk analysis through dynamic analysis and the design of struc- tural components. The project seeks to answer the question: "Given a set of procedures for these steps, what is the actual degree of protection against earthquake damage provided?" Alternative methods of analysis and design are being considered. Specifically, these alternatives are built around three methods of dynamic analysis: (1) time-history analysis, (2) response spectrum modal analysis, and (3) random vibration analysis. The formal reports produced thus far are: 1. Arnold, Peter, Vanmarcke, Erik H., and Gazetas, George, "Frequency Content of Ground Motions during the 1971 San Fernando Earthquake," M.I.T. Department of Civil Engineering Research Report R76-3, Order No. 526, January 1976. 2. Gasparini, Dario A. and Vanmarcke, Erik H., "Simulated Earthquake Motions Compatible with Prescribed Response Spectra," M.I.T. Depart- ment of Civil Engineering Research Report R76-4, Order No. 527, January 1976. 3. Vanmarcke, Erik H., Biggs, J.M., Frank, Robert, Gasparini, Dario, Gazetas, George, Arnold, Peter, and Luyties, William, "Comparison of Seismic Analysis Procedures for Elastic Multi-degree Systems," M.I.T. Department of Civil Engineering Research Report R76-5, Order No. 528, January 1976. 4. Frank, Robert, Anagnostopoulos, Stavros, Biggs, J.M., and Vanmarcke, Erik H., “Variability of Inelastic Structural Response Due to Real and Artificial Ground Motions," M.I.T. Department of Civil Engineering Research Report R76-6, Order No. 529, January 1976. These reports are based upon a series of Internal Study Reports which were prepared during the course of the work. Those prepared to date are: 1. Arnold, Peter, "The Influence of Site Azimuth and Local Soil Condi- tions on Earthquake Ground Motion Spectra," January 1975. 10. . 12. Arnold, Peter, "Structural Response to Earthquake Ground Motions by Random Vibrations," January 1975. Gasparini, Dario, "SIMQKE" A Program for Artificial Motion Genera- tion," January 1975. Frank, Robert and Gasparini, Dario, “Evaluation of Seismic Safety of Buildings: Progress Report on Statistical Studies of Responses of MDOF Systems to Real and Artificial Ground Motion,” January 1975. Gazetas, George, "Study of the Effect of Local Soil Conditions on ‘the San Fernando Earthquake Response Spectra, Using Regression Analysis," January 1975. Frank, Robert, "A Statistical Study on the Nonlinear Response of MDOF Systems to Real and Artificial Earthquake Ground Motions,” August 1975. Frank, Robert, "A Study of the ‘Effective Period’ Change for SDOF Systems Responding Inelastically to a Sinusoidal Base Motion,” August 1975. Frank, Robert, “A Statistical Study of the Response Spectra of ‘the 39 Real Earthquake Records," August 1975. Gazetas, George and Vanmarcke, Erik H., "Evaluation of Random Vibration Analysis of Elastic MDOF Systems," August 1975. Gasparini, Dario, "Some Parametric Studies Using the Earthquake Simulation Program SIMQKE," August 1975. Frank, Robert, "A Study on the Effect of the Distribution of Stiffness on the Response, Elastic and Inelastic, of MDOF Systems to Real and Artificial Earthquake Motions," September 1975. Frank, Robert, "A Study on the Effect of the Motion Intensity on the Nonlinear Response of MDOF Systems," September 1975. The project is supervised by Professors John M. Biggs and Erik H. Vanmarcke of the Civil Engineering Department. They have been assisted by Dr. Stavros Anagnostopoulos, a’ Research Associate in the Department. The research assistants on the project have been Peter Arnold, George Gazetas, Dario Gasparini, Robert Frank, William Luyties, and Richard Haviland. 5. TABLE OF CONTENTS Title Page Abstract Preface Table of Contents Introduction 1. Random Vibration Based - Prediction of Response Spectra 1.1 Moments of Spectral Density Functions 1.2. Stationary Response Variance 1.3. Transient Response Variance 1.4 Prediction of Maximum Response 2. The Program SIMQKE and Properties of Simulated Earthquakes 2.1 Artificial Motion Generation Procedure 2.2 Generation of Power Spectral Density Functions From Smooth Target Spectra 2.3. Improving Response Spectral Matches 2.4 Properties of Generated Motions Accelerograms and Peak Accelerations Response Spectra Spectral Density Functions 3. Parametric Studies Using the Program SINQKE 3.1 Effect of Smoothing Cycles 3.2 Effect of Duration on G(w) and Sy 3.3. Variation of G(w) Derived From a Typical Set of Smooth Design Spectra ‘Summary and Conclusions References List of Figures and Tables Page 1 2 3 5 3B SRR S a 4a 43 47 58 61 64 INTRODUCTION This report documents the algorithms and capabilities of the earthquake simulation program SIMQKE. As an introduction, motivations for earthquake simulation and alternate methodologies are briefly ex- plored. Theoretical developments which are the basis of the algorithms used by SIMQKE are presented next. Finally, the properties of the gener- ated earthquakes and parametric studies utilizing the program are reported. The program itself and a brief user's manual are appended. The limitations and paucity of recorded accelerograms together with the widespread use of time-history dynamic analysis for obtaining structural and secondary systems' response are the primary motivations for the develop- ment of simulation capabilities. Individual real earthquake records are limited in the sense that they are conditional on a single realization of a set of random parameters (magnitude, focal depth, attenuation character- istics, frequency content, duration, etc.), a realization that will likely never occur again and that may not be satisfactory for design purposes. For example, Jennings, Housner and Tsai (19) noted that a significant gap in the present day collection of accelerograms is that the shaking in the vicinity of the causative fault in a truly great (Richter Magnitude 8) earthquake has never been recorded. Their objective was, then, to pre- dict and simulate such anearthquake. Similarly, Hou's (15) methodology had such parameters as magnitude and focal distance as input. His clearly stated objective was to predict likely motions for particular sites. From the above needs and Housner's recognition (16) of an earth- quake as a transient stochastic process, numerous mathematical models have been proposed and are used. References (15) and (19) give brief histories of simulation techniques. The intent herein is to focus on one commonly used method of numerical simulation, the one based on the fact that any periodic function can be expanded into a series of sinusoidal waves: n x(t) = TA; sin(w,t + 6;) a art ‘ A, is the amplitude and 4, is the phase angle of the ith contributing sinusoid. By fixing an array of amplitudes and then generating different arrays of phase angles, different motions which are similar in general appearance (i.e., in frequency content) but different in the "details", can be generated. The computer uses a "random number generator" subroutine to produce strings of phase angles with a uniform distribution in the range between 0 and 2r. The total power of the steady state motion, x(t) is J (A%/2). Assume now that the frequencies u, in Equation 1 are chosen to Iie at equal intervals du, Figure 1 shows @ function G(w) whose value at w, is equel to A;/2hw so that G(w,)Aw =A;/2. Allowing the nunber of sinusoids in the motion to become very large, the total power will become equal to the area under the continuous curve G(w), which is in effect the spectral density function. Formal definitions of G(w) can be found in many textbooks (Cran- dal] and Mark, 1963; Y.K. Lin, 1970). G(w) expresses the relative importance (i.e., the relative contribution to the total power) of sinusoids with fre- quencies within some specified band of frequencies. When G(w) is narrowly centered around a single frequency, then Equation 1 will generate nearly sinusoidal functions as shown in Figure 2a. On the other hand, if the spectral density function is nearly constant over a wide band of frequencies, components with widely different frequencies will compete to contribute equally to the motion intensity, and the resulting motions will resemble portions of earthquake records, as illustrated in Figure 2b. Of course, the total power and the relative frequency content of the motions produced by using Equation 1 do not vary with time. To similate in part the transient character of real earthquakes, the stationary motions, x(t), are usually multiplied by a determinstic intensity function such as the "boxcar", trape- zoidal (Hou, 1968), exponential (Liu, 1969), and compound (Jennings, 1968) functions shown in Figure 3. Bycroft (1960) and Brady (1966) simulated “white noise" (for which G(w) is theoretically constant for all frequencies to represent earthquake ground motion. Actually, the simplest workable form of G(w) is that corres- ponding to a band-limited white noise. The spectral density is constant in the frequency range from 0 to w,, as shown in Figure 4a: = 6 Osu uy, G(w) (2) Based on Kanai's study (1961) of the frequency content of a limited number of recorded strong ground motions, Tajimi (1960) suggested the following widely quoted form for the spectral density function of ground motion (Fig. 4b): T1+4r? (wfu,)*I6, (3) 11 (w/ag)?F + 422 (w/a,)” G(w) = Soviet researchers (Barstein, 1960; Bolotin, 1969) suggested a probabilistic approach to earthquake engineering based on the autocorrelation function of Glw) FIGURE 1 - SPECTRAL DENSITY FUNCTION G@) | & x(t) o7 t + wo ¢a) X(#) Gla) Seas t < + con cb) FIGURE 2 - RANDOM PROCESSES WITH CORRESPONDING SPECTRAL DENSITY FUNCTIONS: a) Narrow Band Process; b) Wide Band Process -10- nates Ee)$€ =@at™ t cod
FIGURE 3 - INTENSITY FUNCTIONS USED TO SIMULATE THE TRANSIENT CHARACTER OF REAL EARTHQUAKES: a G Cw) 6&) i aay 6 6 3 + © w 4 cad Cb) ° FIGURE 4 a) BAND-LIMITED WHITE NOISE S.D.F. b) KANAI-TAJIMI S.D.F. “ue a form corresponding to the spectral density function of Eq. 3. Sample functions x(t) with spectral densities corresponding to Eq. 3 can be obtained by filtering "ideal white noise" (i.e., with w, == in Eq. 2) through a simple oscillator with natural frequency wy and viscous damping ¢g (Franklin, 19653 Penzien and Liu, 1969). These may be interpreted as the “predominant ground frequency" and the "ground damping", respectively. The values u,=4r and ¢,=0.60 have bben suggested for firm ground sites. 