Download as pdf
Download as pdf
You are on page 1of 38
EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS, VOL. 15, $3-72 (1987) MULTIPLY SUPPORTED SECONDARY SYSTEMS PART I: RESPONSE SPECTRUM ANALYSIS R.A. BURDISSO* AND M. P. SINGI Deparonent of Engineering Science and Mechanics, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061, U.S.A, SUMMARY. ‘A response spectrum procedure is developed for seismic analysis of multiply supported secondary systems. The formulation is based on the random vibration analysis of structural systems subjected to correlated inputs applied at several supports. For a proper response spectrum analysis of a multiple support system, the support inputs are required to be defined in terms of the auto and cross pseudo-acceleration and relative velocity floor response spectra. Also information about the floor displacements and velocities as well as their correlationsis required. The response of the secondary aystem is expressed as a combination of the dynamic and pseudo-static response components. The dynamic component is associated ‘with the inertial effects of the support accelerations, whereas the pseudo-static component is due to the displacement of the supports relative to each other. Herein, the correlation between these two parts of the response is included through a term called the cross response component. Each of these components of the response can be calculated by a response spectrum method. The application of the proposed method is demonstrated by numerical examples. INTRODUCTION For seismic design of important industrial facilities, the earthquake input loading is often prescribed in terms of the ground response spectra.!~? For analysis of the primary systems, rational analytical methods which can directly use the ground spectra as inputs have already been developed. However, for analysis of the secondary systems, especially the systems with multiple supports (such as piping attached at several points of a main structure), the methods employing spectra as inputs arcin the development stage, although significant progress has been made in understanding the dynamic behaviour of these systems as a result of the continued research interest. See the list of references. In the current industrial practice, the time history and the single floor response spectrum methods are commonly used to analyse such multiply supported secondary systems. The time history method, though analytically most accurate for a given earthquake motion time history, does not provide unique results suitable for a design. To obtain the design response, that is the response which can be used for the design of these subsystems, one must consider a set of time histories as inputs in the analysis. This will, however, requirea large computational effort and thus may not be economically feasible. The response spectrum method employing the design spectra as the inputs directly in the analysis, on the other hand, is computationally inexpensive. Currently in this approach, it is a common practice to use the envelope of all the support point floor spectra as the inputs to the fixed base model of the secondary systems. To account for the effect of the relative displacements between the support, some ad hoc and approximate methods* ate employed. These methods, however, lack analytical rationale, do not account for the correlation between the support motions properly and may, sometimes, give overly conservative results. In this paper a comprehensive response spectrum approach is developed to overcome some of these aforementioned shortcomings. The method is based on random vibration analysis of the multiply connected secondary systems subjected to correlated random excitations at the supports. Random vibration analyses of, * Graduate Research Assistant + Professor. (0098-8847 /87/010053~20$10.00 Received 6 August 1985 © 1987 by John Wiley & Sons, Ltd. Revised 7 February 1986 34 R.A. BURDISSO AND M. P. SINGH secondary systems for inputs defined in terms of power spectral density function have also been described by Lee and Penzien,? Gasparini et al.® and Singh.’ The approach described here, though based on random vibration principles, does not require any random characterization of the support inputs; rather, the support inputs defined in terms of the so-called auto and cross floor response spectra-are used directly. These floor spectra can be obtained directly from the ground response spectra; the methods for this are developed in the companion paper.* The use of these floor inputs in the response spectrum approach developed herein is demonstrated on two examples of the multiply connected secondary systems. EQUATIONS OF MOTION ‘The equations of motion of a secondary system attached at several points of a primary structure, considered as a free-free system, can be written as*? M, My |{U, Cy Ca |SUs| [Ks Ke |fUs fo 1) we welled e[e alo+Le ected -fo where the subscript ‘a’ is associated with the degrees-of-freedom of the support points ands’ with the degrees- of-freedom of the active or unattached mass points of the secondary system. The displacement vectors {U,} and {U,}, respectively, denote the absolute displacements of the unattached points and the support points of the structure. The vector {U,} is of dimension n, equal to the degrees-of-freedom of the unattached masses, and {U,} is of dimension m, equal to the degrees-of-freedom of the support points or the masses on the primary structure. A