Professional Documents
Culture Documents
(Asce) Be 1943-5592 0000275
(Asce) Be 1943-5592 0000275
Abstract: The displacement-based seismic-design (DSD) methods, owing to their simplicity and efficiency, have been more and more
recognized in structural seismic-research communities during the past few years. However, the dynamic P-Δ effect, as has long been
well realized to be a key issue in structure earthquake engineering that could amplify the structure’s seismic responses or even trigger
the structure’s instability, is still not well solved practically because of the complicated nonlinear mechanism. Therefore, in aim to achieve
a practical and general purpose solution to consider the P-Δ effects in various DSD methods for single-degree-of-freedom (SDOF) systems,
first, the existing approaches of considering P-Δ effects in current seismic analysis and design were evaluated by carrying out a large set
of nonlinear time-history analyses, and then new design formulas and recommendations on threshold of neglecting P-Δ effects and the
allowable design thresholds were promoted on the basis of the statistics data. At last, the proposed procedure was illustrated by a
seismic-design example. DOI: 10.1061/(ASCE)BE.1943-5592.0000275. © 2012 American Society of Civil Engineers.
CE Database subject headings: Seismic design; Regression analysis; Displacement; Bridges.
Author keywords: Seismic design; Dynamic stability; P-Δ effects; SDOF system; Displacement-based seismic design; Regression
analysis.
F yp M yp ∕H M yp
λF ¼ ¼ ¼ ð11Þ
Fig. 1. Effects of P-Δ on a SDOF system F y0 M y0 ∕H M y0
Using stability indices defined in Eqs. (6) and (7), λF;pa and λF;pr
can be rewritten as
1þμ
λF;pa ¼ θe ð17Þ
2
θΔ
λF;pr ¼ ð18Þ
2
Downloaded from ascelibrary.org by Pontificia Universidad Catolica del Peru on 04/13/20. Copyright ASCE. For personal use only; all rights reserved.
Fig. 6. Statistical curves of λΔ in the inelastic range and different prediction curves
mean of λΔ , especially in the engineering allowable stability index Fig. 7 as well. As demonstrated in this figure, formulas proposed
range, where a serious underestimation of real P-Δ effects occurs. by Rosenblueth (1965) and Priestley et al. (2007) are independent
Because Rosenblueth’s formula does not change with ductility fac- of ductility factor μ, whereas curves for Bernal’s (1987) and
tors, the underestimation gets larger as the ductility factor increases. Paulay’s (1978) formulas become lower with the increase of μ.
This implies that simply extending elastic design approaches to the All these trends are not coincided with the real mean curves except
inelastic range is not authentic. the expression of Eq. (21). It was found that Rosenblueth’s formula
Strength Increasing Coefficient λF lacks conservatism under all ductility values, illustrating again that
it is inappropriate to simply extend the elastic strength increasing
Statistical results of strength increasing coefficient λF are shown in coefficient expression to the inelastic range.
Fig. 7. It is obvious that statistics mean curves of λΔ and λF differ Intended to provide a simple compensation method for P-Δ
largely with one another, so the current code approach that does not effects to be directly used in DSD, however, Priestley’s suggestion
distinguish the P-Δ effect formulas for λΔ and λF is misleading. gives a serious underestimation of the mean statistical value.
As shown in Fig. 7, in the engineering allowable range, mλF is
Paulay’s formula produces the smallest prediction among all the
basically a linear function of θΔ , with the line slope becoming
expressions. It can be inferred that it is inadequate to handle the
larger as ductility μ increases, but remaining constant after μ is
larger than 3. A regression analysis of statistics mean of λF is per- problem of P-Δ effects under seismic action in the inelastic range
formed when the structure exhibited a ductility factor within the only from the perspective of energy. Bernal’s formula was re-
range 1.5 to 6, which yields gressed with the analysis data from four ground motions and with
underlying force-based seismic-design concept. As discussed pre-
λF ¼ 1:05 þ ½2 e0:85ðμ1Þ θΔ ð21Þ viously, in the force-based seismic-design concept, the maximum
displacement after the P-Δ compensation will be larger than the
When θΔ equals to zero, the theoretical value of λF should be 1.0;
target value, implying that the compensation is not enough. This is
however, in Eq. (21) a slightly larger value 1.05 was gained for the
confirmed by Fig. 7. Comparatively, the formula proposed in this
reason that regression weights of data in a small nonlinear zone
when θΔ is near zero were reduced to get a concise and con- study coincides well with statistical results. Because of the random-
servative expression. Formulas of strength increasing coefficient ness of the seismic action, to take the mean strength increase λF
proposed in this study and by former researchers are drawn in does not guarantee the resulted mean ductility equals to the design
one. However, as shown later in the example, even with much static (pushover) analysis to the target displacement, θe is permitted
parameter uncertainties, results are generally satisfactory. to exceed 0.10. Using the Takeda hysteretic model for reinforced
concrete structures, Priestley et al. (1996) contended that to obtain
stable structural response without producing significant P-Δ dis-
Recommendations on Design Thresholds for the placement, the stability index θΔ should be less than 0.30.
