Computers and Structures 74 (2000) 131145

www.elsevier.com/locate/compstruc

Structural responses considering the vertical component of earthquakes

Alfredo Reyes Salazar a, Achintya Haldar b,*
b a noma de Sinaloa (UAS), Culiancan, Sinaloa, Mexico Facultad de Ingeniera, Universidad Auto Department of Civil Engineering and Engineering Mechanics, University of Arizona, Tucson, AZ 85721, USA

Received 31 December 1997; accepted 18 November 1998

Abstract The guidelines in the NEHRP Provisions and the Mexican Code regarding the eects of the vertical component of earthquakes on the response of frames are re-evaluated. Using a time domain nonlinear nite element program developed by the authors, the seismic responses of frames are evaluated realistically by simultaneously applying the horizontal and vertical components of earthquake motion. Three steel frames and 13 recorded earthquake motions are considered. The same response parameters are then estimated using the two codes, and their error is evaluated. It is found that, if the frames remain elastic, the NEHRP Provisions estimate the maximum horizontal deection at the top of the frames and the bending moment in the columns very accurately; the Mexican Code overestimates them. If the frames develop plastic hinges, the Mexican Code conservatively overestimates them, but the NEHRP Provisions underestimate them in some cases. Both codes signicantly underestimate the axial loads in columns. The underestimation increases as the frames develop plastic hinges. The underestimation is more for interior columns than for exterior columns. If the ratio R of the PGA of the vertical and horizontal components of an earthquake is higher than normal, the underestimation increases as R increases. The underestimation is not correlated with frame height. The vertical component may increase the axial load signicantly. Since they are designed as beamcolumns, the increase in the axial load will have a very detrimental eect on the performance of the columns. In light of the results obtained in this study, the design requirements for the vertical components need modication. At the very least, further study is required. # 1999 Elsevier Science Ltd. All rights reserved.
Keywords: Seismic response; Seismic design; Steel frames; Load combinations; Design criteria; Lateral deection; Vertical acceleration

1. Introduction The inuence of the vertical component of an earthquake on the overall seismic response of structures has long been of considerable interest to the profession.

* Corresponding author. Tel.: +1-520-621-2142; fax: +1520-621-2550. E-mail address: halder@u.arizona.edu (A. Haldar)

Several design codes tried to address the issue in many dierent and, it is hoped, conservative ways. Despite this, many steel structures suered a considerable amount of damage during the Northridge earthquake of January 1994. Severe cracks developed in many structures during the earthquake. Researchers mostly attribute this damage to defects in welding and material, and to design-related causes. Several recorded ground motions during the Northridge earthquake indicate that the vertical component was much larger

0045-7949/00/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 5 - 7 9 4 9 ( 9 9 ) 0 0 0 3 1 - 0

132 Table 1 Strong motion earthquakes Earthquake

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145

Station

Acceleration (cm/s2) PGAV PGAH 342 825 359 998 276 615 250 326 551 389 380 573 472 PGAV/PGAH 0.60 0.27 0.38 0.44 0.51 0.60 0.61 0.62 0.68 0.73 0.86 1.07 1.11

1 Northridge earthquakes 2 3 4 5 6 7 8 9 10 11 12 13

El Centro Los Angeles, 1526 Edgemont Ave Los Angeles, Wadsworth V.A. Los Angeles, 10660 Wilshire Blvd Los Angeles, Grith Observatory Jenson Filtration Plant Los Angeles, Wadsworth V.A. Topanga Fire Station Sherman Oaks, 1525 Ventura Blvd Los Angeles, 4929 Wilshire Blvd Los Angeles, 10751 Wilshire Blvd Conoga Park, Santa Susana Los Angeles, 4929 Wilshire Blvd

206 225 135 441 142 369 152 201 377 285 326 613 526

than is usually considered normal in design. This observation prompted a discussion about whether the excessive vertical acceleration may have caused the damage since, in the past, similar steel structures behaved well when the vertical component of the earthquakes was not so strong. Although extensive studies are now being conducted in related areas, it is important for the profession to reconsider the adequacy of the design provisions outlined in the model building codes to consider the eect of the vertical component of earthquakes. For numerical evaluation, earthquake motions are generally represented by three components: two horizontal and one vertical. The peak ground acceleration of the vertical component is usually smaller than those of the two horizontal components. Since the horizontal motion of the ground has the most signicant eect on the structural response, it is that motion which is usually thought of as earthquake load. Therefore, most building codes with earthquake provisions require that an equivalent lateral load as a result of the horizontal ground motion be used in simplied empirical approaches [1]. The eect of the vertical component is considered indirectly. Obviously, if the vertical component is much stronger than is usually considered normal, then the simplied code approaches may underestimate the seismic load, and the structure will not perform as intended. The ratio of the peak ground acceleration of the vertical component (PGAV) to the maximum horizontal peak ground acceleration (PGAH), denoted hereafter as R, can be used to study the inuence of the vertical component on the overall seismic response behavior of structures. For normal earthquakes, this ratio is