dot over a time dependent vector quantity denotes its time derivative. [MJ [C,.] and [K,,], respectively, are the mass, damping and stiffness matrices associated with the active degrees of freedom, and thus are of dimension n x n; similarly [My.],[ Ca] and [ Ky], which are of dimension m xm, are the mass, damping and stiffness matrices associated with the support points, The other sub-matrices in equation (1) introduce the coupling effects which exists between the supports and active degrees-of-freedom through the inertial, damping and elastic forces. By taking the right hand side of equation (1) to be zero, we imply that the support points are connected to the rest of the primary structure by springs of zero stiffness. Thus, no force is transmitted between the rest of th primary structure and the support points to which the secondary system is attached. That is, the motion is assumed to propagate only in one direction, from the primary to the secondary structure and not backwards. Such systems are also called systems in cascade.® This neglects the dynamic interaction and feed-back from the secondary system to its primary system in the analysis. The assumption that the two systems are decoupled greatly facilitates the analysis and, therefore, it has been used quite commonly in practice:*~7-10-16 In many practical situations this assumption can also be justified. Yet, however, there are cases where the dynamic interaction between the two systems must be considered. Several approximate analytical procedures have been proposed to include this dynamic interaction. See References 17-22. In this paper, however, the analytical development of a response spectrum approach only for the cascaded secondary systems is presented. In this formulation, the total response is partitioned into the pseudo-static and dynamic components. A similar partitioning of the response wasalso utilized by Lee and Penzien.* Asfura and Det Kiureghian,'? on the other hand, have adopted the absolute displacement formulation without any partitioning. Their formulation, however, cannot be used to calculate the forces in the members of the secondary system which are attached to the supports. These forces depend on the relative movements of the supports with respect to each other as well as with respect to the attached secondary system masses. The formulation proposed in Reference 19, however, has no provision to calculate these relative motions as well as the forces associated with them. As itis necessary to know these forces for design ofa secondary system, the absolute displacement formulation as developed by Asfura and Der Kiureghian’” is inadequate. ‘The pseudo-static component of response is due to the relative displacement between the supports, without any dynamic influence. The dynamic component comes from the inertial forces induced in the unattached ‘masses due to the support accelerations. Obviously, the dynamic response of the attached degrees of freedom is MULTIPLY SUPPORTED SECONDARY SYSTEMS. PART I 55 zero. Thus, we write {U0} = (UFO}+ {UL} @ {U,(0} = {UPO} @) where {U‘} and {UP}, respectively, are the dynamic and the pseudo-static components of the active degrees of freedom and {U,} are the support motion time histories. Substituting equations (2) and (3) into (1), the set of equations associated with the dynamic component of response can be written as [Ma] (08+ OP} + [Cu] (Os-+ UF} + [Ke] (Us + UF} ~ [Mg] (U,}— [Cu] {Ua} ~ [Kea] (Ua @ Since the pseudo-static response does not include any dynamic effect, it can be obtained by eliminating the forces associated with the mass and damping matrix in equation (1). That is, K, K, UP ‘0 oe Kg) {U2 5 [es xl {cct = oh ® From equation (5), the pseudo-static response of the unattached degrees-of-freedom can be written in terms of the prescribed support inputs as (UB = (“[Kal* KD (0) = LAH.) © where matrix [4] iscalled the pseudo-static influence matrix. Substituting equation (6) into (4)and rearranging, terms, we obtain [Mo] {08} + [C.c] (U8) + [Kae] {U8} = (EMuc] [Ku]? [Kal - [Mal (0, +([Cu] [Kes] * [Kua - [Cus] {05} @ If the damping matrix is assumed proportional to the stiffness matrix, the terms dependent on {U,} in equation (7) vanish. In a more general case these terms will not be zero. However, since these terms are associated with damping terms, their magnitude compared to the other terms will be relatively small; and thus they can be neglected. With these assumptions, the equations of motions associated with the dynamic response can, then, be expressed as [Mg] {08} + [Cy] {08} + [Ky] {U3} = [71 {04} @) where [7] = ((M..] [Kg]! [Ka] -[Mq]) is a dynamic influence matrix in which each column represents the distribution of force in the unattached degrees-of-freedom due to the acceleration of each support. DESIGN RESPONSE ‘A response quantity, S(Q, linearly related to the displacement response can also be expressed as a sum of the dynamic and pseudo-static response component as SO) = S*(H+S"0) 0 where the dynamic component, S°, corresponds to the vector {U¢} and the pseudo-static component, 5?, corresponds to the vector, {UP}. Often, the stresses caused by the pseudo-static component are considered as the secondary stresses for which higher allowable stress intensity is permitted by the ASME Boiler and Pressure Vessel Code.