6, is a measure of ground intensity. Extensions of Eq. 3 have been proposed to model ground motions whose spectral density shows one dominant spectral peak (Liu and Jhaveri, 1969). Theoretical spectral density shapes such as those described by Eqs. 2 and 3 are obtained by examining, smoothing and/or averaging of the squared Fourier amplitudes |f(w)|2 of actual strong motion records. This stems from the basic fact that, for stationary random pro- cesses, the expected value of |f(w)|7 and the spectral density functions G(w) are proportional (Jenkins, 1961; Rosenblueth, 1964). Ground motion models which account for the time-varying nature of the relative frequency content have also been proposed, (e.9., S.C. Liu, 19703 H. Kameda, 1975). In reference to Eq. 1, it suffices to allow the amplitudes A; to vary slowly with time. The spectral content can then be described by an “evolutionary" spectral density function G(w,t), i.e., CH ,t) is propor- tional to A? (28). These more sophisticated models are particularly use- ful in describing the frequency content of structural response (37), or of motions at the point of support of secondary systems attached to struc- tures. However, time-invariant models in which G(w) reflects the frequency content during the most intense part of the ground motion are believed to be sufficiently accurate for the purpose of seismic response prediction, for all but certain nonlinear systems. “1e- In summary, the problem in simulation with the model represented by Eq. 1 is, then, one of selecting a proper shape and intensity of either the power spectral density function or the Fourier amplitude spectrum and estimating the duration of the motion. Geophysical models which predict. these parameters of an earthquake at a specific site are therefore ideally suited for use in conjunction with earthquake simulation and random vibra tion analysis (37). Indeed, Berril1 (2) has recently formulated such a model using the 1971 San Fernando earthquake as his data base. The method uses a simple two-parameter source model to estimate the source excitation strength in terms of Fourier amplitudes and then uses an amplitude decay expression and scatter statistics to obtain expected Fourier amplitudes of ground acceleration at a site. For the San Fernando earthquake, agree- ment was found between model prediction and independent observation of |f(w)| to within a factor of two or three (2). With such geophysical pre- diction, simulated earthquakes can indeed represent possible intense motion at a site and could logically be specified for design purposes. Development of design criteria has evolved along different paths, however. In the nuclear industry, for example, a set of smooth response spectra proposed by Newmark, Blume and Kapur (26) have been adopted for use in aseismic design. These design spectra can be completely defined by a peak ground acceleration and are intended to be applicable to all firm ground sites. These developments and the need for time histories have spawned recent efforts in synthesizing earthquakes whose response spectra “match” or are compatible with specified smooth response spectra. Hadjian (13), for example, has proposed utilizing real motions and -13- performing three types of modifications. The first is linear scaling of the entire motion to raise or lower the entire response spectrum; the second is modifying the digitization interval to shift the location of spectral peaks and valleys; the third is utilizing a 1-DOF filter to decrease the Power of the motion at desired frequencies. Similarly, Tsai (34) has proposed beginning with a motion whose res- Ponse spectrum closely resembles the design spectrum (at the appropriate ing damping value) and then utilizing spectrum suppressing and spectrum rai: techniques. Lowering of the spectrum is achieved by passing the selected motion through a set of frequency suppressing filters, (2-DOF mechanical systems). Raising the spectrum is achieved by superposing on the selected ground motion sinusoidal components of suitable amplitudes, frequencies and phase angles. A frequency domain analogue of the above procedure has been proposed and used by Rizzo, D.E. Shaw and S.J. Jarecki (29). A discrete Fourier transform of a real motion is first generated; it is then appropriately modified and, finally, an inverse Fourier transform is computed; i.e., a new time history is synthesized. The original Fourier transform is modified by a multiplicative function when it is necessary to lower the response spectrum while an additive constant is used if the computed response spec- trum is to be raised. Alternatively, Scanlan and Sachs (31) first generate an artificial motion by sinusoidal superposition using amplitudes A, derived from the values of the zero damping target response spectrum. It has been shown that the latter is related to the Fourier amplitude spectrum of the motion (17). Subsequent iterations to provide better spectral matches use a -14- linear correction applied to the amplitudes, as follows: ( Sy) ' v target Ay wre SV) computed These methodologies do not clarify the relationship between the response spectral values for arbitrary damping and the Fourier amplitudes or the values of the spectral density function of the ground motion. This relationship has been clarified by Vanmarcke (35), (37) utilizing random vibration theory. Since this work is the basis of the procedures used in the earthquake simulation program SIMQKE, it will be presented in detail in the following section. -15- 1. RANDOM VIBRATION-BASED PREDICTION OF RESPONSE SPECTRA This section presents a random vibration analysis of linear one- degree systems with different natural frequencies uw, (from 0 to =) and either undamped or with small viscous damping ratios c. Moments of spectral density functions are discussed first, stationary response statistics are then derived, and the required modifications to deal with the transient character of the excitation and the response are discussed subsequently. In the solution developed in this section, the response spectral value y,.,, i.e., the level below which the absolute value of the response y(t) will remain, with probability p, during a strong-motion duration s, is expressed as a multiple of o,(s), the standard deviation of the one-degree system response (for given w, and z) evaluated at s, as follows: Yssp * "ssp y's) ea The problem of determining the peak factor r... is formidable, requiring 33 the solution of the so-called first-passage problem. Exact solutions to this problem do not yet exist, but good, practical approximate solutions are now available. When p=0.5, Equation 4 yields the median response spectra, While it is not uncommon to find the factor r..4 5 treated as @ constant, it may actually lie anywhere between 1.25 and 3.50 for typical ground motions, as will be seen in this section. The methodology presented herein can be used for an earthquake response analysis of rather general linear systems. In this section, however, complete results are derived for one-degree systems only. -16- 1.1 Moments of Spectral Density Functions In random vibration analysis of seismic response, an earthquake can be represented as a limited duration segment of a stationary random function with a given spectral density function G(w). This representation is at once powerful and simple. It allows the application of some important re- sults of stationary probabilistic dynamic analysis, while the specification of a limited duration and a sudden start captures the essential transient character of earthquake ground motion. The integral over frequency of G(w) equals the average total power, or the variance o” for motions which fluc~ tuate about a zero mean value, e.g., ground acceleration and linear system response. For the band-limited noise motion (Eq. 2), the variance i | G(w) dw = Gu, (5) and for the Kanai-Tajimi spectrum (Eq. 3), the variance is: 76, a, a = | Glu) dw = 4 (1 +406) (6) 9 a Actually, a more useful way for dealing with the frequency content of ground motions is through the normalized spectral density function: 6*(u) = 4 ew) (7) or its cumulative spectral distribution: ® F*(u) = | G*(u) dw (8) 0 which increases from 0 to 1 as w goes from 0 to @. Note the analogy between normalized s.d.f. and the probability density function (p.d.f.) “7 of any random variable: both are nonnegative and have unit area. The moments of the spectral density function G(w) are: = [. Ga) dw = o% | w'6(w) dw = oat (9) ° ° in which Ay is the ith moment of the unit area spectral density. It is clear that o?=2, and 4*=1. A measure of where the spectral mass is con- centrated along the frequency axis is (see Fig. 1): a= fT = Ae (10) which is analogous to the root mean square (r.m.s.) of a random variable. A convenient measure of the spread or the dispersion of the s.d.f. about. its center frequency is (Vanmarcke, 1969; 1972): 8° fi, > TAPAS qa) which is dimensionless, always lies between 0 and 1, and increases with increasing bandwidth. Pursuing the analogy between &*(w) and the probability density function of a random variable, 6 is equivalent to the ratio of the standard deviation to the root mean square value. It is clear that 6 will be large if G(w) has two or more fairly widely separated peaks. Important time domain interpretations of the spectral parameters 2 and 6 are discussed ‘in reference (36). In the frequency domain, these two parameters provide a summary description of G*(w). In fact, it is possible to develop (Cheby- chev-type) bounds or F*(w) in terms of @ and & (Vanmarcke, 1972). In evaluating higher spectral moments, problems due to lack of con- vergence are sometimes encountered. These are due to the fact that various Proposed algebraic expressions (e.g., Eq. 3), while providing a good fit to -18- computed power spectra in the central frequency region, do not properly represent the ground motion in the upper frequency tail. Owing to record- ing and processing limitations, strong motion accelerograms provide little information about motion frequency content beyond a circular frequency y= 2n/AT, where AT is approximately 0.02 seconds. Consequently, it is difficult to evaluate contributions to, say, the second spectral moment, due to frequencies beyond w,; one is, in effect, restricted to computing partial moments of G(w), in a limited frequency range (0,w,). The values of the spectral parameters corresponding to these partial s.d.f.'s given in Equations 2 and 3 are listed in Table 1. Spectral shape parameters can also be computed from the squared Fourier amplitudes |f(w)|* of actual strong motion records. (Sixsmith and Roesset, 1970). It suffices to substitute Ir(w)|? ror G(w) in the definitions, Eqs. 10 and 11. Some results are pre- sented in Table 2. 1.2. Stationary Response Variance A basic result of stationary random vibration of linear systems is the following relationship between the spectral density functions (s.d.f.) of input and output (see, for example, Crandall and Mark, 1963): - 2 Gy(w) = Glu) |H(w)| a2) ‘in which Gy (w) = the output s.d.f., G(w) = the input s.d.f., and |H(w)| = the amplification function of the linear system, i.e., the amplitude of the steady-state response of the system to a sinusoidal input with unit amplitude and frequency w. Also, the variance of the response, o?, is equal to the area under 6,(w): -19- Q 6 (Eq. 8) (Eq. 9) Band-Limited White Ji ye Noise Spectral Density ©, = 0.58 w, oe Kanai Tajimi Spectral Density for w < Wy = 4, ~ 2.1, 0.67 and ¢,=0.6 9 2 TABLE 1 - APPROXIMATE VALUES FOR PARAMETERS OF RELATIVE FREQUENCY CONTENT a é (rad/sec) El Centro 1940 NS 31.35 0.73 El Centro 1940 EW 25.51 0.64 Olympia N 10 W 36.07 0.65 Olympia N 80 E 30.85 0.62 Taft N69 W 27n 0.66 Taft $ 210 27.46 0.64 TABLE 2 - SPECTRAL PARAMETERS COMPUTED FROM SQUARED FOURIER AMPLITUDES JF(w) |? OF EARTHQUAKES (SIXSMITH AND ROESSET, 1970) -20- w= | orde= | ot JH(w) |? dw (13) ° ° For a linear one-degree system with natural frequency w, and damping ratio %s whose input and output are support acceleration and relative displacenent response, respectively, the squared amplification function is: -1 IW(o) |? = Clas = 07)? + 42%0Zu7] (1a) The relative displacement response variance o% is obtained by inserting Equation 14 into Equation 13. Also the standard deviation of the pseudo- acceleration response is o, = whoy. Note that, at the extremes of the frequency scale, this formulation leads to the following desirable, not conmonly recognized, results. First, a4 approaches the standard deviation of the ground acceleration (c) when wo, * © (and H(w) + 1): oy * whey + C | Glu) du} =o i) : At the other extreme, when w, > 0 and |H(w)| >"> (see Eq. 14), we obtain: (16) 2 = a 2 oy | wo” Glu) dw r. displ. 