expected to be around 2/3. For the widely used earthquake time histories recorded during the El Centro earthquake of 1940, this ratio is 0.60, as shown in Table 1. For the 12 earthquake time histories recorded during the Northridge earthquake listed in Table 1, this ratio varies between 0.27 and 1.11, and ve of them have a ratio greater than 2/3. Any one of these 12 earthquake time histories can be used to represent the Northridge earthquake in future designs. Thus, the date collected during the Northridge earthquake gives the profession an opportunity to re-evaluate the adequacy of the provisions suggested in design codes on how to consider the eect of the vertical component in design. This study specically addresses two major seismic design guidelines for buildings, namely, the National Earthquake Hazard Reduction Program (NEHRP) Recommended Provisions for Seismic Regulations for New Buildings [2], hereafter denoted as the NEHRP Provisions, and the Mexico City Seismic Code [3]. The design requirements in other codes are expected to be similar. In the 1994 edition of the NEHRP Provisions, a new requirement was added to consider the combined eects of the horizontal and vertical components on the structural response. It is addressed indirectly in the section on `Combination of load eects'. It is suggested that the eect of gravity loads and seismic forces be combined in accordance with the factored load combinations as presented in the American Society of Civil Engineers Minimum Design Loads for Buildings and Other Structures (ASCE 7-95) [4], except that the eect of seismic loads, E, shall be dened as E=QE+0.5 CaD to consider the eect of both the horizontal and vertical components of an earthquake,

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145

133

where QE is the eect of horizontal seismic forces, Ca is the seismic coecient based upon the soil prole type and the value of Aa as determined from Section 1.4.2.3 or Table 1.4.2.4a of the NEHRP Provisions, and D is the eect of the dead load. The commentary of the Provisions further adds that `0.5 Ca was placed on the dead load to account for the eects of vertical acceleration. The 0.5 Ca factor on dead load is not intended to represent the total vertical response. The concurrent maximum response of vertical acceleration and horizontal accelerations, direct and orthogonal, is unlikely and, therefore, the direct addition of responses was not considered appropriate.' In the Mexico City Seismic Code, the eect of the vertical component is considered to be a fraction of the eect of the horizontal component. It states that `For the buildings located in seismic zones C and D the eect of the vertical component should be considered. This eect shall be taken as 2/3 of that of the largest horizontal component. This eect while combined with gravity and horizontal component eects should be taken as 0.3 of the above equivalent vertical eect'. Eectively, the code recommends that the eect of the vertical component should be estimated as 20% (the product of 2/3 and 0.3) of the eect of the largest horizontal component. This criterion can be interpreted another way. If the horizonal maximum response is H, than the vertical maximum response will be 2/3 H. Assuming that both maxima occur at the same time and using the square root of the sum of squares (SRSS) rule, the total response considering both components can be calculated as q 2 2a3H 2 1X2H. Obviously, the requirements H in the Mexico City Seismic Code appear to be much more conservative than those of the Provisions because of this assumption. In light of the extensive damage suered by steel structures during the Northridge earthquake and the signicant amount of information collected during the earthquake enabling detailed analytical studies, it is very desirable to compare the accuracy of the two codes in estimating the eect of the vertical component. The main objectives of this paper are: (1) to evaluate the eect of the vertical component analytically for several recorded earthquakes for several steel frames representing dierent dynamic properties in terms of their maximum lateral displacements and the maximum axial loads and bending moments in the members; (2) to evaluate the eect of the vertical component according to the NEHRP Provisions and the Mexican Code; and (3) to compare the analytical results with the codes' recommendations, in order to evaluate the adequacy of current design practices.

2. Analysis procedure In order to meet the objectives of this study, the nonlinear seismic responses of structures subjected to both horizontal and vertical components of an earthquake need to be evaluated as realistically as possible. The authors, with the help of other research team members, developed a highly ecient time domain nite element-based algorithm to estimate the nonlinear seismic responses of steel frames considering geometric and material nonlinearities. This sophisticated algorithm can also be used to estimate the seismic response of structures, instead of using simplied approaches such as the equivalent lateral load procedure and the modal analysis procedure suggested in the NEHRP Provisions. This type of elaborate analytical procedure is not expected to be used routinely by the design profession; however, it can be used to study the adequacy of the simplied methods suggested in the design codes. The fundamentals of the analytical procedure are available in the literature [5,6], but cannot be described here due to lack of space. Therefore, only the essential features required for the purposes of this paper are discussed briey below. Nonlinear bahavior of a frame can be produced by changes in the geometry, including the PD eect and/or material properties. The eects of geometric nonlinearity are changes in the member lateral stiness due to the eect of axial force, the change in member length due to the bowing eect and axial force, and the nite rigid body deformation of a member with small to moderate relative rotation. Most of the currently available nite element-based nonlinear analysis techniques for frames are based on an assumed displacement eld. In order to capture the eects of change in the axial length of an element due to large deformation, several elements are needed to model each member. The necessity for a large number of elements coupled with the use of a numerical integration scheme to obtain the tangent stiness matrix for each element several times during the analysis makes this approach uneconomical. Alternatively, the assumed stress-based nite element method [79] can be used to derive an explicit form of the tangent stiness. In this approach, the stresses on an element can be obtained directly instead of using the less accurate method of taking the derivatives of the displacement functions as in the assumed displacement eld approach. The method is very ecient and economical because of this feature and the use of fewer elements in describing a large deformation conguration, and because it needs no integration to obtain the tangent stiness. This procedure is particularly applicable to steel structures. It gives very accurate results and is very ecient compared to the displacement-based approach [6,8,9]. This method is used in this study.