* Here in this analysis, these stresses can be explicitly obtained. To consider the ensemble of all possible ground motions in the analysis, here the site motions are represented bya zero mean random process, The maximum response or the design response due to such random motions 36 R.A. BURDISSO AND M. P. SINGH can be expressed as an appropriately amplified value of the-root mean square response Ry= Ce, (19) maximum response, ¢, = root mean square value of response S(t) and Cy = the peak factor by which o, is amplified to give the design response. To obtain o,, the covariance function of the response S(t), defined by equation (9) is first obtained, Reltysta) = ELS) S"(ta)] + ELS") S°(t)] + B[S*(t,)5P(t2)] + ELS") S*()] ay In equation (11) the first two terms represent the contribution of the dynamicand pseudo-static components, to the total response, while the last two terms account for the cross correlation between them, The variance of the response is obtained by setting ¢, = r, in equation (11), and then the design response can be written as follows: Rh= Ci(chgt 02+ 2C yp) (12) where 03, = variance due to the dynamic component, 02, = variance due to the pseudo-static component and Cop = the cross-covariance between the dynamic and pseudo-static components. Herein the three terms in equation (12) are referred to as the dynamic, pseudo-static and cross response components, For design purposes, the earthquake motions are usually prescribed in terms of the ground response spectra for the primary structures and in terms of the floor response spectra for the secondary system. It is, thus, desired to evaluate R, in terms of such response spectra. Herein, therefore, the response spectrum methods employing the ground and floor response spectra as inputs, are developed for calculating the contributions of the dynamic, pseudo-static and cross response components. DYNAMIC RESPONSE CONTRIBUTION In equation (12) the contribution of the dynamic response to the total response, here denoted by Rgg, is Ria = Cita (13) To obtain oy, the solution of equation (8) is required. Employing the modal analysis approach and standard manipulations involving orthogonal properties of the fixed-base normal modes, a decoupled modal equation (for a classically damped system) can be written as 40+ 2B 40+ oF 4,(0) = {P}™O,} (4) where q, = jth principal coordinates; w, = jth modal frequency, B, = the modal damping ratio, and {y;} = jth modal shape vector. {P,} = the influence vector = {y,)"[r]; each element of this vector represents the contribution of a support motion to the response in the jth mode. A superscript T over a vector quantity represents its transpose. Here we have assumed that the secondary system is classically damped. However, analysis can also be made for a non-classically damped system, as indicated by Singh.?> In terms of q,any component of the displacement vector, or a response quantity of interest which is linearly related to the displacement, can be obtained as sto= ¥ palo as) where p is the so-called modal response in the jth mode; t can be evaluated from the displacement mode shape by a simple linear transformation. For given support motion time histories, equation (14) can be solved to define q, which in turn can be used to obtain the time history of response S(t) by equation (15). However, for random site motions, the motions of the support points of the secondary system will also be random processes. To simplify the analysis for such random inputs, we assume that the ground motion, the motions of the support points defined by {U,} as wellas, the induced dynamic response of the secondary systems are stationary random processes. Although these MULTIPLY SUPPORTED SECONDARY SYSTEMS. PARTI 7 assumptions are not strictly valid for earthquake type of ground motions and responses, they have been found to be acceptable in the calculation of primary system response and in generation of floor response spectra.**"?* With these assumptions, the stationary value of the variance of the dynamic response can be shown to be as, follows:?* cu= YY pe DY PaPa { © (0) Ht do (16) i wie Jw in which P,, is the kth component of the modal force influence vector { P,}; H, is the frequency response function defined as x, ‘An asterisk over a frequency response function denotes its complex conjugate and i is the imaginary number TM AW Lora} + 48,80, at . +2iwo,0,(0,f,—0,f,) ]HHI® (o) ® os X(0)= (of +48F020") ® We now resolve HH? into its real and imaginary parts as HH = (N(o)+ioM(@)} |H,? |H,? (10) where N() and M(q) are defined as N(@) = 08 + 0(48,f,0,0,— © ~ 02) + 020? (Ila) M(o) = 2{0,0,8,0, ~ B,,) + 0 (Bo, — Byo)} (1b) Substituting equations (10) into (8) and separating the real and imaginary parts, we obtain O40) = (5 s2v,lOHOX CH, + 3, E sarah ADT) +i02(0)] 1° A Jaro a2) where ‘ Tle) = (@20? +4,8,02,0,07)N(@) ~20,048,00,— B00? Mo) (13) 2(e) = 206040, BN) + (0303 + 4B,B0.0427)M(O) (14) Equation (12) defines the real and imaginary parts of the cross spectral density functions as O50) = ¥ Fv RWADX (0), PO(0) + XY Lr lovs Ore? Polo) 5) and O40) = LY wrlb.oZ(o)[H,? [HP bo) (16) 16 M. P. SINGH AND R. A. BURDISSO For k = I, equation (15) provides the auto spectral density function of the kth floor. In equations (15) and (16), and also quite often in the paper later, we need to split the terms containing the products of the frequency response functions such as |#,|? and |H, ? into their linear sum by partial fractioning as follows: T(o)]H,) |H,P = (4 +0" BH, + (Cy+*D,)|HP (17) Z(o)|H,} |? = (4, + o*B,)|H,P + (Cy +0*D,)|H,P (a7) Effeeting such reformations in