0 The integrand w""G(w) is actually the spectral density function of the round displacement”, and therefore o? is equal to the variance of the Ground displacement when w, = 0. This follows from Eq. 12 and from the fact that the amplification function of a hypothetical system whose input is the second derivative of the out~ put, equals w*. Of course, in this section, both ground acceleration and ground displacement are assumed to be stationary random processes. -21- As shown in Fig. 5, the earthquake excitation spectral density func- tion often varies relatively smoothly in the immediate vicinity of the system's natural frequency uw,» while |H(u)|* exhibits a sharp peak at Oy This effect is more pronounced as the system damping decreases. It Teads to the following useful approximation for the pseudo-acceleration ji 2 = who2: response variance, 03 = whos: a 4, of J G(u) [H(w) |? dw = wt6(u,) | [H(u)|? dw -u,6(u,) + | G(w) dw Q a 4 a7) 4 fn 03 = Gw,w, Ce - 1) + fern ° Figure 5 illustrates the meaning of the two terms on the right side of Equation 17. The first term accounts for the contribution in a narrow frequency range around uw,» the second the contribution in frequency range below (0,w,). The term ,G(u,) is subtracted because the area it represents would otherwise be counted twice. The relative importance of the second term in Equation 17, the partial area under G(w), increases for higher natural frequencies. For lightly damped systems with intermediate natural frequencies, the first term will strongly dominate. Equation 17 correctly predicts o; when the excitation is an ideal white noise, i.e., G(u)=G,, for all w. We have then: o? = Gu, [(n/4e) - 1] + 6,u, = (¥6,0,)/4¢ as) A widely used approximation for the variance of the response to wide band excitation is obtained by substituting G, in Equation 18 by G(w,). It is evident that this result is of little use in predicting responses at higher frequencies. (Yo *0) 39nva AoNanbaad ¥ WOU (4 Yo NnOdY 39NVE MOWUYN ¥ Wows (@ +3SNOdS3Y JO JONVIYVA OL SNOILNGTUINOD - ¢ aunDTS ca) (0) 22+ ken | Yeo 2H | for -23- It is convenient to express the response variance o2 in terms of the ground motion variance o*, i.e., to evaluate the ratio:(*) o Fee ge tS lo)a, (qe = 1) + Ma) (19) where 6*(,) = the unit-area ground motion s.d.f., and F*(w,) = the normalized cumulative spectrum. It is not important that the expression for F*(w,) in Equation 19 be accurate in the range where the first term predominates. But F*(w) should approach 1 at high values of wy» when the contribution due to the first term vanishes. 1.3. Transient Response Variance The steady-state value given by Eq. 17 may never be closely approached when the one-degree system's natural frequency or damping are very small, i.e., when the product tw, is small. As was first shown by Caughey and Stumpf (1961), the response variance will build up from zero (at the tine when the earthquake strikes) to a maximum value, near the end of the (equi- valent stationary) motion duration, s. The frequency content of the one-degree system response will evolve in a way which can perhaps most conveniently be described by the time-dependent spectral density function Gy(w,t) (Corotis and Vanmarcke, 1975). For a broad class of functions G,(wst), the time- dependent variance of the transient response can be obtained by integration over all frequencies: of (t) = | Gy (ust) dw (20) a The function G,(wst) will depend on the input s.d.f. G(w) and on the system For light damping, the factor [(/4z) - 1] may be replaced by 1/4z. -24- properties. For any linear system with impulse response function h(t), it is possible to define the truncated Fourier transformation, or the "time- dependent transfer function", as follows: t H(w,t) = J h(t~2) eae (21) ° which converges to the transfer function H(w) when t +=. When the systen ‘is suddenly exposed to a steady excitation with spectral density function G(w), the time-dependent response s.d.f. will be given by G,(wst) = Glu) |H(w,t) |? (22) In the case at hand, G(w) is wide-band and smoothly and slowly varying and ‘the system is a simple oscillator with impulse response function evcent v ° sinw, t t ej4 h(t) (23) 0 t<0 where w, = w,/T=62 = the damped natural frequency of the system. The transient squared amplification function |H(w,t)|* has the following form (Caughey and Stumpf, 1961; Hammond, 1968; Corotis et al., 1972): ot [H(w,t)[2 = [H(w) |? {i = 22% C(cosw, t+ sin, t) coswt wy 1,28 (+ Ge sina, t cose, t w + . -uygt 7 sinu, t sinwt - eM (o,6)? = a, + u? + 2u? sin?u, 2} (24) -25- Integrating this expression over all frequencies yields approximately | Intwst) |? aw « 2 =e ey (25) : i The foregoing integral increases from 0 to the stationary value wW(4eur), which will be achieved when t >> 1/¢u,. Comparison between Eq. 25 and the stationary value motivates the definition of a fictitious time-dependent damping: Se chee c So that the right side of Eq. 25 can be written as m/4c,u2. OF course, 57S when t >. Actually, the parameter c, is particularly useful in that it allows the entire set of spectral shapes |H(w,t)|? (Eq. 24) to be crudely approximated by: -1 [(w,t)]? = C(wz ~ 07)? + acZu2u?] (27) which has the familiar form of the squared amplification function |H(w) |? given by Eq. 14. The damping parameter ¢, decays from a very high value down to the actual system damping c: the rate of decay is governed by the Product cu,. The approximate form yields, for all values of t, not only about the same total area, but also the same central frequency (w,) and about the same "bandwidth" as the exact form (7). The main advantage re- sulting from the use of Eq. 27 is that all the stationary results obtained ‘in the previous section can now be applied to the transient response situa- tion, simply by substituting & by ry. Note that the use of Eqs. 26 and 27 also provides a convenient way for treating the response of the undamped system, for which the stationary con- -26- dition is, of course, never closely approached. In this case - 1 Sey emt aE (28) A result of particular importance is the pseudo-acceleration response variance evaluated at the end of the motion duration, s: o(s) = wpoy(s) = | G(w) |H(w,s) |? dw 0 4 n 03(5) = Glo, a, (ger = 1) ‘| 6(w) dw (23) ° Also, the ratio of the transient acceleration response variance to the ground acceleration variance is: 2 _ ols) a oe a 7 * F = (oda, (ee 71) + F*(a,) (30) These results are analogous in form to Equations 17 and 19 and, in fact, converge to them when the product (zu, 5) grows large. Again, if the damping is Tight, little accuracy is lost by replacing the factor [(n/4c,) - 1] by 1/4c, in Eqs. 28 and 29. 1.4 Prediction of Maximum Response To evaluate a peak response within the format of Eq. 4, the factor r. must be estimated. As stated previously, this requires the solution of the “first passage" problem. Most of the literature on the first passage problem, however, deals with the stationary response to Gaussian white noise of a lightly damped linear |-degree system. The reader is referred to Crandall (1970) for an -27- excellent state-of-the-art review. A relatively simple approximate pro- cedure has been derived by Vanmarcke and his associates (Vanmarcke, 1969, 1975; Corotis et al., 1972) to predict the maximum responses of a rather general linear system exposed, suddenly and for a limited time, to steady- state Gaussian excitation. Considering transient response, it leads to the following approximate equation for the peak factor r, oP 7 "spt [200 2n U1 = exp(-6,(s) “7TOTET | (31) where n= (Ry(S) So/2n) (-1og (32) s,p 15 plotted in Figure 6 for several values of c,. The peak factor, then, is a function of the spectral moments of the response als), dy(s) and the “equivalent stationary response duration” sy. An estimate for the latter ‘is given by Vanmarcke (37) as: s wol-2fope0 foR(s/2) - }] (33) At very low periods, the response spectral moments are equal to those of the ground motion; i.e., als) = and 6y(s) = 55 further, since stationarity is achieved after a few cycles, 0*(s) /07(s/2) = 1. At moderate and high periods, the peak factor is determined using the paraneters 9,(s) =u, 5 8y(3) = (4e5/n)* and 95(5) /08(s/2) = (1 = M5) (1 =e) (34) In the transition period range, the input parameters are combinations of the ones given. MUM VALUE sp REDUCED MAXI -28- a *5| w Cro Stationary response solutions 100 1000, —ns FIGURE 6 - PEAK FACTOR r, |, FOR PREDICTION OF MAXIMUM NON-STATIONARY RESPONSE -29- The values of the factor (-1og p)"! in Eq. 32 corresponding to P=0.5, 0.9 and 0.99 are about 1.4, 10 and 100 respectively. The factor (-10g p)"' may be substituted by (1-p)"! when the probabitity p is very close to one. When p= e"!=0.368, (-109 p)"!=1 and n equals the average number of cycles of response motion (0,5/2n). Conclusions Equations 4, 29 and 31 in this section define the relationship between the response spectrum for an arbitrary damping value and the power spectral density of ground motion. Conversions from Sy to G(w) can therefore logically be made and used for generation of “response spectrum compatible" simulated motions. Such a capability has been implemented in the program SINQKE and is further discussed in the following section. -30- 2. THE PROGRAM SIMQKE AND PROPERTIES OF SIMULATED EARTHQUAKES The program SIMQKE has these major capabilities: it computes a power spectral density function from a specified smooth response spectrums it generates statistically independent artificial acceleration time histories which match a specified response spectrum and it refines the spectral match through an iterative procedure. It also performs a baseline correction on the generated motion to ensure zero final ground velocity and, of course, it calculates response spectra with the time histories as input. A general flow chart is shown on Figure 7. This section discusses some of the above algorithms and comments on the properties of the resultant simulated earthquakes. An explanation of the input to SIMQKE and a listing of the program are included in Appendices A and B. 2.1. Artificial Motion Generation Procedure The method used by the program for artificial motion generation is that discussed in the introduction, i.e., superposition of sinusoids (Eq. 1) having random phase angles and amplitudes derived from a stationary power spectral density function of the motion. To simulate the transient charac ter of real earthquakes, an envelope function of the form shown in Fig. 3b is used by the program. The final simulated motion, z(t), then becomes: z(t) = 1(t) TA, sin (ust +4,) (35) n It is stationary in frequency content with a peak acceleration close to the target peak acceleration. The program empirically uses a frequency -31- OPTION 1 OPTION 2 OPTION 3 Read Target Sy Read G(w) From Read G(u) at NKK From a gy Bhigh © gy Points Fron sys TO 454 terpolate S| Tnterpolate Gi hateeiied Tterpolate erpo a Modi ficati to NKK Points” to NKK Points peiiieceions [ Compute Er Compute NFQ Phas Nodify G(w) From Sy At — + ired cele Angles As Desire t Interpolate G(w) to NFQ Values From .51,, to 2.0 “hy Compute Accelerogram. NFQ Contributing Sinu- vids At Any Step Multiply Motion By Envelope I(t) If Perform Baseline Correction Adjust Peak Accelera- tion Compute Response Modify 6(w) Compute NFQ Phase} Angles Motions FIGURE 7 - SIMQKE FLOW CHART -32- range bounded by 0.5 w,4, and 2.0 upsgp, where wy, and u,;9, define the high range for which response spectral values are required. The frequency step used is 44 halfband width 5, = 0.005 u,3 it is intended to be a fraction of the smallest min? 2.2 Generation of Power Spectral Density Functions Fron Smooth Target Response Spectra The preceding section established a relationship between the response spectrum and the spectral density function of ground motion at a site. Soe- cifically, Equations 4 and 29 may be combined to yield: w? (Sy)? ar is hn &(u,) = —- ee | Gla) du (36) "Seq [a | with r, , defined by Eq. 31 and c. by Eq. 26. It is seen that the relation- ship between Sy and G(w) is not unique. It depends on the chosen strong motion duration s, and the exceedance probability level p assigned to Sy and on the damping level ¢ involved. Eq. 36 is used iteratively; the pro- gram starts at the lowest natural frequency w,, when the contribution of the integral term on the right side of Eq. 36 is negligible. At an arbi- trary frequency w;, the integral of G(w) up to w, is evaluated numeri- cally and G(w, , ,) can be found. The G(w) derived from Eq. 36 is subsequently used to generate arti- ficial motions by using Eq. 35. For any of these motions, the specified smooth response spectrum and the calculated response spectrum at the corres~ ponding damping value closely "match". -33- 2.3 Improving Response Spectral Matches It should be noted that the peak acceleration which results from the