134

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145

The other major source of nonlinearity in frames is material nonlinearity. Material nonlinearity occurs as a result of the nonlinear constitutive relationship of the material. In the analysis of steel structures, the three most common assumptions for the material behavior are the elasticperfectly plastic, isotropic strain hardening and kinematic strain hardening models [10]. Considering the complexity of the problem under consideration and the usual practice in the profession, the material nonlinearity of steel will be considered to be elasto-perfectly plastic in this study. The von Mises criterion [10] is very appropriate, and is used in this study. The development of the static governing equations using the assumed stress method is not described here due to lack of space. Only the dynamic governing equations required for the nonlinear seismic analysis of frames and the solution strategy are presented very briey below. The equation of motion of a linear system under dynamic and seismic loadings can be expressed as [11]: MD CD KD F MD g 1

where M, C and K are the mass, damping and stiness . matrices of the frame, respectively; D, D and D are the acceleration, velocity and the relative displacement vec tors, respectively; Dg is the ground acceleration vector, and F is the external dynamic force vector, if present. For the nonlinear case the dynamic and seismic governing equations of motion can be expressed in incremental form as [12]: MtDt D
k

hinge locations. In the past, several analytical procedures have been proposed to predict the deformation of elasto-plastic frames under increasing seismic and static loads. However, most of these formulations were based on small deformation theory. In this study, each elasto-plastic beamcolumn element can experience arbitrary large rigid deformations and small relative deformations. In addition to the elastic stressstrain relationships, the plastic stressstrain relationships need to be incorporated into the constitutive equations if the yield condition is satised. Several yield criteria have been proposed in the literature in terms of stress components or nodal forces. Since the nodal forces can be obtained directly from the proposed methods, the yield used in this study is expressed in terms of nodal forces. When the combined action of axial force and bending moment (this is for plane structures only) satisfy a prescribed yield function at a given node of an element, a plastic hinge is assumed to occur instantaneously at that location. Plastic hinges are considered to form at the ends of the beamcolumns elements. The yield function (or interaction equation) depends on the type of section and loading acting on the beamcolumn element [13]. The yield function for the two-dimensional beamcolumn element has the following general form: f P,M,sy 0 at X lp 3

t CtDt D

t KtDt DDk k MtDt D g 2

where P is the axial force; M is the bending moment; sy is the yield stress, and lp is the location of the plastic hinge. For the W-type sections used in this study, this equation has the following form:  P  P     M   M   10  4

tDt Fk tDt Rk1

nx

where tK(k ) is the tangent stiness matrix of the system of the kth iteration at time t; (t+Dt )D(k ) and (t+Dt )F(k ) are the incremental displacement vector and external load vector of the kth iteration at time t+Dt, respectively; and (t+Dt )R(k 1) is the internal force vector of the (k 1)th iteration at time t+Dt. All other parameters were dened earlier. The step-by-step direct integration numerical analysis procedure using the Newmark b method is used to solve Eq. (2). Explicit expressions for the tangent stiness matrix consisting of geometric and material nonlinearities and the internal force vector are developed for each beam column element using the assumed stress method for each iteration at a given time t. The mathematical details of the derivation are not shown here, but can be found in the literature [79]. As stated earlier, in this study the material is considered to be linear elastic except at plastic hinges. Concentrated plasticity behavior is assumed at plastic

where Pn and Mnx are the axial strength and the exural strength with respect to the major axis, respectively. The presence of a plastic hinge in the structure will produce additional axial deformations and relative rotations in a particular element. This is considered in the stiness matrix and the internal force vector of the plastic stage. Explicit expressions for the elasto-plastic tangent stiness matrix and the elasto-plastic internal force vector are also developed. The mathematical derivations can be found in the literature [79]. Depending on the level of earthquake excitation, in a typical structure all the elements may remain elastic, or some of the elements will remain elastic and the rest will yield. The structural stiness matrix and the internal force vector can be explicitly developed by considering individual elements and the particular state they are in. Since actual earthquake time histories are used in

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145

135

Fig. 1. Three steel frames. (a) Frame 1; (b) frame 2; (c) frame 3.

this study, the inertia and applied forces are available. However, further discussion of damping is necessary at this stage. In a realistic seismic analysis of steel frames, the amount of damping energy that will be generated will depend on the nonyielding and yielding state of the material and on the hysteretic behavior if the material yields. For mathematical simplicity, the eect of nonyielding energy dissipation is usually represented by equivalent viscous damping varying between 0.1 and 7% of the critical damping. The damping is often increased in linear analysis to approximate energy losses due to anticipated inelastic behavior [14]. In a rigorous seismic analysis this practice is not appropriate, since the energy losses due to inelastic behavior would be counted twice. Based on an extensive literature review, it is observed that the following Rayleightype damping is very commonly used in the profession:
t

modes, respectively. Then the following algebraic equation system is solved for a and g [11]:     1  o i        x  o 1 i  a   i  6      xj  g 2 1  oj    oj where oi and oj are the natural frequencies of the ith and jth mode, respectively, and are calculated using the Stodola method in this study. Usually the ith mode is selected as the rst mode, and the jth mode as the higher mode that contributes signicantly to the structural response. A computer program has been developed to implement the algorithm. The program was extensively veried using information available in the literature. The structural response behavior and the members' forces in terms of axial load, shear force and bending moment can be estimated using the computer program. 3. Description of structures and earthquakes Three steel frame structures representing dierent

