COMPUTATION OF SEISMIC RESPONSE OF LIGHT ATTACHMENTS To BUILDINGS Nathan M, Newmark" and Roberto Villaverde!? SUIMARY A simple approximate procedure is introduced for conputing the maxinun response of light secondary systens attached to buildings subjected to earthquakes. A numerical example is provided to describe its use, and the results of a comparative study are presented to illustrate its accuracy. INTRODUCTION Ordinarily, there are a variety of attachments to the floors and walls of large and complex buildings which, because of their different character~ istics and functions, may not be considered as part of the structures that support them, but because of their low mass, stiffness, and damping values may be particularly vulnerable to the effects of earthquakes. Exanples of these attachments are the piping system, electrical or mechanical equip~ ment, and parapets that are usually present in multi-story buildings, industrial plants, or nuclear power facilities. In principle, the analysis of such building attachments -~ also called secondary systems -- may be carried out in conjunction vith the analysis of the primary structures to which they are connected. However, the computa~ tional difficulties, the excessive number of degrees of freedom involved, and the problem of schedule and efficiency introduced by having to analyze together two systems that are customarily analyzed separately, make this procedure costly, cumbersome and impractical. Upon recognition of the necessity of a simplified method to facilitate their analysis, several authors have suggested in the past few years several approximate methods. But although sone of these methods have been proved to give reasonable results for sone particular chses, such as for secondary systens with very small masses and natural frequencies far off the natural frequencies of their supporting systens, it is believed that, in general, all of them are either inaccurate or impractical, and that there is still a need for a simple and reliable procedure for the seismic analysis of secondary systems. In this paper, then, an alternative approximate method is proposed to estimate, accurately and'in a simple manner, the maximun response of secondary systems attached to buildings subjected to earthquakes. The method, derived from the analysis by a variation of the response spectrum method [1] of the composite system forned by a primary and a secondary system, and fron the development of simplified analytical expressions for each of the steps that are needed to carry out such an analysis, is de~ scribed in detail in Ref. 2 and presented here in a summarized forn for its direct application in the solution of practical problems. Texofessor of Civil Engrg. and Institute for Advanced Study, Emeritus, Univ. of Illinois, Urbana, Ill., U.S.A. r ‘| TResearch Assistant, Dept. of Civil Engrg., Univ. of I11, Urbana, 11. 343SCOPE AND LIMITATIONS As presented, the nethod may be applied for the analysis of a mlti- degree-of-freedom secondary system connected to arbitrary points of a nulti-degree-of-freedom primary structure. It exhibits, in addition, the following characteristics: (1) It is simple enough to carry out the necessary computations by hand. (2) It fully takes into account the inter~ action between 2 secondary system and its supporting structure, including the damping effect that each system exerts upon each other. (3) It is formilated in terms of the natural frequencies, mode shapes, and damping ratios of independent primary and secondary systems. (4) It uses the re~ sponse spectrum of the ground motion prescribed for the analysis of its primary system to define the earthquake input to a secondary system. (5) It may be employed to analyze secondary systems that are near or in resonance with their supporting systems. The method, however, is limited to those cases in which the separate primary and secondary systems are linear elastic systems with classical odes of vibration. In addition, it is restricted to the analysis of secondary systems that are comected to a primary system at no more than two points, and have small masses in comparison with the masses of their supporting structures. RECOMMENDED APPROXIMATE PROCEDURE consider a secondary system attached to one or two arbitrary points of a supporting primary system. Let the independent primary system be described by its matrix of unit-partictpation-factor mode shapes [0], its natural frequencies ups, 1= 1, 2, ..-, Np» its generalized massest Mi, 4 = 1, 2, ..., Np, and its modal damping ratios fp,, i= 1, 2, +++, Np» where Np represents the nunber of degrees of freedom of the systen. Sin{larly, let the independent secondary systen be fixed at its point of attachnent with the primary system, and let it be characterized by its modal matrix [¢] (mode shapes also with unit participation factors), its natural frequencies Ws;, j= 1, 2, .+., Ng, its generalized masses mt, j = 1, 2 Ng» and its nodal danping ratios Ey, = 1, 2, -.