equations (15) and (16), we obtain O50) = YA OX)|H,PO lo) monS + DY vera Ww (OAs + 07 Bs) |, + (Cy + 0*D5)|H,? J0,(0) (1s) is! BS rhb 0ldato* Baer +(Cy+@°D,)|H,|?] Oo) (19) where the coefficients ofthe partial fractions, Aj-D, and A,~D,,in equations (17)-(19)are given in Appendix I The real and imaginary parts of the spectral density function expressed in equations (18)and (19) will now be used to develop the methods for generation of different types of the interaction free floor response spectra, identified in the companion paper." AUTO FLOOR SPECTRAL INPUTS. Two kinds of auto floor spectra are required: (1) auto displacement and (2) auto (relative) velocity respons spectra. Customarily, auto pseudo-acceleration spectra are more commonly used in lieu of the displacement spectra; these two, however, provide the same information in different forms. The floor spectra provide the spectral characteristics ofa floor acceleration, and can be expressed in terms of the auto power spectral density, equation (15), as follows: R4(o) = RoYot = C3, {" Oy(o)|HPdeo 20) Rio) = Ci |” 07 Qo) |H 40 en where Ryy(0o), Ryy(o) and R,(c), respectively, are the displacement, pseudo-acceleration and velocity floor response spectra for an oscillator with frequency «, and damping ratio f, placed on floor k;and C,,and C,,are the peak factors for the displacement and velocity responses of the oscillator, respectively. For a given floor spectral density function these spectral quantities can be computed. We would, however, like to define these in terms of the ground response spectra. For this an approach developed by Singh!® for generation of the absolute acceleration response spectra will be used here for generation of the displacement and velocity respons spectra. Since the basic steps are the same, the detailed development will be avoided ‘On substitution of equation (18) into equation (20) and realizing that for auto spectra k = I, we obtain aifoy~ ca [” (SE nmexee ines SF rots feet x [Us +0?B)|17,P-+ (Cy t0"D) |H,P}]H PO (c)}d0 @ MULTIPLY SUPPORTED SECONDARY SYSTEMS. PART I 7 “To express this in terms of the ground response spectra rather than the ground spectral density function, we need to split the terms containing the products of the frequency response functions such as |H, |? and| H,|? into their sum by partial fractions, as was done in equation (17). Eifecting these changes in equation (22) we obtain Ro) = caf XY PWR W[Aslig + Bsloy + Csligit Dslog] + YY wrath lb) CAclig+ Bolog + Coligrt Delay] + [Agligst+ Baloge + Coligit Dstaed} 23) where the coefficients of the pattial fractions A,-Ds, Ag-Dg and A,~D, are defined in Appendix I. The two frequency integrals Iyy, and I7,, involved in this expression, respectively, are the mean square values of the (elative) displacement and velocity responses of an oscillator on ground. These can be expressed in terms of the pseudo-acceleration and relative velocity ground response spectra and the associated peak factors as follows: Ral) OFC pg)? (24a) ©,(c)|H,P? [R,(,)/CyT? (24b) = { * (20(0)|H, where R,,(c,) and R,,(«,), respectively, are the pseudo-acceleration and relative velocity ground response spectrum values at the oscillator parameters, «2, and f,; and C., and Cy», respectively, are the displacement and velocity response peak factors of the oscillator. Equations (23) and (24) define the auto displacement floor spectrum value in terms of the ground response spectrum values. ‘The case when H, or H, is identically equal to H, is referred to as the resonance case. In such a case itis not possible to define the coefficients of the partial fractions in equation (23). This case can also be treated as described by Singh.'° The frequency integrals for this case can be expressed in terms of the following integral: Inlay, a3) 5 f (a8 + 4,080) |H,|*O,(o)deo es) which, for peaked |H,|*, can also be expressed as? Taldys G2) = Tyg F[F(@,)C+ a, {1 — F(o)}] + 20070, (26) where F(o,), C,, and Dj, are defined in Appendix I. For H, = H,, the frequency integral associated with the product of H,, and H, in equation (22)can be defined in term of I as follows: rgop= [ X@|H $060 = Art 42) en (A, +0°B,)|H, PO (o)do . ae htAslo?, Ba) es) ‘The terms for which H, # H, remain the same as those in equation (23), Employing equation (15) in equation (21) and proceeding similarly, the expression for the auto relative velocity floor spectrum can also be defined in terms of the frequency integrals I,,, and Jy, or the ground 8B M.P. SINGH AND R. A, BURDISSO response spectra as follows: Bulag +Calhy + Dag] Riylo) = C23 12 EOLA +X Leal, Aatig + Balog + Colag + Doo ta + Ayghigrt Broly Crolig-+ Diol) 9) where the coefficients of partial fractions involved in equation (29) are also defined in Appendix I. For the resonance case, when H, is identically equal to H,, the frequency integrals involved in the velocity spectrum expression can be obtained as a special case of the following integral: Isfaa, a3) = | (ao8o? + a,ofo|H,"®,(o) G0) which can be computed in terms of I, and Izy, and thus the ground response spectra as Ts(03, g) = Ty g(0f F(O,) Fg + 43007 LagyDoy (31) The expressions for F(o), F, and Dj, are given in Appendix II. The frequency integrals required in the resonance case are of the following form and can be defined in terms of J, as con [ 0X (0)|H,!O,(o)do = +, Ig, 482) (32) Iss(0o,) f (Ay + OBS H,|*©,(o)do = a To} Ay, Bs) G3) ‘The terms for which H,# H, remain the same as those in equation (29) Equations (23) and (29) have been used to develop pseudo-acceleration and relative velocity floor response spectra for the primary structure shown in Figure 1. This structure is the same as the one used in the companion paper.' The modal damping ratio for this structure is assumed to be S per cent in all modes. The seismic input to this primary structure is defined by the averaged spectra of an ensemble of 75 ground acceleration time +975 K/ft Figure 1. A five deprees-of-reedom primary structure y MULTIPLY SUPPORTED SECONDARY SYSTEMS. PART IL 9 histories. These spectra are shown in Figures 2 and 3. Use of these spectra as inputs in the development of floor response spectra requires that various peak factors involved in equations (23), (24) and (29) be also known, Evaluation of these peak factors is, however, not a straightforward task."" One needs to define the input in a stochastic form such as a spectral density function, Such descriptions of the input are usually not available. Some simplified assumptions can be made about the input to get an estimate of the peak factors. Sometimes descriptions of earthquakes in terms of the well-known Kanai-Tajimi or white noise spectral density functions are used to estimate these peak factors. It has, however, been observed" that incorporation of these estimated peak factors in definitions of floor spectra does not necessarily improve the accuracy of the results when ‘compared with the simulation results. Also since these peak factors do not differ significantly from each other, assuming them to be equal simplifies the analysis and gives acceptable results in comparison with the simulation results. Thus, in the process of development of the various types of floor response spectra ‘numerically in this paper, the involved peak factors have been assumed to be equal. This renders equations (23) and (29), and also the expressions for all other floor inputs, independent of the peak factors. Also, see the discussion in the companion paper.' The pseudo-acceleration and relative velocity spectra obtained from equations (23) and (24) for floor 3 for damping ratios of 0-01, 0-02 and 0-05 are shown in Figures 4 and 5. For analysis of a multiply connected secondary system, such floor spectra must be developed for each floor where the secondary system is attached. GENERATION OF CROSS FLOOR SPECTRA Tocharacterize the cross correlation between the accelerations of any two floors spectrally, we introduced! the coincident and quadrature displacement and velocity floor spectra. We will now develop the response spectrum methods, utilizing the ground spectra directly, for generation of these coincident and quadrature floor response spectral quantities. FI/Sx2 9,00 6,00 PSEUDO-ACCEL. 3,00 0,00 0.02 out PERIOD (SEC) Figure 2. Input pseudo-accsleration ground response spectra 80 M. P. SINGH AND R.A, BURDISSO. . 1] 1 8 | 2 IN Ke IN Ti Rie ILA é T s| dew | 0.02 Ot 1 PERIOD (SEC) Figure 3. Input relative velocity ground response spectra 35. 28.00 FI/Sx2 21.00 14.00 PSEUDO-ACCEL. 7,00 0,00 0.02 ot ib 5. PERICD (SEC) Figure 4. Auto pseudo-acceleration floor spectra for Noor 3 MULTIPLY SUPPORTED SECONDARY SYSTEMS. PART Il 81 FT/S 4,00 3,00 05 2 RELATIVE VEL, 1,00 nl 1 ‘et T 0.02 Ot 1 5. PERID (SEC) Figure 5. Auto relative velocity floor spectra for floor 3 0,00 Coincident displacement spectrum ‘The coincident displacement spectrum characterizes the in-phase or the coincident components of the floor motions. As described in the companion paper,’ it is defined in terms of the real part of the cross spectral density function, equation (15), at frequency ©; and damping ratio ;, as follows: Choo a(S reds. X(@|HH,P|HPOfo)do + x L rahe WO ie [(43 + 0° Bs)|H,/? + (Cy+0*D)|H,P IH, 20,(0) eo) 4) As we obtained equation (23) from equation (22), we can also express equation (34) in terms of the frequency integrals of equation (24) as follows: Chula) = a E WOM Al g + Bslag + Cslig + Dslog] + ED anal DLAcligr+ Belay + Coliy + Do las $F Aalyggt Bylagst Calg Det) (35) ‘The frequency integrals involved in the resonance case here are the same as in equations (27) and (28). 82 M. P, SINGH AND R.A. BURDISSO ‘The coincident pseudo-acceleration spectrum can now be defined in terms of the displacement spectrum as Cul) = @FCR A) (36) Figure 6 shows the coincident pseudo-acceleration spectra for floors 2 and 3 of the structure in Figure 1 obtained for three different damping ratios of O01, 02 and 0-05, Again, these spectra must be developed for each pair of floors connected through the secondary systems. Coincident velocity spectra ‘The coincident velocity spectrum at frequency «, and damping ratio f, is defined as Chuo) = raf @§,(0)|H,P do G7) where P,:is the peak factor of the velocity response. Substituting for @,() from equation (18), we obtain an equation exactly similar to equation (34), except that the integrands of each integral will now be multiplied by o*. We further break this integrand into partial fractions. The frequency integrals and partial fractions involved in this expression are the same as those involved in the calculation of the auto velocity floor spectra. ‘Thus the expression for this spectrum is essentially the same as equation (29) except that here we are concerned with two floors. With a proper substitution for the modal displacements of the two floors, we obtain ”( D rev (OW OLAslige + Belay, + Ce ligy + Ds Lag] + ZX, 2 neh ADLAcl y+ Bolag+ Coligr+ Dolo: Chulod + Alig