C aM gt K

where a and g are the proportional constants. The use of both the tangent stiness and the mass matrices is a very rational approach to estimate the energy dissipated by viscous damping in a nonlinear seismic analysis. The constants a and g can be determined from specied damping ratios xi and xj for the ith and jth

136 Table 2 Member sizes Frame 1 2 Story

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145

Exterior columns W14 211 W14 145 W14 370 W14 277 W14 211 W14 193 W14 665 W14 455 W14 426 W14 398 W14 342 W14 311

Interior columns W14 283 W14 211 W14 550 W14 370 W14 257 W14 211 W14 730 W14 665 W14 455 W14 426 W14 398 W14 342

Girders W18 175 W18 119 W24 335 W24 279 W24 192 W24 131 W36 650 W36 439 W36 280 W36 245 W36 210 W36 194

1 23 12 34 56 78 14 56 78 910 1012 1315

dynamic characteristics are considered in this study: a three-story, an eight-story, and a 15-story steel frame, all shown in Fig. 1. They will be denoted hereafter frames 1, 2, and 3. They represent short, intermediate and tall buildings, respectively. The geometry of these three frames is shown in Fig. 1, and their member sizes are given in Table 2. The story height for these frames is a constant of 3.66 m, and the width of each bay is 7.32 m. In all these frames, the columns are assumed to be made of Grade 50 steel and the girders of A36 steel. All three frames are assumed to have rigid connections. These frames are designed to remain elastic when subjected to all 13 earthquake time histories. This will help to compare results, as will be further elaborated later. In the seismic analysis of these frames, uniform gravity loads were applied to the girders and the equivalent nodal forces were calculated, as required for the assumed stress-based nite element formulation used in this study. One node was placed at the midspan of each of the girders. Each node is considered to have three degrees of freedom. The natural periods of the three frames are 0.31, 0.54 and 0.68 s, respectively. These frames are assumed to be located in the Los Angeles area. These three frames with dierent dynamic characteristics are subjected to the 13 strong motion earthquakes identied in Table 1. The rst earthquake in the table is the El Centro earthquake of 1940. The El Centro earthquake time histories are widely used in the profession to represent a typical earthquake, and are used here for reference purposes only. The other 12 are strong motion time histories of the Northridge earthquake of 1994, recorded at dierent locations. They are presented in increasing order for the ratio R. It is important to note that any one of the 12 records can be used to represent the Northridge earthquake, although they have dierent peak ground accelerations, strong motion durations, frequency contents,

and R parameters. As mentioned earlier, the R values for these records vary between 0.27 and 1.11, and at least ve of them have an R value greater than 2/3, which is usually considered to be normal in the profession. These earthquakes are denoted hereafter as earthquakes 113.

4. Results and observations In order to analytically evaluate the eect of the vertical component on the seismic responses of the frames, the following two cases are considered. . Case 1the frames are excited by the horizontal component containing the maximum PGA only. . Case 2the frames are jointly excited by the horizontal component containing the maximum PGA and the vertical component. Using the algorithm discussed earlier, the nonlinear response of the frames is estimated for 2 and 5% of the critical damping (x ), in terms of the maximum lateral displacement at the top of the frame DMAX, the maximum axial loads, and the maximum bending moments at the ground level for the interior and exterior columns. All the frames are also subjected to static applications of the dead load as suggested in the NEHRP Provisions, and the corresponding responses are evaluated. For comparison purposes, the error term is dened as: E code specified value analytical result code specified value 7

Considering the dead load, earthquake load, and load combination suggested in ASCE 7-95 [4], Eq. (7) can be rewritten for the NEHRP Provisions as:

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145 Table 3 Maximum top displacements (DMAX) for frame 1 EAR (1) 1 2 3 4 5 6 7 8 9 10 11 12 13 x (2) 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 H (cm) (3) 2.88 2.05 6.92 5.38 1.82 1.72 6.39 4.54 1.79 1.49 3.79 3.08 1.08 0.81 4.26 2.65 7.44 4.74 2.32 1.78 1.96 1.52 4.48 3.59 2.74 2.29 1.2D+HV (cm) (4) 2.86 2.06 6.92 5.37 1.82 1.72 6.40 4.54 1.79 1.48 3.78 3.08 1.08 0.82 4.26 2.66 7.46 4.74 2.32 1.78 1.96 1.52 4.47 3.59 2.74 2.28 NEHRP (cm) (5) 2.88 2.05 6.92 5.38 1.82 1.71 6.39 4.54 1.79 1.49 3.79 3.08 1.08 0.81 4.26 2.65 7.44 4.74 2.32 1.78 1.96 1.52 4.47 3.59 2.74 2.29 MEX (cm) (6) 3.46 2.46 8.30 6.46 2.18 2.05 7.67 5.45 2.15 1.79 4.55 3.70 1.30 0.97 5.11 3.18 8.93 5.69 2.78 2.14 2.35 1.82 5.36 4.31 3.29 2.75 ENEHRP (%), Eq. (8) (7) 1 0 1 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0

