Professional Documents
Culture Documents
Substitute-Structure Method
Tim Huff, Ph.D., P.E.1; and Shahram Pezeshk, Ph.D., P.E., F.ASCE2
Abstract: The design of bridge structures for seismic loading by displacement-based procedures has become preferred over force-based pro-
cedures in modern specifications. One key step in displacement-based design (DBD) is the estimation of inelastic displacement. Current
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 05/13/16. Copyright ASCE. For personal use only; all rights reserved.
AASHTO specifications rely on a linear response spectrum analysis with an amplification factor, Rd, for conventionally designed bridges and
on the substitute structure method (SSM) for analysis of isolated bridges to estimate inelastic displacements. Both methods are used in this
study for a selected site and target acceleration response spectrum. The procedures for both are outlined in detail. Displacement spectra from
nonlinear response history analyses (NLRHA) are generated for seven ground-motion sets and compared to results from the two simplified
methods. The SSM is shown to produce inelastic displacement estimates that more closely match NLRHA results than do results from the
AASHTO Rd method over a wide range of periods for the specific conditions of (1) a large modal magnitude earthquake, (2) a class D subsur-
face profile, (3) accelerations characteristic of the selected site, and (4) reduced initial damping to mimic tangent-stiffness-damping. DOI:
10.1061/(ASCE)SC.1943-5576.0000279. © 2015 American Society of Civil Engineers.
Author keywords: Earthquakes; Seismic analysis; Ground motion; Inelastic; Displacement spectra.
Introduction The purposes of this study are to (1) demonstrate that the
AASHOT Rd method may underestimate inelastic response over a
Current AASHTO (2011) estimates of inelastic displacement for wide range of periods for large magnitude ground motions on Site
bridges of conventional design (nonisolated) rely on the equal dis- Class D profiles, and (2) propose alternative methods for estimating
placement rule (EDR). EDR states that the inelastic displacement of inelastic displacement spectra.
a structure is equal to the displacement that the structure would To accomplish these objectives, a detailed discussion of the
have experienced if all elements remained in the elastic range, and AASHTO Rd method is provided. The SSM is fully described and
is complemented in AASHTO with a short-period amplification other methods proposed in the literature are discussed. Important
factor, Rd. For isolated bridges, AASHTO (2010) adopts the substi- issues related to damping levels for both SSM and NLRHA are
tute structure method (SSM) of analysis. SSM uses the secant stiff- highlighted.
ness to determine an effective period. Hysteretic behavior is treated Seven ground-motion sets were scaled to a target response spec-
as added viscous damping. trum for a hypothetical bridge site on an AASHTO Class D subsur-
Other methods have been proposed for estimating the inelastic face profile. NLRHA analyses were performed on each set to
displacement without resorting to nonlinear response history analy- develop the mean inelastic displacement spectra. The Rd- and
sis (NLRHA). Among these are methods that propose rules for a SSM-based inelastic spectra were generated from the same target
displacement amplification factor, Cm , to be applied to the elastic spectrum and compared to the results from NLRHA.
displacement as an estimate for inelastic displacement. More
recently, Bozorgnia et al. (2010) proposed a ground-motion predic-
tion model to estimate the inelastic displacement. Parameters Cm AASHTO Rd Method for Conventional Design
and Rd are directly analogous, simply different terminology.
In a recent study, Khose and Singh (2014) reported that the Given the complexity of performing NLRHA and the potential dif-
equivalent linearization methods for inelastic displacement, in the ficulty in interpreting the results, it is certainly desirable to have a
displacement-controlled spectral range, may underestimate response simplified, yet accurate, method of estimating inelastic displace-
results. Further research will be helpful in developing tools to esti- ment. Response spectrum techniques are the method of choice in
mate inelastic response through simplified analytical methods in the most engineering offices. This is not likely to change soon.
design office. Nonlinear behavior results in energy dissipation, which histori-
cally has been treated as added equivalent viscous damping. The
stiffness is not constant during loading for nonlinear systems. The
issue then becomes determining which values for effective damping
1
Civil Engineering Manager, Tennessee DOT, 505 Deaderick St., and effective stiffness, when used in a response spectrum analysis,
Suite 1100, Nashville, TN 37243 (corresponding author). E-mail: tim will produce results similar to those obtained in NLRHA.