Brolag'# Cioligr+ Diop) (G8) tis noted that for k= |, the coincident and auto spectra are the same. Thus the same algorithms can be used to compute the two spectra. By TW T mal T TT] I Sx2 a 0 0 F 1 5,00 PSEUDO-ACCEL -8.00 Ui 0.02 on 1 5. PERIOD (SEC) Figure 6. Coincident psevdo-acceleration floor spectra for floors 2 and 3 MULTIPLY SUPPORTED SECONDARY SYSTEMS, PART II 83 Using this approach, the coincident velocity spectra were obtained for floors 2 and 3 of the structure shown in Figure 1. These are shown in Figure 7 for the oscillator damping ratios of 001, 0-02 and (05. Quadrature displacement spectra The quadrature displacement spectrum at the frequency @, and damping ratio f, is defined as 240)= G4” ool titeo @9) where Qy; is the peak factor for the displacement response. If desired, the quadrature pseudo-acceleration spectrum is obtained from the displacement spectrum as Qpisl) = OF Qeui() (40) Substituting equation (19) into equation (39), we obtain. do) = 0%, & vasbsteeao [ [4.+ 0784, +(C,+0D,)|H,P]|H,PO lode (an) To express the frequency integrals in equation (41) in terms of the ground response spectra, we resolve the products involving |H,|*, |H,|? and ||? into their partial fractions, as in equation (17), and obtain the following: Out) = 03, XE, veh AaaTige + Beslog + Casliy t Dy rlogit Agaliget Biz log + Cia Nyt Dir bay) (42) 6 TW Tar 1 i i | | || | ||| | | sll | | Pomoc mie a | 3 0.02 Ont 1 5 PERIOD (SEC) Figure 7. Coincident velocity oor spectra for floors 2 and 3 84 M.P. SINGH AND R. A, BURDISSO. whete the coefficients of partial fractions A,,~B, , and: A,,~B,, are defined in Appendix I. For the resonance case when H, is equal to H,, the frequency integrals are evaluated as in equations (27) and (28). Equations (40) and (42) have been used to obtain the quadrature pseudo-acceleration spectra, shown in Figure 8, for floors 2 and 3 of the structure in Figure 1. Again, such spectra must be developed for each pair of the connected floors. Quadrature velocity spectrum The quadrature velocity spectrum at frequency «, and damping ratio , is defined as 20) = 04 |" ooh Peo 3) where Q,, is the peak factor. Substitution for «®!,,() from equation (19) will give an equation similar to equation (41) except that the integrand will now be multiplied by @. Here again we resolve the product terms involving |H,|?, |H,|? and |H,]? into their partial fractions to express the frequency integrals in terms of the ‘ground response spectra, and obtain the following expression for the quadrature velocity spectrum: Qudeo) = 08, XE, eh aYe(WOTA ral gt Bis lag + Crs hig FD slag Aral igs + Bralogs + Cali Dialog] (44) where the coefficients of partial fractions A,,-D,, and A,4~D,, are defined in Appendix I. For the resonance case when H, is equal to H, the frequency intervals are obtained as in equations (32) and (33). Equation (44) has been used to obtain the quadrature velocity spectra for floors 2 and 3 of the structure in Figure 1. These spectra are shown in Figure 9. Similar spectra must be developed for each pair of the connected floors. 75. 69.00 FI/sx2 45.00 30.00 PSEUDO-ACCEL. 1.60 o.02 at 5. PERIOD (SEC) Figure 8. Quadrature pseudo-acceleation flor spectra for Noors 2 and 3 ~ | MULTIPLY SUPPORTED SECONDARY SYSTEMS. PART I 85 5 -00 2 FT/S 50 i 1,00 RELATIVE VEL. 0,50 00 1 8. PERIOD (SEC) Figure 9, Quadrature velovty floor spectra for Doors 2 and 3 FLOOR INPUTS FOR PSEUDO-STATIC RESPONSE Asshown in the accompanying paper, we require the root mean square values of the absolute displacements of each supporting floor as well as the correlation between these response quantities for calculating the pseudo- static response term. These quantities are obtained from the dynamic analyses of the primary structure. The expressions to obtain these are developed in the following sections. Maximum absolute displacement response ‘The mean square value of the absolute displacement response can be written in terms of the spectral density function of displacement as follows y(0) do (45) where ®y,(co) = spectral density function of the absolute floor displacement. From thestationary random vibration analysis, itis known that the cross or auto spectral density function of the absolute displacement, ®4(c2), and absolute acceleration, ®,,,(), are related by the following expression: Py) = OD, (0) (46) ‘Thus, the maximum displacement in terms of the acceleration spectral density function can be written as zLUZ] pug] = [" 28teo «n The right hand side of equation (47) is, however, the same as the following expression with slur = f° e4(e)|mn Pao (48) 86 M. P. SINGH AND R. A, BURDISSO Equation (48) is nothing but the auto relative displacement floor spectra value at «, = 0 divided by peak factor. This is given by equation (23), with Ijy and Iz, at c, = 0 defined as follows: Ty = Rego = O/CagP? = DylCa,? (49) Tag = (Rig(@r = O/C yg?’ = Ve /Crg? (50) where D, = maximum ground displacement and Ca, = peak factor for the ground displacement random process; V, = maximum ground velocity and C,, = peak factor for the ground velocity random process. No numerical problem is encountered in evaluation of the coefficients As,Bs,... .Ag, Bg, -- - » Ay, By, €te.,for w, = 0. Thus, to obtain the maximum floor displacement, the algorithm for the evaluation of the displacement auto floor spectrum can be directly used. Absolute displacement correlation coefficient The cross correlation between the absolute displacements of two floors can be written in terms of the cross spectral density function as ELUU) = f @h(o)do 61) Using equation (46), this can also be written in terms of the cross spectral density function of the absolute accelerations as suuil= |" o3e)zrdo (32) We notice that this correlation is the same as the coincident displacement cross spectrum obtained at © That is ELUU,) = [Caylos = O/ Pa? (53) Equation (53) can be directly obtained by using equation (35) for «, = 0. Here again, I,,,and I, are expressed in terms of maximum ground displacement and maximum ground velocity, as in equations (49) and (50). ‘The floor displacement response and their correlations must be developed for each connected floor. This input information can be prescribed in terms of the root mean square values of the floor displacements and the correlation coefficients defined as Su = BLU UMELUE DELU7D? (64) Fora secondary structure attached to floors 2, 3 and 4 of the structure in Figure 1 as well as for the system connected with floors2, 3 and ground floor, as considered in the companion paper,' this input information was obtained for the ground input defined in Figures 2 and 3. This is shown in Table I. Column 2 gives the maximum floor displacement whereas columns 3 to 6 give their correlation coefficient values. It is noted that in computing this input, we require maximum ground displacement and velocity. A reasonable estimate of these can be made by considering the proportionality of the maximum ground motion ment response and correlation coefcients Correlation coefficients Maximum Support disp. [ft] Ground Ground 0300 1 0-991 0978 0970 Floor 2 0309 oot I 0998 0994 Floor 3 031s 0978 0-998 1 0997 Floor 4 0319 0970 0994 0997 i MULTIPLY SUPPORTED SECONDARY SYSTEMS. PART II 87 parameters to the parameters of a standard earthquake, as expressed by Newmark etal. It, however, turns out that any reasonable value can be assumed for these parameters without affecting the response of the supported secondary system. This is due to the fact that the pseudo-static response is caused by the relative displacement between the supports. Thus, it can be obtained equally well by the relative or the absolute displacement characteristics of the support motions. The advantage of using the absolute displacement characteristics is that. these inputs can be obtained from the same computing algorithms as those used for calculating the auto and cross floor spectra. FLOOR INPUTS FOR CROSS RESPONSE, In the calculation of the cross response terms, we need (1) the coincident and quadrature cross spectra (2) the velocities of each floor and (3) the correlation between the coincident and quadrature components of the velocities of the connecting floors, as explained in the companion paper.! The development of the cross spectra was given in an earlier section. It has also been shown’ that the other two response inputs can also be obtained asa limiting case of the auto velocity, coincident and quadrature velocity spectra obtained at «, = 0. Thus, the algorithms developed for these in the earlier section can be directly used without any numerical problem. This input information can be defined in terms of the maximum floor velocity and normalized correlation coefficients between the coincident and quadrature components of the velocities; the normalization is done by dividing these correlation values by the root mean square velocities of the connected floors. Table II shows the floor velocities and normalized coefficients between the coincident components for floors 2, 3, 4 and ground. ‘The normalized coefficients given in this table, when multiplied by the velocities of the associated floors, give the correlation between the coincident components of the velocities of the two floors. Similar information about the quadrature component correlation is provided in Table III. These two tables were used in the analysis of the secondary system considered in the companion paper." CONCLUSION ‘Analytical methods are developed to obtain the inputs required in the seismic analysis of the multiple support secondary systems. In particular, the inputs in terms of the (1) auto floor response spectra, (2) cross floor Table Il. Floor coincident velocity response and correlation coefficients Correlation coetcient Maximum Floor Floor Floor Support vel. [fy] Ground 2 3 4 Ground 400 1 701 0549 451 Floor 2 0505 o7o1 1 0362 0390 Floor 3 0588 0599 0962 097s Floor 4 0.665 0451 0890 975 1 Table HI Quadrature velocity coefficients Quadrature velocity coefficient Maximum Floor Floor. Floor Support vel. [fs] Ground 2 3 4 Ground 0400 o —1038 0837 — 0684 Floor 2 0505 0038 0 0366 0353 Floor 3 0588 0837-0366 0. 