137

EMEX (%), Eq. (9) (8) 17 16 17 17 17 16 17 17 17 17 17 17 17 15 17 16 16 17 17 17 17 16 16 17 17 17

ENEHRP

1X2D H 0X5Ca D 1X2D HV 1X2D H 0X5Ca D

where the term 1.2D+H + 0.5CaD represents the combined eect of dead load, horizontal seismic load and vertical seismic load according to the NEHRP Provisions; the term 1.2D+HV represents the combined eect of dead load, horizontal and vertical seismic loads according to analytical results obtained by the algorithm discussed in Section 2, H is the eect of the horizontal component containing the maximum PGA acting alone (case 1), 0.5CaD represents the eect of the vertical component, and HV represents the eect of both the horizontal and vertical components acting simultaneously. Similarly, for the Mexican Code, Eq. (7) can be expressed as: EMEX 1X2D H 0X2H 1X2D HV 1X2D H 0X2H 9

Code, and 0.2H represents the eect of the vertical component. All other terms in Eq. (9) were dened earlier. A positive error in Eqs. (8) and (9) implies that the codes overestimate the load eect due to the vertical component; in other words, the codes' recommendations are conservative. A negative error indicates that the codes underestimate the load eect, and thus are unconservative. The responses of the three frames can be compared in light of the error terms just discussed 4.1. Eect of the vertical component of DMAX Frame 1 is considered rst. The DMAX values for the two damping ratios and all 13 earthquakes are shown in Table 3. Column 3 contains analytical DMAX values for excitation by the horizontal component only. Column 4 contains the same information when the frame is subjected to 1.2D plus both the horizontal and vertical components. Columns 5 and 6 contain the combined eect for the DMAX values according to the NEHRP Provisions and the Mexican

where the term 1.2D+H + 0.2H represents the combined eect of deal load, horizontal seismic load and the vertical seismic load according to the Mexican

138

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145

Table 4 Maximum top displacements (DMAX) for frame 1 (plastic case) EAR (1) 1 2 3 4 5 6 7 8 9 10 11 12 13 x (2) 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 Case 1 (cm) (3) 12.59 8.20 12.26 8.08 9.09 8.54 13.75 11.72 8.94 7.44 13.08 9.25 8.66 6.51 11.32 6.62 14.40 7.12 10.26 7.10 9.83 7.63 12.67 12.38 10.66 8.03 1.2D+HV (cm) (4) 13.73 8.2 12.45 8.08 9.10 8.54 13.77 11.72 8.94 7.44 11.35 9.25 8.66 6.51 11.31 6.62 14.23 7.12 10.24 7.10 9.79 7.61 14.11 12.36 10.66 8.07 NEHRP (cm) (5) 12.59 8.20 12.26 8.08 9.09 8.54 13.75 11.72 8.94 7.44 13.08 9.25 8.66 6.51 11.32 6.62 14.40 7.12 10.26 7.10 9.83 7.63 12.67 12.38 10.66 8.03 MEX (cm) (6) 15.11 9.84 14.71 9.70 10.91 10.25 16.50 14.06 10.73 8.93 15.70 11.10 10.39 7.81 13.58 7.94 17.28 8.54 12.31 8.52 11.80 9.16 15.20 14.86 12.79 9.64 ENEHRP (%), Eq. (8) (7) 9 0 2 0 0 0 0 0 0 0 13 0 0 0 0 0 1 0 0 0 0 0 11 0 0 1 EMEX (%), Eq. (9) (8) 9 17 15 17 17 17 17 17 17 17 28 17 17 17 17 17 18 17 17 17 17 17 7 17 17 16

Code, respectively. Using Eqs. (8) and (9), the corresponding error terms are calculated and are shown in columns 7 and 8, respectively. As stated earlier, the frame is designed so that it did not develop any plastic hinges when excited by any of the 13 earthquakes. From the results given in Table 3, several important observations can be made. The maximum analytical horizontal deections of the frame are observed to be almost the same for excitation by the horizontal component alone or by both the horizontal and vertical components. This is expected. Since the frame is symmetric and did not develop any plastic hinges, the eect of the vertical component in the estimation of the horizontal deection is expected to be small. The DMAX values estimated according to the NEHRP Provisions (column 5) are very similar to the analytical results. This is also expected, since the eect of the dead load on the DMAX calculation is negligible. The corresponding error according to Eq. (8) is also negligible (column 7). However, the situation is quite dierent for the Mexican Code. The results in column 6 indicate that the Mexican Code overestimates the DMAX values, and this overestimation is about 17%