algorithm just described is a random variable. To satisfy requirs- ments concerning the value of the peak acceleration of a motion, the program raises or lowers the generated peak acceleration to match a specified value. This artificial modification only affects the response spectral values in the very high frequency range. Additionally, if the computed response spectrum varies excessively above and below the target, a cycling option in the program may be used to smoothen the match. In each cycle, the calculated real response is compared with the target at a set of control frequencies. The ratio of the desired response to the computed response is obtained at each control frequency and the corresponding value of the power spectral density is modified in proportion to the square of this ratio, i.e., at any cycle i Sylo) : . Glo); 41 = Sw); XN (37) yw where Sy(w) is the target spectral value. With the modified spectral density function, a new motion is generated and a new response spectrum is calculated. This procedure should not be expected to be convergent at all contro? frequencies; the response at a control frequency is depen- dent not only on the spectral density function value at that frequency, but also on other values at frequencies close to the frequency of interest. For this reason, no convergence criteria is implemented in the program; the user simply specifies a desired number of iterations. =34- 2.4 Properties of Generated Motions The simulation procedure and the program allow spectral density functions to be specified directly. For example, a Kanai-Tajimi spectral form or a band-limited white noise may be specified and simulated motions may be generated. The following comments, however, are based on studies of fifteen artificial motions, all generated from the same key input parameters as follows: ENVELOPE FUNCTION: Fa ot tee DAMPING RATIO = 0.02 PEAK ACCELERATION = 0.3 g TARGET RESPONSE SPECTRUM = in/sec Period - sec -01 18 +025 It is only for "the Sy to G(w) option", then, that the following comments are applicable. -35- Accelerograms and Peak Accelerations Figure 8 shows one (typical) simulated time history. The peak acceleration can be made to equal 0.3 g (116 in/sec2) by raising or depressing the actual computer - generated peak acceleration. The Fifteen actual peak accelerations ranged from 74.1 in/sec® to 118.1 in/sec® with a mean peak acceleration equal to 94.3 in/sec®, The mean spectral response, then, would have fallen below the target values at very high fre- quencies if the peaks had not been modified. Response Spectra The target spectrum used was an “eyeballed" smooth version of the mean response spectrum of 39 real earthquakes (39). Response spectra for four artificial motions are shown in Figures 9 and 10; the target spectrum is superimposed on each. It must be noted that no iterations (as described on p. 33) were performed to smooth the generated spectra. (The effect of smoothing will be considered in the following section.) The variability of the generated responses about the target increases with the natural period T; in extreme and isolated cases, variations of +100% from the target occurred. Considering the average spectrum of the 15 motions, such differences were smoothed; indeed, Table 3 indicates that the "match" was very good throughout the period range. Additional information on the variability of structural responses is reported in another project report (39). Multiple time history analyses using statistically similar artificial earthquakes were performed for three 4-DOF systems. Coefficients of variation of responses (relative displacement, absolute acceleration, interstory distortion) varied from S$ 9 NOILV431399¥ -36- ACCELERATION TIME HISTORY PEEEUEVUEL ETC LEP EP EQ ee gerry e gerry ay TIME SECONDS FIGURE 8 - TYPICAL SIMULATED ACCELEROGRAM -37- RESPONSE SPECTRUM ne MAXIMUM VELOCITY ¢IN/SEC) tee ome ' — Tne" NATURAL PERIOD (SECONDS) RESPONSE SPECTRUM se sane MAXIMUM VELOCITY CIN/SEC) sme tt tear ‘a Tae NATURAL PERIOD —CSECONDS) FIGURE 9 - TYPICAL RESPONSE SPECTRA FOR SIMULATED MOTIONS =38- RESPONSE SPECTRUM — sme MAXIMUM VELOCITY CIN/SEC) sme me . se NATURAL PERIOD (SECONDS) RESPONSE SPECTRUM ne i 4 sme MAXIMUM VELOCITY CIN/SEC) i 4 ame * sae Tae NATURAL PERIOD (SECONDS) FIGURE 10 - TYPICAL RESPONSE SPECTRA FOR SIMULATED MOTIONS Ns wnuioads L39u¥L HLOOWS ¥ WOH G3ATYIO NOTLONN4 ALISN3Q WALI3dS - LL 3UNOTd (SONO93S) —«OIU3d TWENLYN fd -39- (€##93S/Z#4N1) - (429 “we NOILONN4 ALISN3Q IVYL93¢s SDIVNDHLUVS GALVINWIS SL 40 WneLIIdS NVBW HL HLIM WOYLI3dS LIDYVL 4O NOSTUVdNOD - € JTevL sse-et o0s"ct 00g" £96°ZL 299°LL o0e"€ 09°61 vd2°8L 809°2 989°0z €29°8L tsere 829° zz 602 °6L 219° L S16°9L Sps"6L #98" 9ub"02 9ve"6L Olt 250-12 0°02 8" L9°02 0°02 OplL” 19°61 0°02 Liss 62°61 0°02 L6Sb" t6L6L S¥2"6L gece: t60°9t 289°91 2962" S/9°EL SEL“EL £062" 250° LL 6£€70L SS6L" 809°L ls°9 SbSL" 188° s2e"p eg2l" U6E"E 828°2 7660" $66°2 692"2 6080" SLL'2 lz" 8990" 6b" 86E°L ozs0° G02" 20L"L €2b0" 026° 898" veeo" 819° 299° 220° 21° LLg" L220" 0b" 06e* o8to" $62" 982° 2p10° 222" giz" Sto" SIAWNDHLYVI SL JO 2s/Ub 3 TVA aoruad ASNOdS3Y 3OVUIA |SNOdS3Y LIDUVL 82 lz 92 sz £2 22 la 02 6L al at gt SL bl €L ra LL Hamenenanc -40- about 0.14 to 0.3. Greater C.0.V.'s were generally associated with the longer period structures (39). Spectral Density Functions Figure 11 shows the spectral density function generated from the target response spectrum of the fifteen earthquakes. It is similar to the Kanai-Tajimi form (Eq. 3) and shows a peak at the intersection of the velocity and acceleration ranges of the smooth response spectrum. To gain a better understanding of the relationship between response spectra and spectral density functions, some parametric studies are reported in the following section. -41- 3. PARAMETRIC STUDIES USING SIMQKE The objectives of the studies reported herein were: (i) to give insight into the variations of the ground motion spectral density func- tion G(w) as derived from parametrically different target response spectra and corresponding to different durations, and (ii) to quantify the effect of "smoothing" on the calculated response spectrum. 3.1 Effect of Smoothing Cycles For this study, key input parameters were: Duration - 30 sec. Maximum Ground Acceleration - 0.18 g Target Response Spectrum - A design response spectrum for 5=0.02. It is defined by the values in Table 4 (which define straight line segments on a log-log plot). T (sec) Sy Cin/sec) -01 0.11 -03 0.33 215 6.0 5 20.0 2.0 20.0 TABLE 4 For the constant acceleration and constant velocity ranges, Table 5 contains second moment statistics of the response spectral values at the end of a specified number of smoothing cycles. A visual measure of the smoothness achieved is given by Figures 12a and 12b. Response spectra were also computed for c= 0.005, 0.01 and 0.05 at the end of the fourth cycle of iteration to determine how a single earth- -42- S3T0A9 SNIHLOOWS 40 Y3SWNN “SA SOILSILWIS 3SNOdS3Y g 718vL 220° sol | ato 60° gb6 | sces2 9 37049 920° es" | 0°02 290" eL'ol | 8z79sz § 31949 £80" so | 26°61 130° ge°2zi | serve y 37049 960" eel | 62°6L 590° z29°9L | oz'ps2 € 31949 lau “ez | 0s°6L £50" oLvez | 6b eve 2 37949 ue ae | BL6L SEL zoree | o2'eve L 37049 a © w a © w oas/ut oz = “s yabuey aBuey £3190, a, juersu0D oasoz>1>g° 1288/Ul ee" 1s2_=°s 396401 aBliey Uolyesa[aooy 4ue3su0D vas ¢° > 1 > 1° -43- quake matches a set of specified response spectra. Figures 13, 14 and 15 show the computed spectra with the corresponding smooth design spectra superposed. For this set of specified spectra, the computed spectra tend to fall below the target as the damping increases. Such a generalization cannot be extended, however, to other sets of design spectra (e.g., NBK). 3.2 Effect of Duration on G(w) and Computed Sy Key fixed input parameters for this study were: Damping - 0.02 Maximum Ground Acceleration - 0.3 g Target Response Spectrum - The smoothed average response spectrum of 39 scaled real earthquakes. It is identical to that defined on page 34. Five values were used for the strong motion duration: seconds. 5, 10, 20, 40 and 60 Figures 16 through 20 indicate that G(w) decreases as the duration increases, the effect being more marked in the high period range. This is reasonable since a given response spectrum level Sy is expressed as the product r, (s) (See Eq. 4). The veak factor r. p increases with sop°y the number of response cycles or, alternatively, it decreases as the dura- tion of the base motion decreases. It follows that the response standard deviation oy(s) must increase for the response to attain a specified level Sy- This, in turn, implies (from Eq. 9) that the area under the G(w) func- tion must increase. The increase in G(w) is more marked at high periods because the response variance o)(s) becomes increasingly dependent on indi- ¥ vidual G(w) ordinates as the period increases. This can be deduced from Equation 17 and Figure 5. WaL03dS 3SNOdS3Y O3LVINITV GBHLOONS - ZL 3UNDdT4 saoA 9 “200 = 2 (Gd Sel p *20°0 = 2 (e (SONO93S) QOIYad WaNLYN ¢SaNo93S) GOIY3d WYNL¥N ~44- (93S/NI} ALIDOT3A WAWIXVW wet WNYL9adS 3SNOdS3y WNY1I3dS 3SNOdS3y (93S/NI) ALIDO13A WAWIXVH lo’ = 2 YOd WNULIIdS ASNOdSTY GILVINIWI - vL aUNDLS (saNo93s) OO1Y3d WaNLYN -45- WNYLI3dS SSNOdS3y CQ3S/NID ALIION3A WOWIXVH soo" = 3 YOd WNYLITdS ISNOdSTY GALVINDTWD - EL JUNDIA ¢SGNOI3S) QOIY3d WYNLVN WNYLIAdS 3SNOdS3y (€93S/NI) ALIDO13A WAWIXVH MAXIMUM, VELOCITY CIN/SEC) -46- RESPONSE SPECTRUM NATURAL PERIOD (SECONDS) FIGURE 15 - CALCULATED RESPONSE SPECTRUM FOR z= .05 -47- Note also that the spectral density becomes increasingly sensitive to duration as the duration decreases. In fact, its shape at high periods for short duration motions does not appear to be typical of a real ground motion. The implication is that, for the given duration and damping, such high responses (as given by the specific smooth target spectrum) are not to be expected. By examining the five computed response spectra, it is evident that the spectral match in the long period range becomes much smoother as the duration increases. Short durations preclude significant long period sinusoidal components in Eq. 4. The responses of long period structures remain transient during short motions and, therefore, more dependent on the initial details of the motion (random phase angles). 