.huff@tn.gov One option for engineers is to use the initial stiffness and the
2
Professor, Dept. of Civil Engineering, Univ. of Memphis, Memphis,
elastic, initial-stiffness-based viscous damping in a linear response
TN 38152. E-mail: spezeshk@memphis.edu
Note. This manuscript was submitted on June 8, 2015; approved on
spectrum analysis to estimate nonlinear response. In this case, the
October 13, 2015; published online on December 30, 2015. Discussion pe- assumption implies that for a given initial stiffness and a given
riod open until May 30, 2016; separate discussions must be submitted for ground motion, a yielding structure will experience the same dis-
individual papers. This paper is part of the Practice Periodical on placement as a nonyielding structure. This equal displacement rule
Structural Design and Construction, © ASCE, ISSN 1084-0680. (EDR) assumption has been used in practice extensively and has
2ð m 1Þð1 aÞ 2
j hys ¼ (9) ðSAEL ÞTEFF TEFF
p m ð1 þ am aÞ ðSDINEL ÞT ¼ g
BL 2p
2
Eq. (9) is based on a single cycle at maximum displacement. TEFF
¼ Rj ðSAEL ÞTEFF g (16)
However, during a large portion of ground shaking, structures 2p
respond at much less than the maximum displacement, so the effec-
tive hysteretic component may logically need to be taken as some- where (SAEL)TEFF = elastic spectral acceleration at period TEFF.
thing less than that given by a theoretical expression [Eq. (9) for this Note that for TEFF, Eq. (15) is valid only in the case of a bilinear
model]. Typically, the Dwairi model results in a lower damping oscillator.
value than the bilinear model. Table 1 lists the m and a combina- For either the AASHTO Rd method or the SSM, an inelastic dis-
tions used in this study, along with the calculated damping values placement spectrum may be generated given an elastic acceleration
for the SSM analysis. Other parameters in Table 1 will be discussed spectrum. This elastic acceleration spectrum could be the mean of a
later. suite of records or the code-based design response spectrum for a
Once the effective damping has been established, a model for given site.
response modification due to the increased damping is needed. A curious aspect of the SSM is the predicted solution at low duc-
AASHTO has its own models and several others have been pro- tility values. The method predicts inelastic displacements smaller
posed (Priestley et al. 2007). Three of the models are given here, than elastic displacements over a wide range of periods when the
along with the AASHTO models, in Eqs. (10)−(14). Eq. (10) is the ductility is reduced to a value dependent on the postyield stiffness
AASHTO model for bridges with isolation devices, which are typi- ratio, a. When the effective period is within the constant velocity
cally characterized by a classic bilinear load-deflection hysteretic region of the target spectrum (i.e., when SAEL is proportional to
curve. Eq. (11) is the AASHTO model for conventionally designed 1/T), the inelastic response displacement can be obtained as given
bridges, such as those relying on plastic hinging in columns. Eq. by Eq. (17).
(12) is the EC8 model for far-field conditions and Eq. (13) has 2 2
been proposed for sites where pulse-type ground motions are SD1 λ1 T SD1 T
ðSDINEL ÞT ¼ g Rj ¼ g λ1 R j
expected. Eq. (14) is an early model based on work by two of λ1 T 2p T 2p
the pioneers in the field of earthquake engineering, Professors (17)
Nathan M. Newmark and William J. Hall.
0:30 Eq. (17) can be simplified to Eq. (18).
j EFF
BL ¼ 1:70; AASHTO ð2010Þ (10)
0:05 ðSDINEL ÞT ¼ ðSDEL ÞT λ1 Rj (18)
While the purpose of this paper is not to synthesize a set of ground motions for design. Further assistance in obtaining infor-
design ground motions for the hypothetical site, either of the last mation for a site may be found with (1) hazard deaggregations
two sets might arguably be interpreted as such, given that match (USGS 2015) to establish MW, R, and SAEL data; and (2) the
to spectral shape was identified in a recent study (NEHRP OpenSHA platform at OpenSHA.org (Field et al. 2003) for
Consultants Joint Venture 2011) as the single most important fac- inferred VS30 values.
tor in ground-motion selection at far-field sites. Other important For each selected record pair, scaling was performed on the
factors include moment magnitude (MW), source-to-site distance geometric mean of as-recorded, horizontal components to mini-
(R), tectonic setting (active tectonic versus stable continent), and mize the mean-square-error (MSE) at periods between 0.09 and
site conditions (VS30, profile depth, etc.). The NEHRP report is an 10.0 s. This is a much wider range than would normally be used in
excellent reference for engineers who need to select and scale practical ground-motion selection and scaling. With 10 record
NGA number Event Station Rjb (km) Rrup (km) VS30 (m/s) SF
1190 Chi-Chi, Taiwan CHY019 50 50.5 478.3 6.2271
1838 Hector Mine Whitewater Trout Farm 62.9 62.9 345.4 7.1297
1521 Chi-Chi, Taiwan TCU089 0 8.9 680.0 1.3418
1284 Chi-Chi, Taiwan HWA035 44 48.4 500.8 5.4097
1153 Kocaeli, Turkey Botas 126 127 274.5 4.6694
719 Superstition Hills-02 Brawley Airport 17 17 208.7 3.4581
2112 Denali, Alaska TAPS Pump Station #08 104.2 104.9 424.8 9.1359
880 Landers Mission Creek Fault 27 27 345.4 3.7383
1214 Chi-Chi, Taiwan CHY057 56.7 56.9 411.5 6.7623
169 Imperial Valley-06 Delta 22 22 274.5 1.2269
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 05/13/16. Copyright ASCE. For personal use only; all rights reserved.