(column 8). Thus, for frame 1, the NEHRP Provisions estimate DMAX very accurately, but the Mexican Code overestimates it by about 17%. In order to study the behavior of the same frame subjected to stronger earthquakes, all the earthquake time histories are scaled up so that six to eight plastic hinges develop in the frame. The frame is reanalyzed, and the results in term of DMAX are given in Table 4. For 2% damping, the frame developed six to eight plastic hinges, but remained elastic for 5% damping. For 2% damping, when the structure lost its symmetry due to the development of the plastic hinges in the frame, the NEHRP Provisions underestimated DMAX by more than 11% in some cases, and overestimated by over 13% in other cases. This underestimation or overestimation cannot be correlated with the R parameter. It appears to be problem-specic. As before, the Mexican Code overestimates the DMAX values in this case too; however, the amount of overestimation is, in some cases, smaller than that of Table 3, and could be as small as 7%. For 5% damping when the frame remains elastic, the eect of the vertical component on the NEHRP calculation remains negligible

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145

139

for the NEHRP Provisions, but the Mexican Code overestimates it by about 17%, as before. In summary, whether the frame remains elastic or develops plastic hinges, the Mexican Code always overestimates the DMAX values; however, the NEHRP Provisions could unconservatively underestimate DMAX if plastic hinges develop in the frame. The benecial eect in terms of the reduction in DMAX as a function of damping can also be noted from Tables 3 and 4. However, the amount of reduction varies from earthquake to earthquake and depends on the degree of yielding occurring in the structure. If no plastic hinge develops in the structure, the reduction could be around 20% (Table 3), and if plastic hinges develop, the reductions could be larger than 40% (Table 4). Frames 2 and 3 are considered next. Results similar to Tables 3 and 4 for frame 1 were estimated for frames 2 and 3. They cannot be shown here due to lack of space. The major conclusions made for frame 1 are also valid for these frames. If frames 2 and 3

remain elastic, the error according to the NEHRP Provisions is almost zero, but according to the Mexican Code, the conservative error is about 17%, as before. If plastic hinges develop, the error according to the NEHRP Provisions could be on the unconservative side by about 13% for frame 2, and about 9.5% for frame 3. Thus, the trends are very similar for all three frames. The heights of the frames cannot be correlated with the corresponding errors, particularly when the errors are negative or unconservative.

4.2. Eect of the vertical component on bending moments in columns The eect of the vertical component on the evaluation of the bending moments for the interior and exterior columns at the ground level of frame 1 is considered next. Results for recorded time histories and scaled up time histories, similar to Tables 3 and 4, are shown in Tables 5 and 6, respectively. The

Table 5 Maximum moments at ground level columns for frame 1 EAR x (1) 1 2 3 4 5 6 7 8 9 10 11 12 13 H (kN m) 1.2D+HV(kN m) NEHRP (kN m) Interior Exterior (7) (8) 769 547 1768 1362 488 454 1528 1085 463 385 991 803 263 198 1006 627 2862 1192 573 442 482 374 1197 962 702 587 516 373 1161 899 335 313 1005 720 318 268 660 538 187 145 668 423 1222 789 389 304 333 263 793 641 473 398 MEX (kN m) ENEHRP (%), Eq. (8) EMEX (%), Eq. (9) Interior Exterior (13) (14) 17 17 17 17 17 17 17 17 17 16 17 17 16 16 17 17 17 17 17 17 16 16 17 17 17 17 16 16 16 17 16 16 16 16 16 16 16 16 15 15 16 16 17 16 16 16 15 15 16 16 16 16

Interior Exterior Interior Exterior (2) (3) (4) (5) (6) 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 767 545 1766 1360 486 452 1526 1083 461 383 989 801 261 196 1004 625 1860 1190 571 440 480 372 1195 960 700 585 496 353 1141 879 315 293 985 700 298 248 640 518 167 125 648 403 1202 769 369 284 313 243 773 621 453 378 769 547 1769 1362 488 454 1528 1086 463 385 991 803 264 198 1006 626 1863 1193 572 441 486 377 1197 961 703 587 513 369 1158 895 332 310 1002 717 315 165 656 534 184 142 666 420 1218 786 386 301 332 261 790 637 470 395

Interior Exterior Interior Exterior (9) (10) (11) (12) 922 656 2121 1634 585 544 1833 1301 555 461 1188 963 315 237 1206 752 2234 1430 687 530 578 448 1436 1154 842 704 612 440 1386 1073 395 368 1199 857 374 314 785 638 217 167 794 500 1459 940 460 358 392 308 944 762 560 470 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 1 1 0 0 1 1 0 0 1 1 1 1 2 2 0 1 0 0 1 1 0 1 0 1 1 1

140

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145

Table 6 Maximum moments at columns for frame 1 (plastic case) EAR x (1) 1 2 3 4 5 6 7 8 9 10 11 12 13 H (kN m) 1.2D+HV(kN m) NEHRP (kN m) Interior Exterior (7) (8) 3295 2180 2836 2042 2439 2261 3130 1952 2307 1916 3182 2045 2074 1568 2607 1563 3099 1787 2433 1760 2525 1860 3307 2305 2537 2048 2129 1429 1922 1338 1595 1484 2186 1279 1512 1258 2016 1341 1358 1024 1714 1028 2066 1174 1597 1156 1654 1234 2037 1510 1657 1343 MEX (kN m) ENEHRP (%), Eq. (8) EMEX (%), Eq. (9) Interior Exterior (13) (14) 14 17 17 17 17 17 16 17 17 17 23 17 17 16 16 17 12 17 17 17 16 16 13 17 16 17 18 17 17 16 16 17 19 17 17 16 19 17 16 17 16 16 13 16 17 17 16 16 15 17 16 16