-, Ng in which Nq denotes the number of degrees of freedom of’such an independent secondary systen. Let, then, the following variables be defined as follow: SD(u, &) = response spectrum displacenent ordinate corresponding to a natural frequency w and damping ratio £. Yay 7 ff = generalised ee ratio in Jth secondary and ith prinary ‘The ith generalized mass of a system with N degrees of freedom is defined x Mi= 5 ow ew) nel where ¢, (1) is the amplitude of the ith mode shape of the system at the Jevel of its uth mass, and M, represents the value of such a nth mass. 344¢,(1) + amplitude of the jth mode shape of the independent secondary system at the level of its nth mass, T £49} =65,G) 4nG)-9 G0 +++ “ty 4g} = vector of element distor tions in the jeh node of fhe independent secondary system, 7 : a . Ther. ky vee My vector of normalized differential flexibilities 2 = Ke an ty GMS 8) ts 4 2444) = 4,60 + oyf4,( ~ 4,(4)] = central value of the amplitudes of the points of attachment in the ith primary and jth secondary nodes. 1,2, 11.) Nj#l, represents the jth stiffness constant of the seco yystem, and (1) and (1) are the ampli- tudes of the kth and tth primary masses - the masses of the primary systi to which the secondary system is attached - in the ith mode of the inde— pendent primary systen. For secondary systens with a single point of attachment, (€6/f,.} = {0}, 8, = 0, and #,(4,4) = 6, (4). In the above expressions, k,, Assume now that the assembled system (primary and secondary systems together) is an N4N_ degree-of-freedom system whose natural frequencies are the frequencis 8£ its independent primary and secondary components. Classify a mode of this assembled systen as a resonant mode {f its natural frequency is a frequency comon to both independent components, and asa nonresonant mode if its frequency is any other. Let R denote the number of these resonant modes, and for each node of the assenbled systen identify with subseripts T and J the parameters of the separate primary and second~ ary systems, respectively, vhose frequencies are the closest to or coincide with the frequency of such @ mode. Thus, if the base of the primary system is excited by a ground motion, and if this ground motion is specified by its response spectra, the vector of maximum distortions* of the secondary system may be calculated by 5 2 (3, s @ where {x.}°), which represents the coubined maximum response of the ® Secondar} aysten in tvo resonant nodes with equal frequency, and ({-) which denotes the maximm secondary response in the rth noaresonant*node, may be determined as follows: ‘Fig presented in this paper, the method is formulated to obtain specifically maximum distortions. It is important to note, however, that if few adjust- ments are made, it may be used as well to estimate other responses (Ref. 2). 345Resonant nodes tx3) = x) 1293 sp, &) @ where ate =e, @ ° Ea 83 1 =i CO) BEG +E and T z Fane e200 FaHa) 02 (2,3) © z Dy ~ oT Dra Fy Ff im which, depending on the relation between the damping and mass ratios of the separate primary and secondary systems, “|, is given by one of the following formulas: 1 1,3) yoy le 1€, =! cases foc Wy |s | 8, tan 7 VEEL / + 8) © where 5! and §! are of the form gla +2/lu, 8&1, rean a and &, and & are given by ma = 5, Fe, @) case 11: |2,(1,9) Mig] 216, = UL + RG, 6, 71/4882) ® Dye, a = r where = £, + WU lw, s(E,)1 ao) In Eqs. 7 and 10, S(E_), a function of £1, r= 0, m, n, stands for the duration of the equivélent white noise that best represents the gromd motion under consideration. It may be calculated as suggested in Ref. 2, or it may be assumed on the basis of the characteristics of the expected average earthquakes and site conditions in the area of interest. Thus, for example, it is reported in Ref. 3 that for a group of earthquakes recorded in relatively firm ground along the west coast of the United States such an equivalent earthquake duration may be taken as 12.5 sec. Notice that when |, (I,J) “yy [= [6 - &, a wir I Fs given by Hq. 5 becones indeterminate. Hovever, it is shown in Ref, 2 that the amplification factor 346if sone of the second-order terms that were neglected in the derivation of that amplification factor vere included in this equation, its denominator would be different from zero. Consider, thus, that ye O when [2 (,5) My | = 1 7 |. jonresonant sodes with a frequency of the primary systen x, © Op pe fe ca3M) 6 2 Oy a yee sere J rear 1 sO, ot, ap where Fe 7 (eae cor Ax a2) Fy 7 sum(l-85) (A, (/A,1EG#69 / 048 | as) P= acs Aas aa ia which . 2 yg? 2 8D = God IGE oP as) 87 Go, - 54 Me, 9) as) 1” Gogo,” Fags! Cp, and "sgn" is a function that reads as "the sign of." Nonresonant Modes with a frequency of the secondary systen © = (ay sp w, . > an where ' w= cae F arate ey a as) 1 sa in which 21) = a2 pw? = a2 2 BAC = [Gay MO) gy 1 + ED as) (20) It {s important to note that the above expressions have been presented in its most general form and that, consequently, their application does not necessarily require the use of such a general form. Rather, without over- ooking that this is an approximate method, one should interpret these expressions and use a simplified version of then. Thus, for example, al- though Eq. 1 indicates the use of all the assembled system modes, and Eqs. 18 and 11 consider respectively all the modes of the primary