3.3 Variation of G(w) Derived From a Typical Set of Smooth Design Spectra For the aseismic design of nuclear power plants, Newmark, Blume and Kapur (26) have proposed design response spectra for varying values of the damping ratio. In this study, NBK spectra for ¢=0.005, 0.02, 0.05, 0.07, 0.1 were in turn used as target spectra, a simulated motion was generated from each and the response spectrum for the corresponding damping was computed. Other key input parameters were s'=20 sec, a,,,=1.0 9. Figures 23, 24 and 25 indicate that for ¢20.05, the design spectra are consistent in the sense that the implied G(w) is essentially constant. For ¢=0.02 and 5=0.005, the resultant G(w) is substantially different and, in fact, not typical of real earthquakes. Moreover, the computed response spectra are noticeably lower than the target spectra in the long period range. For an estimated duration of 20 seconds then, the design spectra for ¢=.02 and 0.005 in the long period range appear to be overly conservative. GCW) - CIN##2/SECe*3) MAXIMUM VELOCITY CIN/SEC) -48- SPECTRAL DENSITY FUNCTION Ps sane sane hawt t i t © ent Seg NATURAL PERIOD (SECONDS) RESPONSE SPECTRUM wet sane sae met NATURAL PERIOD (SECONDS) FIGURE 16 - G(w) and Sy FOR s=5 sec GOW) - CIN®#2/SEC#«3) MAXIMUM VELOCITY ¢IN/SEC) -49- SPECTRAL DENSITY FUNCTION “ one tenet NATURAL PERIOD (SECONDS) RESPONSE SPECTRUM we tment ame met tm Tae" cane zane NATURAL PERIOD (SECONDS) FIGURE 17 - G(w) and Sy FOR s=10 sec GOW) - CIN#*2/SEC#*3) MAXIMUM VELOCITY CIN/SEC) -50- SPECTRAL DENSITY FUNCTION met met ne met ‘a NATURAL PERIOD (SECONDS) RESPONSE SPECTRUM sane] sme NATURAL PERIOD —CSECONDS) FIGURE 18 - G(w) AND Sy FOR s=20 sec GOW) ~ CIN**2/SECe#3) CITY CIttvezcy MAXIMUM VEL; -5- SPECTRAL DENSITY FUNCTION sane NATURAL PERIOD (SECONDS) RESPONSE SPECTRUM NATURAL PERIOD (SECONDS) FIGURE 19 - 6(w) AND Sy FOR s = 40 sec GW) - CIN®*2/SEC#*3) Y CIN/SEC) MAXIMUM VELOCT -52- SPECTRAL DENSITY FUNCTION ane sme amet eet = 7 Te NATURAL PERIOD — CSECONDS) RESPONSE SPECTRUM sane amet sane amet NATURAL PERIOD (SECONDS) FIGURE 20 - G(w) AND Sy FOR s= 60 sec GW) - CIN*#2/SEC#*3) MAXIMUM VELOCITY CIN/SEC) -53- SPECTRAL DENSITY FUNCTION er Tn oo ae NATURAL PERIOD (SECONDS) RESPONSE SPECTRUM net set set sane omen a co = NATURAL PERIOD (SECONDS) FIGURE 21 - G(w), COMPUTED Sy, AND SMOOTH NBK TARGET Sy FOR 5 = .005 54. SPECTRAL DENSITY FUNCTION 220] GCW) - CIN*#2/SEC#*3) [ an ae NATURAL PERIOD (SECONDS) RESPONSE SPECTRUM one sane sae MAXIMUM VELOCITY CIN/SEC) se nt se an me NATURAL PERIOD (SECONDS) FIGURE 22 - G(w), COMPUTED Sy, AND SMOOTH NBK TARGET Sy FOR ¢= .02 -55- SPECTRAL DENSITY FUNCTION reer - GcW) - CIN®#2/SEC#*3) 1a a i vam NATURAL PERIOD (SECONDS) RESPONSE SPECTRUM _ sae tment MAXIMUM VELOCITY CIN/SEC) tae amet — fe ae NATURAL PERIOD —CSECONDS) FIGURE 23 - G(w), COMPUTED Sy, AND SMOOTH NBK TARGET Sy FOR y= .05 -56- SPECTRAL DENSITY FUNCTION ° i GB x | ? = é a Bao g 3 z 3 7 3 at von a sae NATURAL PERIOD (SECONDS? RESPONSE SPECTRUM vane sae MAXIMUM VELOCITY CIN/SEC) sae NATURAL PERIOD —-CSECONDS FIGURE 24 - G(w), COMPUTED Sy, AND SMOOTH NBK TARGET Sy FOR 5 = 0.07 -57- SPECTRAL DENSITY FUNCTION a 1 3 ad A tI 8 BR, é g z ’ z 8 a bet NATURAL PERIOD SECONDS RESPONSE SPECTRUM ne 3 g S ne z 8 D men g g z 5 g = NATURAL PERIOD (SECONDS) FIGURE 25 - G(w), COMPUTED Sy, AND SMOOTH NBK TARGET Sy FOR 5 =0.1 -58- SUMMARY AND CONCLUSIONS Motivations For Earthquake Simulation Simulated earthquake motions are used for time-integration analyses where no satisfactory real motion records are available. Evaluation of structural response and floor response spectra using simulated motions is frequently an important step in the seismic design of constructed facilities. Simulation Model ‘A common model for synthesizing earthquakes is that of superposing sinusoidal components with random phase angles. The input parameters for such a model are, then, the amplitudes and phase angles of the contributing sinusoids as well as the characteristics of the variation of motion inten- sity with time, especially the duration of the motion. The amplitudes are determined from estimates of the Fourier spectrum or the spectral density function of the ground motion. These amplitudes may be assumed to be vary- ing in time or constant for the duration of the earthquake. Although the constant amplitude assumption unrealistically implies a stationary frequency content of the motion, satisfactory simulated motions can be generated with such models. There are several alternatives for obtaining G(w) or f(w). Ideally, geophysical models should predict them for particular sites. The logic of this approach has been recognized by others (31) and Berril] (2) has recently proposed such a model. In reality, current design practice utilizes and specifies smooth response spectra. The zero damping spec- trum may be used (Ref. (31) and (19)) as an estimate of the Fourier -59- amplitude spectrum of the ground motion. Alternatively, the specified response spectrum Sy may be used to derive the spectral density function G(w) by following the procedures described in Section 1 of this report. This latter approach constitutes the unique capability of the program described herein: SIMQKE. Another possible methodology (in fact, an option of SINQKE) involves assuming a standard shape, such as the Kanai-Tajimi (Eq. 3) for the spec- tral density function. Assuming the ground damping value is considered constant, this model requires determination of the duration, the central Frequency w, and the "intensity" G . The latter quantity could be estimated by relating the variance of the ground acceleration to the area under the spectral density function. Compatible Motions Compatible motions may mean motions which have been scaled to match a specified peak acceleration or a Spectral Intensity as defined by Housner (19). Given the present accepted methodology of specifying smooth design response spectra, however, they are herein assumed to be motions whose response spectrum envelopes, in some way, a smooth design spectrum. Alternate methodologies for achieving this objective are reviewed in the introduction. The program described herein, SIMQKE, achieves such capability by: (i) deriving the spectral density function from the response spectrum which is to be matched, (ii) adjusting the generated peak accelera- tion to match the target value, and (iii) adjusting the ordinates of the spectral density function to smoothen the match. Step 3 may not be necessary depending on the objectives of the user. Section 2 has demonstrated that without smoothing, the average response ~60- spectrum (of a set of simulated motions) matches the smooth target spectrum very closely. Further, the variability of the resultant spectra was quantified in Table 5 for a typical case. Issues On The Use of Simulated Motions The common practice of generating simulated motions (for use in design) from a single spectral density function appears to be justi- fied only if one earthquake source accounts for nearly all of the total site seismic risk, or if the local geology is principally responsible for shaping the frequency content of the ground motion. Otherwise, it is necessary to re-examine the data on seismicity and attenuation to determine the relative likelihood of occurrence of the various types of site ground motion and frequency content. These issues need consider- able further study. Another project report (39) explores the relationships among alternate methods of dynamic analysis and discusses the variability in structural responses obtained from each method. It is shown therein that multiple dynamic analyses based on a set of recorded accelerograms, from different real earthquakes and sites, yield higher coefficients of variation than those obtained from random vibration analysis or from a set of simulated ground motions with conmon expected frequency content and duration. The variability in responses obtained by time integration of simulated motions (or random vibration analysis) reflects only inherent uncertainty due to random phasing, not the uncertainty in such ground motion parameters as duration and dominant ground frequency. 10. n. 12. 13. -61- REFERENCES Barstein, M.F., "Application of Probability Methods for the Design Against the Effect of Seismic Forces on Engineering Structures,” Proc. 2nd WCEE, 1960. Berrill, J.B., Ph.D. Thesis, California Institute of Technology, March, 1975. Bolotin, V.V., Statistical Methods in Structural Mechanics, Holden - Day, 1969. Brady, A.G., "Studies of Response to Earthquake Ground Motion," Ph.D. Thesis, California Institute of Technology, Pasadena, Califor- nia, 1966. Bycroft, G.N., "White Noise Representation of Earthquakes," Journal of the Engineering Mechanics Division, ASCE, Vol. 86, No. EM2, T=16, April 1960. Caughey, T.K. and Stumpf, H.J., "Transient Response of a Dynamic System Under Random Excitation," Journal of Applied Mechanics, 563 - 566, 1961. Corotis, R. and Vanmarcke, E.H., "Time Dependent Frequency Content of System Response," Journal of Engineering Mechanics Division, ASCE, October 1975. Corotis, R., Vanmarcke, E.H., and Cornell, C.A., "First Passage of Nonstationary Random Processes," Journal of Engineering Mechanics Division, Proc. ASCE, Vol. 98, No. EM2, pp. 401-414, 1972, Crandall, S.H. and Mark, W.0., Random Vibration in Mechanical Systems, Academic Press, New York, 1963. Crandall, S.H., "First Crossing Probabilities of the Linear Oscillator,” Journal of Sound and Vibration, Vol. 12, 285 - 300, 1970. Donovan, N.C., Valera, J.E., and Beresford, D.J., "Statistical Uncer- tainty of Design Based on Smoothed Response Spectra," Proceedings of the U.S. National Conference on Earthquake Engineering, Ann Arbor, Michigan, June, 1975. Franklin, J.N., "Numerical Simulation of Stationary and Nonstationary Gaussian Random Processes," SIAM Review, Vol. 7, No. 1, 68-80, 1965. Hadjian, A.H., "Scaling of Earthquake Accelegrams - A Simplified Approach," Journal of the Structural Division, ASCE, Vol. 98, No. ST2, pp. 547-551, February 1972, 4. 15. 16. Ww. 18. 19. 20. a. 22. 23. 24. 25. 26. 27. -62- REFERENCES (cont.) Hammond, S.K., "On the Response of Single and Multi Degree of Freedom Systems to Nonstationary Random Excitations," Journal of Sound and Vibration, Vol. 7, 393-416, 1968. Hou, S., "Earthquake Simulation Models and Their Applications," M.I.T. Department of Civil Engineering Research Report R68-17, Nay 1968. "Properties of Strong Ground Motion Earthquakes," BSSA, 1, 1947. Housner, G.W. Vol. 37, 19 Hudson, D.E., "Some Problems in the Application of Spectrum Techniques to Strong Motion Earthquake Analysis," BSSA, Vol. 52, pp. 417-430, 1962. Jenkins, J.M., "General Considerations in the Analysis of Spectra," Technometrics, Vol. 3, 133-166, 1961. Jennings, P.C., Housner, G.W., and Tsai, N.C., "Simulated Earthquake Motions," EERL California Institute of Technology, Pasadena, Califor- nia, April, 1968. Kameda, H., “Evolutionary Spectra of Seismogram by Multifilter," J.E.M. Division, ASCE, December, 1975. Kanai, K., "An Empirical Formula for the Spectrum of Strong Earthquake Notions," Bull. Earthquake Research Institute, U. of Tokyo, Vol. 39, 8-95, Tokyo, Japan, 1961s Lin, Y.K., Probabilistic Theory of Structural Dynamics, McGraw-Hill, New York, 1967. Liu, S.C., "On Intensity Definitions of Earthquakes," Journal of the Structural Division, ASCE, May, 1969. Liu, S.C. and dhaveri, D.P, "Soectral and Correlation Analysis of Ground Motion Accelerograms," BSSA, Vol. 59, 1517-1534, August, 1969. Liu, $.C., "Evolutionary Power Spectral Density of Strong-Motion Earthquakes," BSSA, Vol. 60, No. 3, 891-900, 1970. Newmark, N.M., Blume, J.A., and Kapur, K.K., "Seismic Design Spectra for Nuclear Power Plants," J. Power Division, Proc. ASCE, Vol. 99, No. P02, 287 - 303, November 1973. Penzien, J. and Liu, S.C., "Nondeterministic Analysis of Nonlinear Structures Subjected to Earthquake Excitations," Proc. 4 WCEE, San- tiago, Chile, 1969. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. -63- REFERENCES (cont.) Priestley, M.B., "Power Spectral Analysis of Nonstationary Random Processes," J. Sound and Vibration, Vol. 6, 86-97, 1967. Rizzo, P.C., Shaw, D.E., and Jarecki, S.J., "Development of Real / Synthetic Time Histories to Match Smooth Design Svectra," Proceedings of the Second International Conference on Structural Itechanics in Reactor Technology, American Nuclear Society, Berlin, Germany, September, 1973. Rosenblueth, E., "Probabilistic Design to Resist Earthquakes," J. Engr. Mechanics Division, ASCE, Vol. 90, No. EMS, 189-219, 1964. Scanlan, R.H., Sachs, K., "Earthquake Time Histories and Response Spectra," J.E.M. Div., Proc. ASCE, August 1974. Sixsmith, E. and Roesset, J., “Statistical Properties of Strong Motion Earthquakes," M.I.T. Department of Civil Engineering Research Report R70-7, January, 1970. Tajimi, H., "A Statistical Method of Determining the Maximum Response of a Building Structure During an Earthquake," Proc. 2 WCEE, Vol. 2, 781-797, Science Council of Japan, Japan, 1960. Isai, N.C., “Spectrum Compatible Motions for Design Purposes," Journal of Engineering Mechanics Division, ASCE, April, 1972. Vanmarcke, E.H. and Cornell, C.A., "Seismic Risk and Design Response Spectra," ASCE Specialty Conference on Safety and Reliability of Metal Structures, Pittsburgh, Pennsylvania, November, 1972. Vanmarcke, E.H., "Properties of Spectral Moments with Applications to Random Vibration," Journal of the Engineering Mechanics Division, ASCE, April, 1972. Vanmarcke, £.H., Chapter 8 of Seismic Risk and Engineering Decisions, Lomnitz, C. and Rosenblueth, E., Editors, Elsevier Publishing Co-, (to be published). Vanmarcke, E.H., “First Passage and Other Failure Criteria in Narrow Band Random Vibration: A Discrete State Approach," M.I.T. Department of Civil Engineering Research Report R69-68, 1969. Vanmarcke, E.H., Biggs, J.M., Frank, R., Gasparini, D., Gazetas, G., Arnold, P., and Luyties, W., "Comparison of Seismic Analysis Procedures for Elastic Multi-Degree Systems," M.I.T. Department of Civil Engineer- ing Research Report R76-5, Order No. 528, January, 1976. Figure 10 n 12 13 14 15 16 7 18 19 20 2 22 23 24 25 -64- LIST OF FIGURES Title Spectral Density Function (s.d.f.) 6(w) a. Narrow Band Processes and Corresponding Spectral Density Function b. Wide Band Processes and Corresponding Spectral Density Function Intensity Functions Used to Simulate the Transient Character of Real Earthquakes a. Band-Limited White Noise s.d.f. b. Kanai-Tajimi s.d.f. Contributions to Variance of Response; a) From a Narrow Range Around w, and b) From a Frequency Range (0, u,) Peak Factor rs,p For Prediction of Maximum Non- Stationary Response Flow Chart for the Program SIMQKE Typical Simulated Accelerogram Typical Response Spectra for Simulated Motions Typical Response Spectra for Simulated Motions Spectral Density Function Derived From a Smooth Target Spectrum Sy Smoothed Calculated Response Spectra for ¢=0.02 Calculated Response Spectrum for ¢ = 0.02 Calculated Response Spectrum for & = .005 Calculated Response Spectrum for ¢ = .01 G(w) and Sy for s=5 sec G(w) and Sy for s=10 sec G(w) and Sy for s=20 sec G(w) and Sy for s=40 sec G(w) and Sy for s=60 sec G(w), Computed Sy, and Smooth NBK Target G(w), Computed Sy> and Smooth NBK Target G(w), Computed Sy, and Smooth NBK Target &(w), Computed Sy, and Smooth NBK Target G(w), Computed Sy, and Smooth NBK Target 22 28 31 36 37 38 39 ~65- LIST OF TABLES Title Approximate Values for Parameters of Relative Frequency Content Spectral Parameters Computed From Squared Fourier Amplitudes |f(w)|* of Earthquakes Comparison of Target Spectrum with the Mean Spectrum of 15 Simulated Earthquakes Target Response Spectrum Used in Study to Quantify Effects of Smoothing Cycles Response Statistics as a Function of the Number of Smoothing Cycles 4a 42 SIMQKE: A PROGRAM FOR ARTIFICIAL MOTION GENERATION USER'S MANUAL and DOCUMENTATION November 1976 Department of Civil Engineering Massachusetts Institute of Technology SIMQKE: A PROGRAM FOR ARTIFICIAL MOTION GENERATION The program SIMQKE has these major capabilities: it computes a poner spectral density function from a specified smooth response spectrum; it generates statistically independent artificial acceleration time histories and tries, by iteration, to match the specified response spectrum. It also performs a baseline correction on the generated motion to ensure zero final ground velocity and, of course, it calculates response spectra with the time histories as input. The object herein is to describe briefly the algorithms used in the first two capabilities. An explanation of the input to SIMQKE and complete flowchart (Fig. 1) are givens furthermore, an example and program listing are appended. BRIEF DESCRIPTION OF THE MOTION GENERATION PROCEDURE The method used by the program for artificial motion generation is based on the fact that any periodic function can be expanded into a ser‘es of sinusoidal waves: x(t) = DA, sin(u,t+9,) a) n * contributing sinu- A, is the amplitude and %, is the phase angle of the n™ soid. By fixing an array of amplitudes and generating different arrays of phase angles, one obtains different motions with the same general appear- ance but different details. The computer uses a random number generator to produce strings of phase angles with uniform likelihood in the range between 0 and 2m. The amplitudes AL are related to the (one-sided) spectral density func- tion G(w) in the following way: a fn Gly,) a0 = (2) Since the total power may be expressed as: 2 R Th =2 Gw,)wo ~ [ 6(w)du (3) G(w,,)du may be interpreted as the contribution to the total power of the motion from the sinusoid with frequency u,. ATlowing th number of sinu- soids in the motion to become very large, the total power will become the area under the continuous curve G(w). The power of the motion produced by using Eq. 1 does not vary with time. To simulate the transient character of real earthquakes, the steady- state motions are multiplied by a deterministic envelope function I(t). The artificial motion Z(t) then becomes: Z(t) = I(t) TA, sin(w,t + 9) (4) n The resulting motion is stationary in frequency content with a peak accel- eration close to the target peak acceleration. In this program, we have ‘incorporated three different intensity envelope functions such as “Trape- zoidal" (Hou, 1968), "Exponential" (Liu, 1969), and "Compound" (Jennings, 1968) functions as shown in Fig. 2, The program artificially raises or lowers the generated peak acceleration to match exactly the target peak acceleration. The response spectra corresponding to the motion (4) are then computed. The response spectrum for one chosen damping value is called the "target" response spectrum which the program will attempt to "match. * To smoothen the calculated spectrum and to improve the matching, an iterative procedure is implemented. In each cycle of the iteration, the calculated response is compared with the target at a set of control frequen- cies (the user specifies the number of control frequencies). The ratio of the desired response to the computed response is obtained at each control frequency and the corresponding value of the power spectral density is modi- fied in proportion to the square of this ratio, i.e., at any cycle i: 3 g(“) 7? 8) ;4y = 6(0), [3 (5) (a) where S, is the target spectral value. With the modified spectral density function a new motion is generated and a new response spectrum is calculated. The procedure should not be expected to be convergent at all control frequen- cies; the response at a control frequency is dependent not only on the spec- tral density function value for that frequency, but also on other values at frequencies close to the frequency of interest as well. Usually, it is not productive to iterate for more than about 4 cycles. If an adequate level of agreement cannot be reached, the user is advised to “start fresh" by generat- ‘ing an entirely new motion (with a new set of random phase angles). For more elaborate explanation of sone features of the program, the reader is referred to Gasparini and Vanmarcke (1976).” Gasparini, D. and Vanmarcke, E.H., “Simulated Earthquake Motions Compatible with Prescribed Response Spectra,"M.I.T. Department of Civil Engineering Research Report R76-4, Order No. 527, January 1976. -4- Option 1 Option 2 Option 3 Input Power Spec tral Input Target Input Power Spectral Deere P ty Function G(w} Response Density Function G(w) ae NK Petts T | i Interpolate Sy Interpolate G(w) Input Desired ‘to NKK Points to NKK Points Modification | Compute Power Spectral Density oe Detinad Funetion | Loop on NPA QF (turer 0 arti ticiat =“ Desired) Loop on ICYCLE (To Smoothen Response Spectrum for Target Damping) _tfirst cycle Compute Phase other than Angle first cycle| first cycle Interpolate G(w) to NFQ Values —_L____ Compute [ Saw Compute Acceleration = Fig. 1. Flowchart of SIMQKE 3 _t Trapezoidal Exponential Compound Intensity Envelope Intensity Envelope Intensity Enveloge 1 t Multiply Motion by Envelope I(t) Base Line Correction 1 Adjust Peak Acceleration tL Compute Response Spec- tra for Target Damping Loop on NDAMP Calculated Resp. Spectra for Other Dampings [ Output (plot) Figure 1 (Continued) -6- Option A: Trapexcidal Icey} 10 + TRISE Ti Bur’ Option B+ Exponential Ie} -at -pt I(t) = Ao(e"= 6” By Ie) INPUT DATA: a 2 INPUT DATA 1 ICASE=3 3, ALFAO 3. BETAO 4 AO 5. DUR INPUT DATA: 2 1CAsE = 4 2 TRISE a Te 4 IPOW 5. ALFAO ‘ -T- USER'S MANUAL SIMQKE INPUT The program SIMQKE can generally be used in three moces ov optic In all options the primary output is an acceleration time-history, but in OPTION 1 the primary input is a target response spectrum; in OPTION 2 the spectral density function is directly specified; and OPTION 3 allows the user to re-input a previously generated power spectral densi and to specify desired changes in that function, function OPTION 1 - INPUT DATA CARD 1 - FORMAT (20A4) TITLE CARD CARD 2 - FORMAT (8610.0) TS : smallest period (seconds) of desired response spectrum TL : largest period (seconds) of desired response spectrum TMINI : smallest period used to determine the range of frequencies to be represented in the simulation. Generally it is equal to Ts. TMAXI : largest period used to determine the range of frequencies to be represented in the simulation. Generally it is equal to TL. YMIN : an estimated smallest velocity response spectral value (in/sec). It is mainly used to determine the minimum ordinate on 2 plot of the spectrum. YMAX : an estimated largest velocity response spectral value (in/ It is used mainly to determine the maximum ordinste on a plo of the response. CARD 3 - FORMAT (15,6F10.4,15) ICASE : If ICASE=1, no intensity envelope is used. 2, trapezoidal intensity envelope is used. 3, exponential intensity envelope is used. 4, compound intensity envelope is used. TRISE : earthquake rise time (sec) of intensity envelope (when ICASE= 1,3; TRISE=0). TLVL : earthquake level time (sec) of intensity envelope (when ICAS: 1,35 TLVL=0). DUR : AQ: ALFAO : + parameter of intensity function (specifies when iCASE=3). TPOwW : BETAO -8- desired duration of accelerogram parameter of exponential function (equals zero when ICASE=3). parameter of intensity function (specifies wien ICASE=3,4). Parameter of compound intensity function (specifies when ICM3E =4). CARD 4 - FORMAT (2F10.4, 110, 815) DELT : AGMX TIX: NDANP NPA : NKK = NRES : NGHK : IPCH : + number of damping values for which S\(w) is desi NCYCLE : discretization interval (sec), standard input is 0.01 second. i desired maximum ground acceleration in "g's" an arbitrary odd integer which acts as a seed for the ~andom phase angle generator. d. number of cycles to smoothen a response spectrum. If NCYCLE=1, no cycling is made. nunber of artificial earthquakes desired from one target re~ Sponse spectrum (with one spectral density function). total number of periods at equal intervals on a logarithmic scale. 