Interior Exterior Interior Exterior (2) (3) (4) (5) (6) 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 2 5 3293 2178 2834 2040 2437 2259 3128 1950 2305 1914 3180 2043 2072 1566 2605 1561 3097 1785 2431 1578 2523 1858 3305 2303 2535 2046 2109 1409 1902 1318 1575 1464 2166 1259 1492 1238 1996 1312 1338 1004 1694 1008 2046 1154 1577 1136 1634 1214 2017 1490 1637 1323 3394 2180 2833 2043 2441 2263 3139 1952 2298 1916 2929 2045 2075 1572 2618 1563 3262 1789 2428 1759 2545 1879 3467 2304 2347 2051 2098 1425 1917 1335 1593 1481 2124 1275 1507 1255 1948 1337 1356 1020 1719 1025 2148 1171 1591 1152 1661 1236 2078 1507 1658 1341

Interior Exterior Interior Exterior (9) (10) (11) (12) 3953 2615 3402 2450 2926 2712 3755 2342 2768 2298 3818 2453 2488 1881 3128 1875 3718 2144 2919 2111 3029 2231 3968 2765 3044 2457 2548 1708 2299 1598 1907 1774 2616 1528 1807 1502 2412 1602 1622 1222 2050 1226 2472 1402 1909 1380 1978 1474 2437 1805 1981 1604 3 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0 5 0 0 0 1 1 5 0 0 0 1 0 0 0 0 0 3 0 0 0 3 0 0 0 0 0 4 0 0 0 0 0 2 0 0 0

major observations made for the DMAX calculations are also valid for the estimation of moments at the ground level of columns. If the frame remains elastic, the errors in the bending moment calculations according to the NEHRP Provisions are almost zero for both interior and exterior columns. However, when plastic hinges develop, the underestimation could be about 5%. The Mexican Code always overestimates the bending moments; the corresponding overestimation errors are about 17% when the frame remains elastic, and as low as 12% when plastic hinges develop. Frames 2 and 3 were similarly analyzed, and the corresponding errors are almost identical to Frame 1. The results are not shown due to lack of space. As in the DMAX evaluation, the Mexican Code is conservative in the estimation of bending moments, but the NEHRP Provisions could underestimate this eect if plastic hinges develop. These errors have no correlation with the R parameter or with the height of the frame.

4.3. Eect of the vertical component on axial loads in columns The eect of the vertical component on the evaluation of the maximum axial loads at interior and exterior ground level columns for all the frames is considered next. The estimation errors according to both codes are calculated using Eqs. (8) and (9) for both interior and exterior columns. For ease of discussion, the errors versus R are plotted. Only underestimation of the axial load with errors larger than 25%, which occurs for the interior column only, is emphasized in the following discussion. Other results cannot be shown due to lack of space. The results for the interior column of frame 1 are shown in Fig. 2 for the elastic case and in Fig. 3 when plastic hinges develop in the frame. Unlike the DMAX and the bending moment evaluation cases, the eect of the vertical component on the axial estimation is observed to be signicant, even when the frame remains elastic. The underestimation error could be very large, on the order of 50% for 2% damping and elastic behavior

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145

141

Fig. 2. Error in the axial load on the interior column of frame 1, elastic case.

Fig. 3. Error in the axial load on the interior column of frame 1, elastic case.

142

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145

Fig. 4. Error in the axial load on the interior column of frame 2, elastic case.

Fig. 5. Error in the axial load on the interior column of frame 2, elastic case.

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145

143

according to the NEHRP Provisions, and about 70% according to the Mexican Code. If inelastic behavior is considered, the corresponding errors increase to about 150% for both codes. It is also observed from the above gures that for a given code, the error is always larger for 2% damping than for 5% damping. However, this observation is not valid if the structure develops plastic hinges (Fig. 3). It is interesting to note that the magnitude of the unconservative error increases as the R value increases. Frame 2 is considered next. The underestimation error for the interior column is shown in Figs. 4 and 5 for the elastic and plastic cases, respectively. The major observations made for frame 1 apply to frame 2. If the frame remain elastic, the unconservative error associated with the Mexican Code is greater than that of the NEHRP Provisions. This observation is not valid if plastic hinges develop. Although the NEHRP Provisions are better than the Mexican Code for the elastic case, the unconservative error associated with it may not be acceptable. it is also observed from Figs. 4 and 5 that the magnitude of the unconservative error

is not a function of the height of the frame; however, it increases as the R value increases. Frame 3 is analyzed last. The results for the interior column are shown in Fig. 6, when plastic hinges develop in the structure. The major observations made for frames 1 and 2 are valid for this frame too. The only additional observation is that, unlike frames 1 and 2, both codes are conservative for the interior column when R values are smaller than about 0.6. For larger values of R, however, they again considerably underestimate the axial force. 4.3.1. Design implications The eect of the vertical component on the axial load evaluation is signicant for all three frames considered in this study. All these members are expected to be designed as a beamcolumn, and the exact form of Eq. (4), according to the AISC LRFD code, is: Pu 8 Mux 1X0; fPn 9 fb Mnx if Pu r0X2 fPn 10

Fig. 6. Error in the axial load on the interior column of frame 3, plastic case.