and secondary systems, in similarity with a conventional nodal analysis one should take into account only those of such modes which significantly affect the response of the secondary system under consideration ATILLUSTRATIVE EXAMPLE To clarify the use of the procedure established above, the maximum distortions of the secondary system shown in Fig. 1 are here calculated for the case in which its primary supporting structure is subjected to a portion of El Centro (May 18, 1940) earthquake whose response spectrum js shown in Fig. 2. (The units of the values indicated in Fig. 1 are T-sec“/m for the masses and T/m for the stiffnesses). The damping matrices of the independent primary and secondary systems are assumed proportional to their respective stiffness matrices, and the damping ratios in their fundamental nodes are considered to be 2 and 0.1%, respectively. Their modal matrices, natural frequencies, modal damping ratios, and generalized masses are as: Primary Systen 15 O04 0. = Lo ces 002 Ms 5 so oa) 5, ae cy = fu.0 02-0. = 20 ep = 0.06 Mg + 0.9 1+ fi.0 02-02] s+ 20en8 6, = 0.08 Hy 15-006 0. = 3.0 exp. = 0.0 =o 15-06 of f+ 30ene8 6, 70.08 Oy Secondary Sys 0.5 0.5 = 1.0 epee = 0.001 at = 0.0045 tel = 15-05 = Fens 6, = 0.00161 mg = 0.0025 2 2 ‘Tous, in accordance with the procedure introduced above, in this exam- ple the secondary system and its supporting structure give rise to a five- .dom assembled system whose natural frequencies in cycles per on t, fy fy fs 1.0 1.0 v2.0 3.0 Then, this assembled system has two resonant and three nonresonant modes. The first two are the resonant ones; the third is a nonresonant mode with a frequency of the secondary system whereas the last two are nonresonant modes with frequencies of the primary system. By inspection, however, of the above frequency values, it may be established beforehand that the resonant modes dominate the response of the secondary system herein being considered. Hence, its mximm response will be approximated by the maximm response in such resonant modes as follows: Observe that the frequency and damping ratio of the resonant modes are (see Eqs. 3 and 4) , 7 2 rad/sec 3 £, * }(0.0240.001) = 0.0105 Observe also that 8, = 0-00075(1.5)/[1.0(0-0045)] = 0.25 and that the central value of the modal amplitudes of the points of attach~ nent is : 401,21) = 0.5 + 0.25(4.5-0.5) = 0.75. 348Consequently the values of |&) - § | and 19 (z,3)"%7,] axe 7 ° = 0.02 = 0.001 = 0.019 8G. = 0.75 0.008578.5 = 0.02372. Evidently, in this case |¢(1,J) Y¥ 7 [is greater than le, 6, | and hence o__ in Bq. 5 should be calculated by sean of Bq. 9.°Tror thie Purpose, th@#, assume in accordance with the values calculated in Ref. 2 that the equivalent earthquake duration for &, = 0.0105 is $(0.0105) = 17.2 sec. ‘Thus, Eq. 10 yields Elm 0.0105 + 2/2m(17.2) = 0.02901, and consequently from Eqs. 9 and 5 one obtains 1 = 1/{2 + (0.02372) ¢0,019)"1/14(0.02901)7}} = 0.94349 f3c1.0-0.94349)(0.75)* can? wow? ue) = 8.87815 0.02372)" - (0.019) With ehto valve of ¥{1) and since from Fig. 2 one has that SD(2 ,0.0105)= 0.201 m, Eq. 2 leads therefore to 0.5 0.392 (4, = a.e7eis( 1.5 - 0.5 }o.201=( 1.784 Sm -1.s -2.677 from which one may conclude that if the response in all the other modes is neglected, the maximum distortions of the secondary system result approx- mately as. aan Daan (1786 fm 2.677 COMPARATIVE STUDY The accuracy of the procedure presented above is evaluated in Ref. 2 by means of a comparative study with "exact" time-history solutions. In this comparative study, a number of secondary systems with different mass ratios, frequency distributions, damping characteristics, location of the points of attachment, and number of these points of attachment, are ana~ lyzed for three different earthquakes, and the averages of their approxi~ mate and exact maximm distortion responses to these three earthquakes are compared. The results of the study are statistically summarized in Table 1, where some statistics of the obtained approximate to exact maximum distor- tion ratios are shown. For each of the categories and damping character- istics considered, the statistics furnished in this table indicate the mean 348value, coefficient of variation (c-o.v.), and the maximum and minimm values of che sample formed by all the approxinate to exact maximm distortion ratios of all the analyzed systens with one of such damping characteristics and within one of such categories. ACKNOWLEDGMENTS This paper is based upon the doctoral dissertation of Roberto Villa~ verde at the University of Illinois at Urbana-Champaign under the direction of Nathan M. Newmark. The study was supported by the National Science Foundation (RANN) under Grants AEN 75-08456 and ENV 77-07190. REFERENCES 1, Villaverde, R., 1980, "Earthquake Response of Systems with Nonpropor~ tional Damping by the Conventional Response Spectrum Method," 7th World Conf. on Eartha. Engrg., Istanbul, Turkey. 2. Villaverde, R., 1980, "Seismic Response of Light Attachments to Build ings," Ph.D. Thesis, University of Illinois, Urbana, Illinois. 3. Newmark, N.M. and Rosenblueth, E., 1971, Fundamentals of Earthquake Engineering, Prentice Hall.