P-Δ Effects As is shown in Fig. 7, when θΔ is relatively small, statistics
mean mλF is larger than all the predictions made by the formerly
In real earthquake engineering practices, two issues are typically of proposed equations. Hence, if the previously discussed criterion for
special concern: threshold of safely ignoring P-Δ effects and upper neglecting P-Δ effects is employed, i.e., the required strength am-
design limit for P-Δ effects. The former limit can be determined by plification to achieve a particular ductility should be less than 10%,
restraining the amplification requirements, whereas the latter one is a more stringent limit will be obtained. Synthesizing the statistical
governed by collapse-prevention criteria. data of various ductility factors in Fig. 7, the threshold for neglect-
On the basis of analyses of multistory reinforced concrete ing P-Δ effects in the seismic design is determined as
frames, Paulay (1978) recommended conditions when P-Δ effects
were small enough to be ignored. For SDOF systems, Pauley’s θΔ ≤ 0:025 ð23Þ
approach indicated that P-Δ effects could be neglected if θΔ <
0:15. Bernal (1987) and Mahin and Boroschek (1991) proposed which is only 1∕6, 1∕8, and 1∕2:5 (on average) of the threshold
that if the required strength amplification to achieve a specific duc- proposed by Paulay (1978), Mahin and Boroschek (1991), and
tility was less than 10%, then P-Δ effects could be ignored. Using Bernal (1987), respectively. This implies that P-Δ effects exert
this criterion, Mahin and Boroschek suggested the threshold of more influences on the SDOF structure than expected before;
θΔ < 0:20, while Bernal gave hence, prudent judgment should be made in determining whether
θe < 0:1∕ðβ þ 1:1Þ; β ¼ 1:87ðμ 1Þ ð22Þ the P-Δ effects can be neglected.
When the structure is designed with a ductility factor of 6, the
The design tolerance for P-Δ effects given by FEMA 450 design limit by FEMA 450 (BSSC 2003) corresponds to θΔ ≤ 0:60.
(BSSC 2003) is θe ≤ 0:10, but if the resistance to lateral forces As shown in Fig. 6 that if the ductility is 6, the structure collapse
is determined to increase continuously in a monotonic nonlinear ratio η is nearly 100% when θΔ gets close to 0.6; hence, regulations
will be above 2; thus, the structural ductility under the MCE will be To verify the design results, inelastic time-history analyses are
above 3. As shown in Fig. 6, the collapse ratio η is nearly invariant performed using 20 artificial acceleration time histories, which are
when ductility is larger than 3. Therefore, on the basis of statistical compatible with the design displacement spectrum shown in Fig. 8.
results shown in Fig. 6, it is recommended that the collapse- Given the column design strength, reinforcement is configured for
prevention requirements will be satisfied if the section, and the real section yield curvature can be obtained
from the moment-curvature analysis of the section. The EPP hys-
θΔ ≤ 0:15 ð24Þ teretic rule is selected for consistency. The design and validation
procedure will be similar if other hysteretic rule-based formulas
When in the elastic situation, the FEMA 450 criterion that θe ≤ are used. The average of the 20 analysis results is regarded as the
0:10 is still recommended. “actual” structure response. Refer to Appendix S1 for detailed
information of the design verification.
From Table 2, one can see that, except for very short period
Design Example range as in case (a), the original DDSD provides good results when
not considering the P-Δ effects. But, when considering the P-Δ
Applications of the regression formulas and design recommenda- effects, it fails to comply well with the design displacement. The
tions are illustrated by the design of a simple bridge pier. The direct displacement amplification factors as shown are seriously underes-
displacement-based seismic-design method (DDSD) (Priestley et al. timated in current seismic codes (as proposed by Rosenblueth), but
2007) is chosen in the illustration, while it is expected that the pro- relatively close to the predictions obtained using the formulas of
cedure will equally apply to other DSD methods. The bridge is this paper. Among the procedures studied, the procedure proposed
located in a region of high seismicity [peak ground acceleration in this paper produces the best compliance with the target displace-
(PGA) = 0.7 g], and the design displacement spectrum is taken ment when considering the P-Δ effects.
from Eurocode 8 (CEN 2001) for firm ground condition. Key de- The collapse ratio of each compensation case under the MCE
sign parameters are summarized in Fig. 8. Five cases are studied to level as defined in FEMA P695 is also presented in Table 2.
cover a wide range of interested structures (structure period from As shown, apart from the P-Δ design on the basis of the proce-
0.17 to 2.93 s). The design limit state is represented by the more dure in this paper, the FEMA P695 collapse-prevention criterion
critical requirement of a displacement ductility of μ ¼ 4 or a drift that limits the probability of collapse to 10% cannot be satisfied.
of 0.035. Basic procedures of standard DDSD are taken first (see This would be the more serious consequence of underestimating
Appendix S1), following with the P-Δ effect design. For compari- the P-Δ compensation value in previous procedures.