144

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145

Pu Mux 1X0; 2fPn fb Mnx

if

Pu ` 0X2 fPn

11

where Pu is the required axial strength, Pn the nominal axial strength, Mux and Mnx are the required exural and the nominal exural strength with respect to the major axis, respectively, f is the resistance factor for compression (or tension) and fb is the resistance factor for exure. Although the eect of the vertical component on the moment calculation is negligible, it may increase the axial load signicantly. Since both the axial load and moment are considered in the interaction equations, the increase in the axial load will have a very detrimental eect on the performance of the columns. This observation indicates the need for modication of the way the eect of the vertical component is considered in design codes, or at least indicates the need for further study. If the R values are greater than the value usually considered to be normal, say 2/3, the underestimation increases as the R value increases. The underestimation error is observed to be more for interior columns than for exterior columns, indicating that the location of the columns may also be important. The underestimation error also depends on the elastic or plastic state of the frames, but no correlation is observed between the underestimation error and the height of the frames.

timation is more for interior columns than for exterior columns. If the ratio R of the PGA of the vertical and horizontal components of an earthquake is more than is usually considered to be normal, the underestimation increases as R increases. The underestimation can not be correlated with the height of the frames. Although the eect of the vertical component in the moment calculation of columns is negligible or conservative in most cases, it may increase the axial load signicantly. Since they are designed as beamcolumns, the increase in the axial load will have a very detrimental eect on the performance of the columns. In light of the results obtained in this study, the design requirements for the vertical components, as outlined in the NEHRP Provisions and the Mexican Code, need modication. At the very least, further study is required.

Acknowledgements This paper is based on work partly supported by the National Science Foundation under grant nos MSM8896267 and CMS-9526809. The nancial support received from the American Institute of Steel Construction (AISC), Chicago, is appreciated. The work is also partially supported by El Consejo Nacional de Ciencia y Tecnologia (CONACYT), Mexico, and La Universidad Autonoma de Sinaloa (UAS), Mexico. Any opinions, ndings, conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reect the views of the sponsors.

5. Conclusions The eect of the vertical component on the seismic responses of frames, as outlined in the NEHRP Provisions and the Mexican Codes, is reevaluated. Using a time domain nonlinear nite element program developed by the authors, the seismic responses of frames, in terms of the maximum lateral displacement at the top of the frame and the maximum axial and bending moments in columns, are evaluated as realistically as possible by applying the horizontal and vertical components of earthquake motion simultaneously. Three steel frames and 13 recorded earthquake motions are considered in the study. The same response parameters are then estimated using the two codes, and the error associated with their recommendations is evaluated. Several important observations are made. If the frames remain elastic, the NEHRP provisions estimate DMAX and the bending moments very accurately; however, the Mexican Code overestimates them. If the frames develop plastic hinges, the Mexican Code still conservatively overestimates them, but the NEHRP Provisions underestimate them in some cases. Both codes signicantly underestimate the axial loads in columns. The underestimation increases as the frames develop plastic hinges. Also, the underes-

References
[1] Salmon CG, Johnson JE. Steel structures: design and behavior, 4th ed. New York: HarperCollins College, 1996. [2] NEHRP Recommended provisions for seismic regulations for new buildings, FEMA 222A 1994. [3] Manual de Diseno de Obras Civiles, Diseno por Sismo. Comision Federal de Electricidad, 1993. [4] Minimumdesign loads for buildings and other structures. American Society of Civil Engineers, 1995 ASCE 7-95. [5] Gao L, Haldar A. Non linear seismic analysis of space structures with PR connections. International Journal of Microcomputers in Civil Engineering 1995;10:2737. [6] Haldar A, Nee K-M. An ecient algorithm for nonlinear seismic response of multiple-support excitations. In: Tenth International Conference on Structural Mechanics in Reactor Technology (SMiRT), vol. K, 1989. p. 35560. [7] Haldar A, Nee K-M. Elasto-plastic large deformation analysis of PR steel frames for LRFD. Computers & Structures 1989;34(5):81123. [8] Kondoh K, Atluri SN. Large deformation, elasto-plastic analysis of frames under non-conservative loading, using

A.R. Salazar, A. Haldar / Computers and Structures 74 (2000) 131145 explicitly derived tangent stiness based on assumed stress. Computational Mechanics 1987;2(1):125. [9] Shi G, Atluri SN. Elasto-Plastic large deformation analysis of space-frames. International Journal for Numerical Methods in Engineering 1988;26:589615. [10] Owen OR, Hinton E. Finite element in plasticity: theory and practice. Swansea: Pineridge Press, 1982. [11] Clough RW, Penzien J. Dynamic of structures, 2nd ed. New York: McGraw Hill, 1993.

145

[12] Bathe KJ. Finite element procedures in engineering analysis. Englewood Clis, NJ: Prentice-Hall, 1982. [13] Mahadevan S, Haldar A. Stochastic FEM-based evaluation of LRFD. Journal of the Structural Engineering Division, ACSE 1991;117(5):1393412. [14] Leger P, Dussault S. Seismic-energy dissipation in MDOF structures. Journal of Structural Division, ASCE 1992;118(5):125169.