0 < NKK < 300 (generally NKK is on the order of 200 and 300). total number of points which describe the target response spec- trum. set NGWK=0 for OPTION 1 if IPCH=0, no punched output is obtained. if IPCH=1, punched output is obtained. CARD 5 - FORMAT (8610.0) AMOR (I) : damping coefficients in decimal parts of critical damping. The first damping entered will be the one for which cycling, if desired, will be done. CARD 6 to CARD (5+NRES) - FORMAT (2F10.4) - Target response spectrum NOTE : TsVv(1) svo(1) TSV(NRES) —s¥O(NRES) Tsv(1) smallest period (sec). TSV(NRES) largest period (sec). SVO(1) target pseudo-velocity value in in/sec. SVO(NRES) target pseudo-velocity value in in/sec. -9- OPTION 2 - (Earthquake is specified in terms of power spectral density) CARD 1 = Same as in OPTION 1 CARD 2 - Same as in OPTION 1 CARD 3 - Same as in OPTION 1 CARD 4 - FORMAT (2F10.4, 110,815) DELT : same as in OPTION 1 AGMX : same as in OPTION 1 IIX : same as in OPTION 1 NOAMP : same as in OPTION 1 NCYCLE : same as in OPTION 1 NPA : same as in OPTION 1 NKK : same as in OPTION 1 NRES : set NRES=0 for OPTION 2 NGWK : number of points that describe the power spectral density function IPCH : same as in OPTION 1 CARD 5 - Same as 1n OPTION 1 CARD 6 to CARD (5 + NGWK): - FORMAT (2F10.4) - Input power spectral density function. wo (1) GHKO(1) WO (Nwk) HO (NGWK) NOTE: WO (1) smallest frequency in rad/sec. W O(NGWK) largest frequency in rad/sec. GuKO(1) power spectral density. GWKO(NGWK) Power spectral density. OPTION 3 - (Input of a previously generated power spectral density func- tion and desired changes) . CARD 1 - Same as in OPTION 1 CARD 2 - Same as in OPTION 1 CARD 3 - Same as in OPTION 1 CARD 4 - FORMAT (2F10.4, 110,815) DELT : same as in OPTION 1 AGMX : same as in OPTION 1 TIX: NOAMP NCYCLE NPA NKK = NRES = NGHK : IPCH : =10- same as in OPTION 1 same as in OPTION 1 same as in OPTION 1 : same as in OPTION 1 is the negative number of periods for whic is provided (e.g., - 300). set NRES=0 for OPTION 3 set NGWK = -1 for OPTION 3 same as in OPTION 1 the rover spectrum CARD 5 - Same as in OPTION 1 CARDS FORMAT : CARD N2: NB: CARD Previously punched output of power spectral density functica values and corresponding periods at NXK points. Periods TQ(I): (10F8.4) (Smallest Period First) Power spectral density values GWK(I) : (6F13.3) FORMAT (2110) N2 = 1 denotes that some portions of the GWK array are to modified. N2=0 denotes that no such modifications are desired. denotes the number of portions of the GWK array which are tc be modified. FORMAT (2F10.4) - if individual values of GWK must be changed, the following cards contain first the period and then the corresponding new value of GWK. These local changes are ter- minated by a card with 99. in the first field. If no local changes are desired, only the 99. card is entered. Assuming no local changes, the following card will be: CARD ma T92 RATIO : FORMAT (8610.4) - the following cards are read only if N2=1 and N3 > 0. beginning period (sec). ending period (sec). ratio by which GWK values from period TQI to TQ2 will be multi- plied. -- Sample Data Duration - 20 sec. Maximum Ground Acceleration - 1.09 Intensity Envelope Function being used - Trapezoidal (As Fig. 3). Target Response Spectrum - A design response spectrum for £ = 0.02 as shown in TABLE I. I(t) Leveling 13 sec 1.0 t TRISE=2 TLVL=15 DuR=20 Figure 3 Tsv (sec) SVO (in/sec) 0.0143 1.0 0.0263 1.9 0.1064 22.0 0.2444 75.0 3.333 95.0 5.0 I 60.0 Table I Input Data: CARD 1: TITLE CARD 2 : TS = 0.02 sec. TL = 3.0 sec. “12+ TWIN] = 0.02 sec. TMAXI = 3.0 sec. YMIN = 0.1 in/sec. YNAX = 500.0 in/sec. CARD 3: ICASE = 2 (Using Trapezoidal Intensity Function) TRISE = 2.0 sec. TLVL = 15.0 sec. DUR = 20.0 sec. AQ = 0.0 ALFAO = 0.0 BETAO = 0.0 IPOW = 0 CARD 4; DELT = 0.01 sec. 1.06 1235 3. (for 0.02, 0.05, 0.1) NCYCLE = 2 NPA = 1 NKK = 200 NRES = 6 NGWK = 1PCH CARD 5: AMOR(1) = 0.02, AMOR(2) = 0.05, AMOR(3) = 0.1, which are the damping value. Note that the first one is the target damping. CARD 6 - CARD 11, as in TABLE I. Output Data: The calculated spectral density function is shown on Fig. 4, The response spectra are shown on Figs. 5,6 and7, for damp- ing ratios of 0.02, 0.05, and 0.10, respectively. Furthermore, the target response spectrum for damping 0.02 is shown on each plot. GCW) - CIN®®2/SEC##3) -13 SPECTRAL DENSITY FUNCT IS}: sone — 1 HH | iH i t sone | { t I a | _ H HHH Wf jit T Pri I ia send Tene ete NATURAL PERIOD (SECONDS) Fig. 4 MAXIMUM VELOCITY CIN/SEC) -14- RESPONSE SPECTRUM tons tone Pa { | sone" Tene tone sere NATURAL PERIOD (SECONDS? Fig. § (€ = 0.02) Target response spectrum and calculated response spectrum for 2% damping MAXIMUM VELOCITY CIN/SEC) -15- RESPONSE SPECTRUM vot aone tone 3 sone yang Fae none ane NATURAL PERIOD CSECONDS} Fig. 6 (& = 0.95) Calculated response spectrum for 5% damning and 2% damped target response spectrum. MAXIMUM VELOCITY CIN/SEC) 16- RESPONSE SPECTRUM sone? sone’ sonte'?® 2one Tene" Tene pert NATURAL PERIOD CSECONDS) Fig. 7 (E = 0.1) Calculated response spectrum for 10% damping and 2% damped target response spectrum. -17- 47 SSHIM=SHENG LAT" sCLASS=C+REGION=200K 7*SR1 WEEKLY JNA 47 EXEC FORCLG®! TIME=10sLINES=10 IBRARY="SYSS,SC4020,SUBR? 4/C+SYSIN DD AMMANAAADADOAANAMAA AO ADDO OROOOOE MOO SIMOKE - A PROGRAM FOR EARTHQUAKE SIMULATION INPUT PARAMETERS REQUIRED IX--A STARTER FOR THE RANDOM NUMBER GENERATOR=IT MUST BE GDD NPA---NUMBER OF DIFFERENT MOTIONS REQUIRED ICASE: FOR STATIONARY CASE TL = THE LARGEST PERIOD VALUE FOR RESPONSE CALCULATIONS TS _- THE SMALLEST VALUE TMINsTMAX=--OPTIONAL MINIMUM AND MAXIMUM PERIODS TO NETERMINE FREQUENCY CONTENT OF THE MOTION, DEFAULT USES TS AND TL NCYCLE~--THE NUMBER OF ITERATIONS TO BE PERFORMED IS THAN THIS NUMBER=-IF NCYCLE = ls NO ITERATION IS "ADF DELT -- TIME INTERVAL USED BETWEEN POINTS NDAMP---NUMBER OF DIFFERENT DAMPINGS TO BE CONSIDERED AMOR==-ARRAY CONTAINING THE DAMPING VALUES. TRISE RISE TIME TLVL ==" INTERVAL AT THE HIGHEST AMPLITUUE NGWK -- DEFINES TYPE OF SPECTRAL DENSITY FUNCTION USFO IF NGWK = 0 + THE PROGRAM GENERATES ITS OWN POWER SPFCTRUM, IF NGWK IS NOT = 09 THEN A PIECEWISE LINEAR POWER SPFCTRUM WILL BE PROVIDED BY USER AND NGWK = NUMBER OF POINTS THAT DEFINE IT. IF NGWK IS NEGATIVEy THEN GWK WILL BE READ ALONG WITH PERTODS FOR RESPONSE CALCULATIONS ABS(NKK) = NUMBER OF POINTS FOR RESPONSE CALCULATIONS. IF NKK IS POSITIVE, THE PROGRAM WILL GENERATE A STRING OF POTNTS. ON A LOGARITHMIC SCALE FROM TS TO TL. IF NKK IS NEGATIVE, THE USER PROVIDES A LIST OF POINTS. (TSV9SVO) = POINTS WHICH DEFINE DESIRED VELOCITY RESPONSE SPECTRUM NRES~--NUMBER OF POINTS WHICH DEFINE DESIRED RESPONSF SPECTRUM IF NRES = 0+ NO DATA NEEN BE GIVEN(NO CYCLING ONLY) (W0+GWKO) ~ POINTS THAT DEFINE POWER SPECTRUM IF NGWk IS NOT = 0. TQ=--OPTIONAL ARRAY OF PERIOD VALUES FO2 RSPONSE CALCULATIONS. AGMX === MAX GROUND ACC INPUT UNIT IN S DUR --= DURATION UNITS SECONDS+INCHES=--UNLESS SPECIFIED OTHERWISE fr DIMENSION TQI(150) DIMENSION YTITL (9) eTITLO(9) DIMENSION TIT (9) ¥TIM(9) oTIMX (9) oTIMY (9) 9 TIX (9) o TITX(9) 9 TITY (9 DIMENSION ACCG (A001) »WB (300) +GWK (300) » TIME (3001) sFRO(300) + 1Q(300) #PLTVMX (10+300) »AMOR(10) » TITLE (20) » IBUF (7600) + FO (1500) +GWG (1500) 4PA(1500) 20W(1500) + TM (104300) WO (300) +GWKO (300) +S¥(300) »TSV(1010) SVO(1010) »S7 (300) + ANEWGK (300) DIMENSION PERCEN (300) EQUIVALENCE (TIME (1) 6FQ.(1)) » (TIME (1501) 9DW(1)) » (GWGCL) @PLTVMX (1) DATA TIx/* te HetRESPHAFONSE's* SPESICTRUNGIM 140 , st v DATA TIM/+ tat yt ACCE te LEROMs 'GRAMS yt tee " 9 s uv DATA BLANKs* u DATA TIT/#RESP1¥tONSE'4# SPENSICTRUte'M D2, AMPI?°NG soo DATA TITX/+ ‘ cane NAtetTURA'S#L PEt stRIODIG! Tot (SECH se tONDSFy ~18- sy oy DATA YTITL/Y ty 8G CW) tet = (Pe INBENSHASSENtCHS3t Gt) 17 DATA TITLO/*SPEC#s#TRAL #9! DENte#SITY"9* FUN 9CTIO%HIN 00 Sy DaTAa TITY/+ HeIMAXT |S *MUM Ne AVELONSPCTT Ye 98 CINE 97: soy DATA TIMK/+ HeATIMENS# (SENPPCONDE DES) 499 tyr sey DATA TIMY/+ ACCEHy"LERANSATIONS 9? 59 GarS¥ee Bey DATA BETAS sBETAL/0.005+0+2/sP1/3«14159/ ICONT=0 c c REQUIRED INPUT PARAMETERS. c 9003 READ (Syl) TITLE CALL STOIDV (15324-9950' +990) 1 FORMAT (2084) READ (547020) TSyTLeTMINI sTMAX] 9 YMIN® YMAX 7020 FORMAT (8610.0 READ (543020) ICASEsTRISE+TLVL DUR» AO¥ALFAQsBETAO, IPOW 3020 FORMAT (15,6F10.4415 READ (54129) DELT+AGMXy1IX+NDAMPsNCYCLE »NPA yNKK »NSESsNGHK, IPCH 129 FORMAT (2F10.44110+815) AGMX=AGMX#32.2812, 4262 FORMAT (2F10.4) FIRST DAMPING VALUE MUST RE ONE WHICH IS CYCLED ON. THE FIRST CURVE VALUE WILL BE PLOTTED(XESPONSE SPECTRUM eae READ (547020) (AMOR(T) 4 I=] sNDAMP WRITE (642) TITLE 2 FORMAT (1H14//42X 92044) WRITE (6430) DELT 30 FORMAT(//7X,"TIME INCREMENT = 1 4F5.3) IF (NKK,LE,0) GO TO 6301 CALL PLTX2(TS+TL¥TOsNKK) Go TO 3 6301 NKK=-NKi G C OPTIONAL INPUT PARAMETERS — IF NKK IS VEGATIVE. C GWK IS REQUIRED ONLY IF NGW< IS NEGATIVE c READ (5413) (TCT) + 1=1 ¢NKK) READ (5,888) (GWK(NKK=I+1) 911 9NKK) READ (547020) N2yN3 14 READ (544262) TCsGWe IF (1C,GT.50.0) GO TOS DO 9 I=1sNKK IF (ABS(TC-TQ(T))«LT.0.0002) GO TO 11 9 CONTINUE GO TO 14 11 GWK(NKK=I+] GO TO 14 5 CONTINUE IF (N2.EQ,0) GO TO 3 00 10 T=1.N3 READ (547020) TQ1+TQ2sRATIO DO 10 J=1yNKK IF (TQ(J)6GTeTOLsAND.TO(J) LT.TQ2) GWA (NKK=J¢1) =GWi (NKK=J+ 1) 9RATIO Wc ~19- 10 CONTINUE 3 00 4325 JENKK=Tel FRO(1)=14/TQ(1) WB (J) =6,2832/T0(1) IF (TMINI.EQ.0.) TMINL=TS WL=6.2832/TMIN1 IF (TMAX16EQ.0.) TMAX1=TL. WS=6«2832/TMAX1 = 19NKK WEND --- THE HIGHEST FREQ FOR GROUND MOTION WBEGIN ~~~ THE LOWEST FREQ FOR GROUND MOTION THE FOLLOWING OPTIONS FOR COMPUTING WEND AND WBEGIN May BF ELIMINATED SINCE BETAL ANO SETAS HAVE SEEN DEFINED INTERNALLY BY THE PROGRAM TO BE +2 AND .005 RESPECTIVELY ee2o00K8e WEND=2.0°WL IF_((5.0®BETAL) «GEe1.0) WEND=WL®(1.*5-*BETAL) WEEGIN=WS*.5 IF (BETAL«LT.0,05) WREGIN=WS#(1.-10+*BETAL) IF(ICASE*GT.1) GO TO 42 WRITE (64134) 134 FORMAT (7X+15HSTATIONARY CASE GO TO 38 42 WRITE (64135) 135 FORMAT (7XsS9HNON-STATIONARY IN INTENSITY BUT STATIONARY IN FREQ SP ZECTRUM) 38 WRITE (64106) AGMX 106 FORMAT (7Xs EXPECTED MAXIMUM GROUND ACC ="9F7425! TNe/SECe/SECs*) IF (NRES«EQ.0) GO TO 6022 READ (544262) (TSV(I) ¢Sv0(1) »T=1 «NRES) CALL_POLATE (NRES «NKK +e TSV9SV09TO9SV, WRITE (64107) TRISEsTLVL.OUR 107 FORMAT(7Xs*TRISE ="4F7.29" TLEVEL WRITE (646016) 6016 FORMAT (//411Xy "ORIGINAL POWER SPECTRUM#9//411X4 "PERIOD" VAX» S*FREQUENCY 7X "SPECTRUM! 912K 9 !R1//: 6022 IF (NGWK«EQ.0) GO TO 4260 IF(NGWK.LT.0) GO TO 9703 FT629% DURATION =14F7,2 OPTIONAL INPUT OF ORIGINAL POWER SPECTRUM IF NGWK TS POSITIVE IF TQ WAS READ IN PREVIOUSY FOR NKK NEGATIVE, THIS. OVERTDES POWER SPECTRUM 'Gwk? READ IN WITH *TO",. OPTIONAL INPUT OF DESIRED RESPONSE VELUCITY SPECTRIM IF CYCLING IS USED. enannn00 READ (544262) (WO(T) «GWKO(T) I=] sNGWK. CALL POLATE (NGWK #NKK+W0GWKO9W9 9GWK) 9703 DO 8011 I=1+NKK JENKK+ 1-1 GWKO (1) =GWK (1) BOLL WRITE (644340) TO(T) sFRO(I) +GWK (J) GO TO 6007 4260 T=(DUR*TLVL)/2. BETA=AMOR (1) CALL SVGW(NKK »WB sGWKO+SV9TsBETA+ 1640904690. 368+GSIIMs¥CP 20° ¥RR DO 6001 LLL=1eNKK LUL=NKP LLL} 6001 WRITE (€ 8901) TO(LLI) sFROCLLI) »GWKO(LLL) »RRILLL, 6 ane ae0 326 327 328 329 330 6008 007 60 8608 4080 6703 1116 iy -20- RITE (648902) WCP.OP RAAT (/7e10X«" CENTRAL FREQUENCY WC = '9F10-30//+10X0 DISPERSIO St PARAMETER Q= '4F10,39/) SET THE MAXIMUM VALUE OF SPECTRAL DENSITY FUNTION FOR PLOT XMAX= 0.0 00 327 112= 14NKK IF (XMAX~GwKO(T12)) 32653279327 XMAX=GWKO(T12) CONTINUE IF (XMAX=70.0) 32943284328 XLAT=KMAX/100« NOUM> (HF IX(XLAT) #1) #100 XMAX=DF LOAT (NDUM) GO TO 330 XMAX=70.0 CONTINUE CALL GWPLOT (NKK+040164.090609XMAXy TQ9GHKOyTITK eT AREA=SQRT (GSUM) /386.4 WRITE (646008) AREA FORMAT (/11X¥"STANDARD DEVIATION OF PROCESS = "4F7,49* GteSt ITOTAL=NDAMP#NKK 1X2 (11X72) #241 DO 585 NTOTAL=1+NPA WRITE (6460) Ix FORMAT(#1*s//10X9*A NEW PHASE ANGLE SET WITH SEED = *110) 00 8608 I=1+NKK GWK (1) =GWKO(T) TLOyYTITL) DELW=BETAS#W =W*DELW CALL DUMMY (Ws FOUT #NKK 9 WB» GWK MM) NFQ=NFQ41 GWG (NFQ FQ(NFQ DW(NFQ) =DELW AREAG=AREAG+GWG (NFQ) ®DELW SIGMS=SIGMS+GWG(NFQ) *DELWoWew IF (W.LT»WEND) GO TO 4080 DO 100 ICYCLE=1+NCYCLE ‘ouT W IS LOWEST FREQUENCY REPRESENTED IN GROUND MOTION. IF (ICYCLE.LE.1) GO To 1116 AREAG=0, MM=1 D0 6703 I=1+NFO W=FQ(1) CALL DUMMX (Ws FOUT #NKK 9 WB 9 GW 9 MM GWG (1) =FOUT AREAG=AREAG+DW (I) #GWG(I) DO 1117 IP=1,NFO GWG (IP) =GWG (IP) #DW(IP) #26 IFCICYCLE.GT+1) GO TO 8603 COMPUTE AVERAGE FREQUENCY AND PERIOD -21- S13MS=SIGMS/AREAG WA=SQRT (SIGMS) Th2602832/WA DEFINE SLOPES OF ENVELOPE IF (ICASE.6T.2) GO TO 6 IF (RISE +6T..0) GO TO 33 TRISE=0.25#0UR TLVL=0. 