son, three P-Δ design procedures are considered: the noniterative
one on the basis of Eqs. (20), (21), (23), and (24), the one proposed
by Rosenblueth and used in current codes, and the one proposed by Conclusions
Priestley et al. to be used in DDSD. Design results are listed in
Table 2. In an effort to obtain a more convenient procedure to consider P-Δ
In case (a), the stability index is calculated as θΔ0 ¼ PΔd ∕ effects in the DSD methods, comprehensive nonlinear time-history
M Base ¼ 0:012 < 0:025, within the limit for neglecting P-Δ analyses of SDOF systems with and without P-Δ effects were per-
effects. In this paper, the design ends, with the P-Δ effects directly formed in this paper, and new design formulas were regressed.
neglected. Thresholds of stability indices to neglect P-Δ effects and maximum
3 P-Δ design 1 Displacement amplification factor λΔ — 1.03 1.14 1.50 1.81 1.73
(procedure in this paper) Strength increasing coefficient λF — — 1.14 1.30 1.42 1.43
New stability index θΔp — — 0.04 0.10 0.15 0.17
P-Δ design 2 Displacement or strength λΔ ðλF Þ — — 1.05 1.15 1.28 1.33
(procedure by Rosenblueth) amplification factor
New stability index θΔp — — 0.05 0.11 0.17 0.19
P-Δ design 3 Strength increasing coefficient λF — — 1.02 1.07 1.11 1.12
(procedure by Priestley et al.) New stability index θΔp — — 0.05 0.12 0.20 0.22
4 Lateral displacement verification Initial design without P-Δ Δ0 m 0.020 0.135 0.250 0.414 0.610
Initial design with P-Δ Δp0 m 0.021 0.176 0.364 0.833 0.890
P-Δ design 1 Δ0p1 m — 0.135 0.271 0.439 0.582
P-Δ design 2 Δ0p2 m — 0.145 0.315 0.510 0.650
P-Δ design 3 Δ0p3 m — 0.152 0.320 0.628 0.757
Ratio of Δ0 to Δd Δ0 ∕Δd — 0.68 1.15 1.09 0.99 1.09
Displacement amplification factor Δp0 ∕Δ0 — 1.05 1.30 1.46 2.01 1.46
Ratio of Δ0p1 to Δd Δ0p1 ∕Δd — — 1.15 1.18 1.05 1.04
Ratio of Δ0p2 to Δd Δ0p2 ∕Δd — — 1.24 1.37 1.21 1.16
Ratio of Δ0p3 to Δd Δ0p3 ∕Δd — — 1.30 1.39 1.50 1.35
Collapse ratio Initial design ηp0 % 0 0 55 95 80
P-Δ design 1 ηp1 % — 0 0 10 10
P-Δ design 2 ηp2 % — 0 25 25 20
P-Δ design 3 ηp3 % — 0 45 70 65
allowable design limits were also determined. The results were The researches done in this paper apply only for SDOF systems
validated by a seismic-design example. The major advantage of the with EPP hysteretic rule. For other hysteretic rules, design formulas
proposed procedure is that it applies to various DSD methods, and and thresholds of the P-Δ effects can be derived in a similar
no modification of their basic formulations is needed to include the manner, while the results provided in this research are expected
P-Δ effects. to be the upper bound for them. For multiple-degrees-of-freedom
Conclusions of this paper can be drawn as follows: (MDOF) systems, the P-Δ problems may be solved in a story-
• The influences of P-Δ effects in the inelastic range differ much by-story manner (CEN 2001) or by the equivalent SDOF structure
from the elastic case, revealing that simply extending elastic approach (Adam et al. 2004). Research efforts are encouraged to be
approaches to the inelastic range is inappropriate. The stability made on these topics in the future.
index should be selected according to the structure’s re-
sponse state. Acknowledgment
• The statistics mean curves of λΔ and λF differ a lot from one
another, which implies that it could be a concept confusion of The financial support of the National Natural Science Foundation
current approaches to use the same formula to deal with both the of China (90815007) is gratefully acknowledged.
response amplification and design compensation problems.
When θΔ ≤ 0:05, λF is larger than λΔ , and beyond this range
the opposite trend exists. Supplemental Data
• The effects of P-Δ predicted by current approaches are all less
than the statistics mean values within practical allowances, Appendix S1, “Details of the design verification,” is available
which indicate the P-Δ effects could be underestimated accord- online in the ASCE Library (www.ascelibrary.org).
ing to current seismic codes.
• On the basis of nonlinear time-history analyses using EPP
References
hysteretic model, it is recommended that P-Δ effects in the
nonlinear seismic design can be neglected when θΔ is less Adam, C., Ibarra, L. F., and Krawinkler, H. (2004). “Evaluation of p-delta
than 0.025, and an upper limit of 0.15 for stability index θΔ effects in non-deteriorating MDOF structures from equivalent SDOF
is suggested for collapse-prevention requirements. systems.” Proc., 13th World Conf. on Earthquake Engineering