0203 Floor 4 0665 ets =0353— —0203 0 88 MP, SINGH AND R. A. BURDISSO response spectra, (3) floor displacements and their correlations and (4) floor velocities and their correlationsare required and can be obtained by these methods. The auto floor spectra are defined for the pseudo-aceleration and the relative velocity response of an oscillator on the floor. The cross spectral characteristics are defined for the coincident and quadrature components of a pair of floor motions; also for each of these components, spectra are obtained both for the pseudo-acceleration and relative velocity responses of an oscillator. Thus in all sx sets of spectrum curves are required to be defined. To define the floor displacement and velocity related inputs no new expressions are required; itis shown that the algorithms developed for the auto and cross spectra can be directly used. ‘The proposed methods require a dynamic analysis of the supporting primary structure to obtain its dyngnye= characteristics in terms of the frequencies, mode shapes, modal damping ratios and modal participation factors. In the development of the methods random vibration principles are employed: yet, however, itis not necessary to define the base input in any stochastic form; the ground response spectra can be directly used. The analytical approach for utilization of these inputs in computing the secondary system response is described in the companion paper.? ACKNOWLEDGEMENTS ‘The work on this paper is financially supported by the National Science Foundation through Grant No. (CEE-8208897, This support is gratefully acknowledged. Any opinions, findings, conclusions and recommend- ations expressed in the paper are those of the writers and do not necessarily reflect the views of the National Science Foundation. APPENDIX I Partial fraction coefficients The coefficients of the partial fractions in equations (18) and (19) are obtained from the sol following systems of equations: (¥Va} = (Wohi ral (¥e} = (4h (85) where the matrix [Y,,] is defined as follows: [ of 0 ot 0 ] 20} (262 1) of 2003 (28? — 1) of | Dad= OY zeny 1 waft) cD 0 1 0 1 and the vectors {V3}, {W}, etc. {V7 = (A, B, C,, Dj} 3.4 (57) Wal) = wFors, W3(2) = wPeo;[— e7 (1 —483)— oF (1 — 4B7)] Ws(4) = 48,B,0,0,, W(3) = ovo; (1 — 462) (1-46?) (58) WA) =0, W.2)=0, W483) = 20,0,(6,07 ~ B03) WA) = 20,07 B,(1 487) — oF 0,B,(1 — 47)) The coefficients As, Ag, etc., in equations (23), (29), (35), (38), (42) and (44) are obtained as a solution of the following equations: Dedt¥s) = Ws Delt) = O%e)s Dad} = 7} Yale) = ahs Lalo} = obs Wal Mich = ro} i) Dadi = Wak Walid = Maks WadMis} = aad | dt") = ra} MULTIPLY SUPPORTED SECONDARY SYSTEMS, PART IL 89 in which the matrices [¥,;] and [¥,/] are the same as in equation (56) except for the proper substitution of the subscripts. The vectors {V,}, {Vg}, etc, are given by {V}" = {Ay B;, C,, Di} i=5,6,...,14 (60) {Wa}" = {op, 4Breoz, 0, 0}; {Wo}7 = (Ag, Bs, 0, O} {Wr}¥ = {Cy, Dy, 0, 0}; {We} = (0, of, 46202, 0} {1¥0}" = (0, Aa, By, 0}; {Wo} = (0, Cs, Ds, OF {Wys}T = {Aw Ba, 0, 0}; {W237 = {Ca, Dy, 0, 0} {Wisk = (0, Ag, Be, 05 {W437 = (0, Cy, Dy, 0} (61) APPENDIX II In this appendix, certain functions required for computing the integrals I(a,, a3) and Is(a, a3) in equations (26) and (31) are defined. F(®) = Ing /(@2T1y)) 2r(ay + a2/3)\/(Em— 2) (Dy ~ 27/3)/(En 21) (62) ra(e9)}e—> | where r = /to,; o, = the highest frequency in the input, I,,, and Iz4, are defined by equation (24), and Cu = alts Bor As dz, 0, 0) Dy, = Bult, Boy 1, 0) (63) Eq = Bult, Bo, 0, 1) uD Fm = Anl?, Bor 0, 22> a3: 0). The functions 4,, and B,, in equation (63) are related to the following integrals: Ags Bos Ais as yy 4) = f (@,08 + 4,080? + a,ato* +a,080°)|Ho|*do ) _Bylt, Boy , B) = f (azo? +b0%)|Hyl2do 65) Equations (64) and (65) can be obtained in closed form to define A,, and B,, as follows: _m, 2Ny+m, | f-+r?)—2r V3) Aas Bt 0 0) = Ore pM EHF VSB _2Na{(1=1?) + 26377} m,/2(147) PUPP + 4for7} __2Na+M,2_ | flr ea 4r°(1 = B2) J(1— Ba) T+r?42r JB) 90, M.P. SINGH AND R. A. BURDISSO = (7 /16B5) [a4(1 = 48) +45 +42 + 4, (1 +485)) Na = ~ (A/L6BOL lay + a,)(1 465) +43 + 42] | im, = (PGBS) (ag + 4;)(1 +485) + a5 + a9] (67) mz = r4/2[2(1 ~263)a, +4; —a,) M, = m,/2iBor | Mz = m,/AiBor J and Q, = 2(n-8) (68) =2 in which 6 = tan~? (289/|1—r|?), and (, = Ort in {itr = 2 J = fo) Bas B00 = ot Mg TB where my = (a+ b)%/r } «a Ny = —br?/2 REFERENCES 1. RA. Burdisso and M. P. Singh, Multiply supported secondary systems Part I: response spectrum analysis, Earthquake eng struc. dyn. 15, 53-72 (1987). 2 RVA.Burdisso and M. P. Singh, ‘Seismic analysis of secondary systems with multiple supports’, Prac. con. struct.anal.des.nucl. power plants Porto Alegre, Brazil (1984). 3. MP. Singh, ‘Response spectrum methods for secondary systems’, Symp. earthquake effects plant equip. Hyderabad (1984). 4 A-Asfura and A. Der Kiureghian, ‘A new floor response spectrum method for seismic analysis of multiply supported secondary systems, Report No. UCB/EERC-84/04, Earthquake Engineering Research Center, University of California, Berkeley, CA, 1984 5. L Meirovitch, Computational Methods in Structural Dynamics, Sijtholt and Noordhof, The Netherlands, 1980. 6 MP. Singh, Seismic response by SRSS for nonproportional damping’, J. eng. mech. die. ASCE 106, 1405-1419 (1980). 7. M. P. Singh and A. M. Sharma, ‘An alternative procedure for direct generation of seismic loor spestra’, Trans. 7h SMIRT conf, Chicago, IL, K(a) (1983), Paper K6/10. 8. MP. Singh, ‘Generation of seismic floor spectra’ J. eng. medh, dio. ASCE 101, 593-607 (1975) 9. R.A. Burdiso and M, P, Singh, ‘Seismic analysis of multiply connected secondary systems’, Report No. VPI-E-84-30, College of Engineering Virginia Polytechnic Institute & State University, Blacksburg, VA, 1984, 10. M.P. Singh, ‘Seismic design input for secondary systems’, J. struct. div. ASCE 106, $0S-S17 (1980) 11, ALM. Sharma and M. P. Singh, ‘Ditect generation of stismic floor response spectra for classically and nonclssically damped structures, Report No. VPL-E-83-44, Virginia Polytechnic Institute & State University, VA, 1983, 12, N.M. Newmark, J. A. Blume and K:K. Kapur, ‘Seismic design spectra for nuclear power plant J. power div. ASCE 99, 287-303 (1973,

You might also like