33 IF (ICASE.LE«1) GO TO 7 8 FICI=1./TRISE FTC2=-1./ (DUR-TRISE-TLVL) GO T0 6 7 FICI=0.5 FTC2=04 © WRITE(69114) WAyTAsNFQsWBEGIN»WEND 114 FORMAT(//10X»*CENTRAL CIRCULAR FREQUENCY = '6F10,4,! RADIANS/SEC.! $//10X+*CENTRAL PERIOD = 'sF 8.4." SECONDS! $//10X» "NUMBER OF PHASE ANGLES 15 S//10% "LOWEST FREQUENCY IN MOTION = '9F10.5,% RAOTANS/SEC? S//10X» "HIGHEST FREQUENCY IN MOTION = '9F10.59! RANTANS/SEC.?) c c COMPUTE RANDOM PHASE ANGLES c 00 31 I=1.NFO IY=1X#65539 IF (1Y.GE.0.) GO To 32 ¥#2167483647+1 YFL=YFL® 6465661 3€-9 PA(1)=6.2832* YFL 31 IX=1Y c c ACCELERATION COMPUTATIONS c 8603 NACCG=DUR/DELT+1.000001 IF (NCYCLE«LE«ICYCLE) GO TO 9801 WRITE (649567) 9567 FORMAT (1H1,9X» *PERTOD' ¥BXs "FREQUENCY ' 9 4X9 "POW. SPEC .DEN. 5X4 S'DES «RESPONSE 4 4X» #CAL «RESPONSE "9 7X4 "DIFFERENCE ® 9X0 *TIME® 9 // WRITE (649008) ICYCLE+TQ(1) 9008 FORMAT (30X« "CYCLE NUMRER '+12y20X9 "LOWEST MODIFIEN PERTOD = to $F10.4%" SECONDS'//) 9801 00 1114 KK=]¥NACCG 1114 ACCG(KK)=0,, NEW KCHEK=1000 NEW DO 12 LM=14NFQ IF (GWG(LM).LT.0.0) WRITE (643000) GWG(LM) +LM GWG (LM) =ABS (GWG (LM) ) 3000 FORMAT (* GWG NEGATIVE. FQUALS 'y£10.3+" FOR LM OF "91S AA=SQRT (GWG (LM) ) NEW ALFA=FQ(LM) #DELT NEW SINASSIN(ALFA) NEW COSA=COS (ALFA) NEW SN=SIN(PA(LM)) NEW CN=COS (PA(LM)) NEW SNA=SINA®CN*COSA®SN CNA=COSA*CN-SINA®SN ACCG (2) =AASSNA+ACCG (2 o00 o00 c 3009 FT= 3010 ACCG (KK) =ACCG (KK) #FT 3011 CONTINUE 22+ DO 12 _KK=3»NACCG (KK .GE+KCHEK) GO To 5012 SNO=SNA| SNA=SNA#COSA*CNA®SINA CNA=CNA®COSA~SNO#SINA GO 10 12 3012 KCHEK=KCHEK+1000 SWASSIN(PA(LM) « (KK=1) @ALFA, CNA=COS (PA(LM) #(KK=1) @ALFA) 12 ACCG (KK) =AR#SNA*ACCG (KK) APPLY INTENSITY FUNTION WITH FOUR OPTION GO TO (3003+3003+3004+3007) sICASE 3003 IF(ICASE.LE.1) GO TO 18 TX=TRISE GO TO 19 18 TX=2. DEFINE MAXIMUM HEIGHTS IN TERMS OF SLOPES 19 D0 16 KK=2,NACCG TI=(KK=1) ®DELT IF (TI.GT.TX) GO TO 15 FI=FICL#TT GO TO 16 15 IFCICASE.LE.1) GO TO 28 IF ((TI=TX=TLVL) 467.04) G9 TO 29 28 FT=l. GO TO 16 29 FT=Lee (TI-TX=-TLVL) OF TC2 COMPUTE ACCELERATION 16 ACCG(KK) =ACCG (kK) #FT. GO TO 3011 3004 DO 3006 KK=2eNacce Kk=-1) #NELT O# (EXP (~ALFAO®TI) -EXP(-BETAO#TI)) 3006 ACCG (KK) =ACCG (KK) #FT GO TO 3011 3007 DO 3010 KK= 2yNACCG TI=(KK=1) #0ELT IF (TI.GE*TRISE) GO TO 3008 FT (TI/TRISE) ##1POW GO TO 3010 3008 IF ((TI-TLVL-TRISE) .LT.0+) GO TO 3009 FTSEXP (“ALFAO#(TI=TLVL)) GO TO 3010 +0 COMPUTE MAX GROUND ACC BEFORE BASELINE CORRECTION c 20 AMAXIM=0. DO 5000 I=1+NACCG IF (ABS(ACCG(T)) «LT.ABS(AMAXIM)) GO TD 5000 AMAXIM=ACCG (1) TMAXIM= (I=1) ®0ELT 5000 CONTINUE IF (NCYCLE.GT.ICYCLE) GO TO 8504 a00 3200 23+ WRITE (645200) AMAXIMsTMAXIM FORMAT(1H +//s10Xs*MAX, ACCEL, BEFORE CORRECTION? ,F12.5// st AT TIME' sF1265/7) 8504 4300 4310 TL=-DELT#0.5 JUSTIFY ACCG TO ZERO FINAL VELOCITY BETA BETAZ=0, BETA3=0, VEL=0. 00 4300 IZ=1+NACCG VEL=VEL+ACCG(1Z) *OELT T1=T1+DELT BETAI=BETAL+VEL*T) BETA2=BETA2+VEL*T1°T1 BETAS=BETA3+VEL*T1¢T1¢T1 BETAI=BETAL*DELT/(T1¢T 1811) BETA2=BETA2*OELT/(T1#T1#T1#T1) BETAS=BETAI*OELT/(T1®T1#T1®T1#T1) C1=300.*BETA1-900.*BETAZ*630, #BETA3 C2=(~1800.*BETA1 +5760. *BETAZ-4200.*BETA3)/T1 C3= (1890. *RETA1-6300. 9BFTAZ+4725,eBETA3) /(T1®T1) DO 4310 IZ=1»NACCG T1s(1Z~1) #0ELT ACCG (IZ) =ACCG(IZ)-C1-C2*TI=C3#T1#TT GET MAX GROUND ACC ACCEL) DO 59 LL=2.NACCG GAMY=ABS (ACCG(LL)) EL*ACCG(LL) ®DELT OISP=DISP+VEL*DELT DAMY=ABS (DISP) VAMY=ABS (VEL) IF (DAMY«LE.OMAX) GO To 52 OMAX=DAMY IF (VAMY«LE.VAMX) GO TO 56 VAMX=VAMY IF (GAMY.LE.GAMX) GO To 59 GAMX=GAMY LUI=LL CONTINUE NO SCALING OF THE ENTIRE TIME HISTORY IS DONE BUT PEAKS APE ADJUSTED IN ORDER TO HAVE ONLY ONE PEAK EQUAL TO THE SPECIFIED MAX GROUND ACC TTT=ABS (GAMX/AGMX) IF(TTT.LE.1.) GO TO 1112 DO 111 K1=1+NACCG DAR=ABS (ACCG(K1) ) -AGMX IF(DAR.LE.0.) GO TO 111 ACCG(K1) =ACCG(K1) /TTT CONT !NUE Go Te 1113 ° o00 nooo 1 12 3 34 a7 aaa 8901 0002 43 10000 e000 au 213 214 -24- BCCG(LL1) =ACCGILLI) /TTT GAMX=AGMX/ 386 64 LIM=NDAMP TF CICYCLE.LT.NCYCLE) LIM=1 CHECK ACCG DIMENSIONS ICK=NACCG+2.®TQINKK) /DELT IF (ICK.GE.8000) WRITE (+34) ICK FORMAT (* ACCG ARRAY NOT ENOUGH FOR NACCG+?* (LARGEST PERTOD)/0T = $ t915) IF (ICK.GE.8000) GO To 9003 RESPONSE CALCULATION AND PLOTTING CALL SPECT (PLTVMX » TMD 9 ACCG »NACCG DELT # TQ »NKK + AMOR « LIM} IF (IPCH.EQ.0) GO TO 35 WRITE (7927) ICYCLE FORMAT ('GWK FOR CYCLF #912 WRITE (79888) (GWK (NKK=T¢1) 9121 NKK FORMAT (6F1343) CONTINUE TF (NCYCLE*LE«ICYCLE) GO TO 44 CYCLING PROCEDURE WHICH MODIFIES G(W) TO SMOOTHEN THE CALCULATED RESPONSE SPECTRUM SUMPOS = 0. SUMNEG = 0. DO 43 T=1+NKK AMULT=SV(1) /PLTVMX (191) RATIOS = ABS (14/AMULT) #100. PERCEN(I) = RATIOS - 100. WRITE (648901) TQ(T) sFRO(T) eGWK (NKK=1*1) oSV(T) #PLTVMX (191) ® PERCEN(I) 6 TMO(19I) oI FORMAT (S(GX oF 1203) 94XyF1Zely! BY 9GX9F 12039110 J=NKK=Te ANEWGK (J) = GWK (J) *AMULT®AMULT AINCRM = ANEWGK (J) -GWK (J IF (AINCRM,GE.0.) SUMPOS = SUMPOS+AINCRM IF (AINCRM.LT.0.) SUMNEG = SUMNEG-AINCRM CONTINUE IF (SUMNEG,LE.1-E-8) GO TO 213 FACTOR = SUMPOS/SUMNEG WRITE (6410000) SUMPOS +SUMNEG +FACTOR FORMAT (//10Xs"SUMPOS = 9F12.3y10X"SUMNEG * 4F12.3) DO 211 1=1yNKK GWK(1) = ANEWGK (I) GO TO 100 9F12,3+10X9 "FACTOR =F OPTION THAT MAKES NO CHANGES IN POSITIVE INCREMENTS WHEN SUMNEG IS LESS THAN 1. XE -8 DO 214 T=1.NKK GWK(T) = ANFWGK(T) GO TO 100 WRITE MAXIMUM RESPONSE VALUE -25- 46, VRITE (64120) GAMX 9 VAMX » DMAX 120 FORMAT(41#//10X+"MAXIMUM GROUND ACCELERATION = "eF6o30" GtISt// 10% "MAXIMUM GROUND VELOCITY = '+F6,39" INJ/SEC.*// S10X+ "MAXIMUM GROUND DISPLACEMENT — 'eF6e39" INS t// 520Xy"SIMULATED GROUND ACCELERATION: //) DO 17 I=1+NACCG 17 ACCG(T) 2ACCG (1) /386.4 WRITE (645203) (ACCG(I) +T=1sNACCG) 5203 FORMAT (SH +15FA.4) IF (IPCH.E9.0) GO TO 36 WRITE (744111) (ACCG(T) »I=1¥NACCG) 4111 FORMAT (BF9.5) KOUNT=0 JK=0 WRITE (791) TITLE WRITE (7922) NOAMP«NKK 22 FORMAT (2110) WRITE (7913) (TQ(T) 9 T=19NKK) 13 FORMAT (10FB.4) 9101 KOUNT=KOUNT+1 JeJKrL JKEIK +8 IF (JKsLT«NACCG) GO TO 9103 JJ=NACCG*1 DO 9104 K=JJsJK 9104 ACCGIK) =040 9103 WRITE (79301) (ACCG(T) + I=J9JK) eKOUNT 301 FORMAT (8F9.5+18) IF (JK.LT.NACCG) GO TO 9101 TEMP=99,. KOUNT=KOUNT#1 WRITE (749102) TEMP »KOUNT 9102 FORMAT (F9,6963X+18) 36 CONTINUE 00 9012 LL=1+NDAMP WRITE (644535) &MOR(LL) 4535 FORMAT (1H1," DAMPING 1 7X+ "RESPONSE # 6X IF (IPCH.EQ.0) GO TO 37 CAM=AMORILL) ® 100. WRITE (759016) CAM 9016 FORMAT ("DAMPING '4F4,19" PER CENT!) WRITE (798a8) (PLTVMX(LL9N) #N=1 ¢NKK) 9015 FORMAT (10F 8.4) 37 CONTINUE 9012 WRITE (694340) ¢(TQ(KK) ¢FRO(KK) sPLTVMX (LL 9KK) ¢ TMD (LL KK) 9KK) 9 S$ KK=1+NKK) IF _(NRES.EQ.0) GO TO 100 WRITE (659567) DO 23 I=1+NKK AMULT=SV (1) /PLTVMX (191) RATIOS = ABS (1./AMULT) #100. PERCEN(T) = RATIOS ~ 100. 23 WRITE (648901) TQ(T) sFRO(L) oGWK (NKK=1¢1) ¢SV(T) sPLTVMK (191) 4 ® PERCEN(I) «TMD (Lol) oI 4340 FORMAT(! 144F 14649110) 00 21 TI=1¥NDAMP DO 21 JJ=14NKK 2. PLTVMX (ITs JJ) =ABS (PLTVMX (IT 9 JJ)) NFC=? 9F663///4 9Xs "PERIOD! 6X0 FREQUENCY? » TIME?//) -26- DO 1000 IT=1+NDAMP re 001 LTVMX (TT 9 J) MOR (TI) CALL DIB2 (NFC 449140 oNKKsTS+ TL oYMING YMAX leo s 9009090 9-25~2. STO+SL+SVoTIX+TITX TITY» 369 369360040. 9XAMOR. 200 CONTINUE 190 CONTINUE 38S CONTINUE IF (NKK.«GT.0)GOTO1100 400 CALL PLTND(KIKI) CALL EXIT END SUBROUTINE DUMMY (WsFOUT» VKK sWB »GWK #MM DIMENSION WB(1) GW (1) JAY=MM 1 IF (W=WB(JAY)) So4e02 2 JAY=JAY+1 IF (JAY.LE.NKK) GO TO 1 FOUT=GWK (NKK) GO TO 6 4 FOUT=GWK (Jay) MM=JAY GO TO 6 5 MM=JAY=1 IF (MM.LE.0) GO TO & SLOPE= (GWK (JAY) ~GWK (JAY=1)) / (WA (JAY) “WB (JAY=1)) FOUT=GWK (JAY=1) *SLOPE® (Ww (JAY=1) ) 6 CONTINUE RETURN END SUBROUTINE SPECT (VMAX +TAyGAyNyDEL »PO+ IP s0MPy ID) 2901 c SUBROUTINE FOR COMPUTATION OF SPECTRA FROM EARTHQUAKE RECORD C DIGITIZED AT FQUAL TIME INTERVALS c DIMENSION VMAX (104300) +TA(10+300) +A (6001) PD (300) ¥OMP(10)+ 1 A (242) 9B (292) TY (3) 9X (3) 9G(2 00 6 J=1sI0 D=DMP (J) DO 6 K=1yIP P=PD(K) IF (P.LT.0,001) P=0.001 W=6.2831854/P c C CHOICE OF INTERVAL CF INTEGRATION c DELP=P/10. L=DEL/DELP+1.-1.E-5 DELT=DEL/L. c © COMPUTATION OF MATRICES A AND 3 c CALL PCNO4 (D4WsDELT+ A+B) c C INITIATION c xa aa0 ona -27~ COMPUTATION OF RESPONSE 1 SL=(GA(T#1)-GA(T))7 BO 5S M=1sL. GUD= GAIT) +SL#(M=1) G(2)= GALT) +SLeM TY (CL) SAC 101) 9X (1) 4A (192) 8X (2) B19 1) G1) 8102) #62 TY (2) =A (21) 9X (1) +A (242) X (2) -B (241) 9G 1) -B (292) 962 LISL1+1 TIME=(L1-1) *DELT MONITORING THE MAX. VALUES IF (ABS (TY(1)) .LE*ABS(DMAX)) GO TO 2 DMAX=TY (1) TO=TIME 2 xaDeTya) 5 x(2)=TY(2) TEST FOR END OF INTEGRATION I=T4} IF (1,£0.N) GO TO 7 Go TO 8 7 VEND=X(2) AIF (1.EQ.(N*TA)) GO TO 10 IF (1,GE.N) GO TO 9 Go TO 1 9 GA(T#1)=0. Go TO 1 10 CONTINUE VMAX (Jo) =W#OMAX TA(J9K) =TO 6 CONTINUE RETURN END SUBROUTINE DIB2 (NFC+IND#NGRAPH»NGD»NPOINT oXL oXRo¥BsYTyDX 409 SNoMy To JgNXoNY 9X9 9ZeTI Ts TI TX TITY NT #NTX#NTY @NPT 9 PTMRK » XAMOR) DIMENSION X(1) o¥ (1) 9Z(1) TIT (1) oTITX (1) TITY (1) @PTMRK (1) IND. GO TO (1924344) ¥IND 1 CALL SMKYV (040) Go TOS 2 CALL SMXYV(091) Go TOS 3 CALL SMXYV(1+0) Go TOS 4 CALL SMXYV (141) 5 CONTINUE CALL SETMIV(150+100+1504150) IF(NFC=1) 11910920 10 NFAS2 GO TO 30 20 NFA=4 ae n 400 100 103 101 104 105 110 200 205 203 201 204 210 303 300 301 302 401 403 500 402 -28- CALL GRIDLV(NFAs XL ¢XRy¥B9YT 90x sD #NoMe Te J9NX ONY) CALL RITEZV(125+250100099092yNTY 919 TITY »NLAST. CALL RITE2V(3000125+100090+2eNTX ele TITX+NLAST) CALL RITE2V(250+925+10009092eNTo ly TIT *NLAST CALL LABLV (XAMOR+7509AA0 969101) CALL INCRV (894) NAU=NGRAPH+NGD IF(NAU) 40144014400 00 7 IT=1+NAU NAUX=NPOINT=1 DO 8 K=1+NAUX TAUX= (TI=1) #NPOINT#K «KD 1K) (KeD) (ke1) (TAUX) Y2=Y (TAUX# 1) IF(Y1-YT) 10091009101 IF(Y2-YT) 11091109103 X2= (X2=X1) 8 (YTHAYL)/ (¥22¥1) OX y2=YT Go TO 110 IF(Y2=¥T) 10491049105 X12 (X22X1) 8 (YTHYL)Z(Y22¥1) 4X1 y1syYT Go TO 110 INDA=1 CONTINUE IF (Y¥1-Y8) 700%201+201 IF (Y2-YR) 20592034203 INDA=1 Go TO 210 X15 (X29X1) 8 (YBOY LIZ (Y22¥1) 4X1 Y1=B GO TO 210 IF (Y2=YB) 20492109210 X22 (X2=K1) 8 (YBAYL) /(Y22¥1) #X1 y2=¥8 CONTINUE IF(INDA) 303%303+302 IF (LI-NGRAPH) 30043004301 CALL LINEV(NXV (X1) @NYV (YL) #NXV(X2) #NYV(Y2) ) CALL LINEV(NXV(X1) »NYV (YI) ¢NXV(X2) @NYV(Y2) ) CALL DOTLNV (NXV(X1) »NYV(Z1) oNXV(X2) #NYV(Z2)) GO TO 302 CALL DOTLNV(NXV(X1) pNYV (YL) 9NXV(X2) @NYV(Y2) ) CALL DOTLNV (NXV(X1) ¢NYV (YL) #NXV(X2) #NYV(Y2) ) INDA=0 CONTINUE CONTINUE IF(NPT) 4024402+403 LL=NPOINT#NPT 00 500 I=1+NPOINT CALL APLOTV(LL4X(1) s¥ (I) #0 NPOINT #NPT #PTMRK TERR. CALL APLOTV(LL +X (I) 9¥ (1) #09NPOINT pNPT #PTMRK » IERR RETURN ENO SUBROUTINE PCNOS(DyWsDELT+As8) -29- C SUBROUTINE FOR COMPUTATION OF MATRICES & AND B c DIMENSION A(2+2) 0B(292) XP (-DWSDELT) Al=WeSQRT(1.-D2) AD1=A1®DELT A2=SIN(AD1) OS (AD) 282 2602-1.) /w2 7 PAS /W2 AT=1 6/02 AB=(A1®A3-DW#AZ) #00 AQ==(A1#A2+0W#A3) AO ALO=AB/A1 All=A0/a1 Al2=A1 1942 AL3=A08R3 AlG=A108AG AIS=A12 #06 AL6=A6%A13, ALT=A9#A6 A(141)=A08 (DW#A27A1 +43) A(L2)=A12 A(291)=A10@DWeAd A(242)=A10 B (161) =(~A15-A16+A6) /DELT-A12#AS-A7#A13, B(192)=(A15*A1 6-86) /DELT*AT B(291)=(-A14-A17-A7) /DELT~A10®AS=A9#A7 B(292)=(A14+A17%A7) /DELT RETURN END SUBROUTINE DUMMX (W+FOUT » NKK » WB 9 GWK » MM: DIMENSION WA(1) +GWK (1 JAY=MM 1 TF (WeWB(JAY)) S902 2 JAY=JAYe] IF (JAY,LE.NKK) GO TO 1 FOUT=GWK (NKK) GO T0 6 Deeper 5 MM=JAY=1 IF (MM.LE.0) GO TO 4 = (WB (JAY) #WB (JAY=1) 1/24 IF (W=X) 79798 7 FOUT=GWK (JaY=1) GO TO 6 ® FOUT=GWK (JAY) 6 CONTINUE RETURN ENO SUBROUTINE POLATE (NyMeXIVsYINyXOUTyYOUT DIMENSION XIN(1) eYIN(1) ¢XOUT(1) »YOUT (1) J=1 IF (XIN(1)=XOUT (19) 2924100 40 30 100 20 ~30- IF UCIN(N) =XOUT(M)) 1009393, Do 30 a] TF (KOUT (1) -XIN(J)) 594004 JeJel GO 10 6 JeJo1 YTEST=(ALOG (YIN(J#1) ) ALG (YIN (J) ))®(ALOG (XOUT (1) )=ALOG (XTN(J) 7 $(ALOG (XIN (Jo 1) ) -ALOG (XIN (J) )) #ALOGCYIN(J)) YOUT (1) =EXP(YTEST) GO TO 30 YOUT (1) =YIN(J) CONTINUE RETURN WRITE (6520) FORMAT (1H1, "PROGRAM STOP*FUNCTION UNDEFINED IN DESIRED INTERVAL*) CALL EXIT END SUBROUTINE PLTX2 (XMIN¢XMAX +X yNPOINT) DIMENSION x(1) POINT=NPOINT=1 SPACE=ALOG10 (XMAX/XMIN) /POINT X (1) =XMIN DO 1 T=2+NPOINT AlsI-1 EXPO=SPACE#AI X(T) =XMIN®10.8#EXPO X (NPOINT) =XMAX RETURN END SUBROUTINE GWPLOT (NKK + TS9TL ¢GMINsGMAX 9 TOs GWy TITXo TITLOSYTITL, DIMENSION TO(1) oGW(1) oTETX(L) oTITLOCL) o¥TITL CL) IF (GMAX«LE.70.0) GO TO 3 IF (GMAX«LF.200.0) GO TO 2 OY=20.0 CONTINUE ESTABLISH SEMILOG COORDINATES CALL SMXYV (140) ESTABLISH MARGINS CALL SETMIV(150+100+150+150) ESTABLISH GRID CALL GRIDIV(16TS+TLsGMINsGMAX41409D¥ 909590 959=-24=7) WRITE Y AXTS LABEL CALL_RITE2V(125+250+100099092+2A1¢YTITL#NLAST) WRITE X AXTS LABEL CALL_RITE2V(300+125410009092936s1eTITX+NLAST) WRITE TITLE CALL RITE2V(250+925+1000909242R91 + TITLOyNLAST JOIN POINTS WITH STRAIGHT LINES. IT=NKK+1-T y1=GW(TT) W (NKK=T) XXL) -31- YV CYL) NV (X2) LY2=NYV(Y2) CALL LINEV(IXL4IY1 91X29 1Y2: 4 CONTINUE RETURN END SUBROUTINE SVGW(NKK »WsGW*SV9S9ByWC+Q9P¥XLAMO pWCP2QP9RR) DIMENSION GW(1) #W(1) SV(1) #RR(L) 214159 2831852 NWENKK=To1 POW=2.*B8W(T)#S IF (POW.GT.5040) GO TO 610 TRANS=1,-EXP (-POW) Go TO 611 610 TRANS=1. 611 BS=B/TRANS WCYS=W(T) QYS=SQRT(4,0#8S/P1) =WCYS#S/ (PI2®AL0G(P)) SORT (24 *ALOG(24®XS9)) RSTAR®QYS*SORT (PI/2) 2.°xXSP®(1.-EXP(ET)) SP. 4 ®RS/ (WT) #PT)) ® (SV (NW) #W (1) /RSP) 882-GSUM: IF (GW(T) «LE 40401) GW(T) =0001 IF (1.6T.1)G0 TO 140 GSUM=0..58W(1) 9GW01) GO TO 1000 140 GSUM=GSUMsGW(T) #(W(T) =W(I=1)) 1000 CONTINUE wCP=0.0 aP=0.0 Gw(T) +GW(T-1))/2. OUMY=W(T)

Вам также может понравиться