NGA number EQ Name Station name Rjb (km) Rrup (km) VS30 (m/s) SF
5 Kern Co Taft Lincoln 38.42 38.89 385.4 2.5013
1149 Kocaeli Atakoy 56.49 58.28 310.0 3.0919
1158 Kocaeli Duzce 13.6 15.37 281.9 0.9681
1177 Kocaeli Zeytinburnu 51.98 53.88 341.6 3.2370
1211 Chi-Chi CHY052 38.7 39.02 573.0 3.4082
1236 Chi-Chi CHY088 37.48 37.48 318.5 1.9942
1495 Chi-Chi TCU055 6.34 6.34 359.1 1.2668
1521 Chi-Chi TCU089 0 9 671.5 1.3418
1546 Chi-Chi TCU122 9.34 9.34 475.5 1.2615
1549 Chi-Chi TCU129 1.83 1.83 511.2 0.8290
Parameter Chi-Chi Kocaeli Imperial Valley Landers Loma Prieta Best Fit 1 Best Fit 2
PGA (g) 0.379 0.393 0.418 0.397 0.409 0.400 0.387
PGV (in./s) 18.13 21.17 17.36 17.37 17.70 18.72 19.05
PGD (in.) 18.93 16.88 9.39 9.91 9.33 15.67 13.72
PGV/PGA (s) 0.124 0.142 0.110 0.114 0.115 0.122 0.135
PGD/PGV (s) 1.044 0.797 0.541 0.571 0.527 0.837 0.720
Acceleration RMS (g) 0.053 0.050 0.071 0.073 0.065 0.059 0.049
Velocity RMS (in./s) 3.76 3.98 4.41 4.44 4.43 4.01 3.53
Displacement RMS (in.) 5.79 4.30 3.29 3.36 3.18 4.71 3.35
Arias intensity (m/s) 3.774 2.343 3.409 4.563 2.328 3.719 3.027
Cumulative absolute velocity (in./s) 977 676 766 1059 559 978 843
Significant duration (s) 27.04 24.47 23.00 30.00 15.76 31.28 28.93
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 05/13/16. Copyright ASCE. For personal use only; all rights reserved.
bridge on very stiff substructures isolated with lead-rubber-bearings periods beyond 0.54 s in each of the plots, the AASHTO Rd method
(LRB). and the elastic spectrum coincide. Clearly, for each of the record
The C m method was not considered further in this study as the sets adopted in this study, the SSM provides more severe and more
intent was to bypass the need to compute an amplification factor accurate results (assuming that NLRHA provides the correct an-
to be applied to elastic displacements, and instead to directly gen- swer) at these high ductility values of 6 and 12. Displacement
erate inelastic displacement spectra. The PEER GMPE was not amplification at periods much longer than 0.54 s is indicated in the
pursued further in this study, since the goal was to generate plots for both NLRHA and SSM results.
inelastic spectra directly from a target elastic acceleration spec- Fig. 8 depicts the results for the case in which m = 2, a = 0.00,
trum without detailed knowledge of the seismological parameters and no reduction in idealized effective damping is taken. Using
needed in the GMPE. the rationale developed in an earlier section (“The Substitute-
For ductility values of 6 and 12, no reduction in the idealized Structure Method for Isolated Bridges”), the limit ductility for
hysteretic damping given by Eq. (9) was incorporated. For a ductil- this condition is m LIMIT = 2.93. As indicated in Fig. 8, the SSM,
ity value of 4, the Dwairi hysteretic damping model of Eq. (8) was in fact, predicts inelastic displacements below the elastic dis-
adopted. For a ductility value of 2, the idealized model of Eq. (9) placements (since m = 2 < 2.93) when the full theoretical effective
was initially used, and a reduced value for effective damping equal damping is used. It is also evident in Fig. 8 that the NLRHA
to 20% of that indicated by the idealized value was used. results do not support the decrease in inelastic displacement
For a large range of periods, underestimated response results, below elastic displacement. Fig. 9 was obtained by reducing the
relative to NLRHA results, were generated by the AASHTO Rd effective damping to 20% of the calculated value and provides
method. Figs. 4–7 illustrate the results for ductility values of 6 and one means of estimating inelastic displacements for low ductility
12, in terms of inelastic spectral displacement versus initial period. values. Fig. 10 represents results for the case of a = 0.00, m = 4,
Recall that the AASHTO Rd method uses the EDR for periods larger and the Dwairi damping rule, which produced results more
than T*, which is equal to 0.54 s in the present study. Hence, for closely matching NLRHA for this case.
The prediction of lower inelastic displacements compared to each of the seven record sets and inelastic displacement spectra
elastic displacements for low ductility values is an issue which were generated.
requires more study. Two simplified analysis methods—SSM and the AASHTO Rd
method—were used to generate inelastic displacement spectra from
the target acceleration spectrum for comparison to NLRHA results.
Summary and Conclusions Methods from the literature were proposed by which appropriate
effective viscous damping levels may be computed for both the
Seven sets of ground-motion record pairs (10 record pairs per set) SSM ( j EFF) and for NLRHA ( j NLRHA).
were selected and scaled to the target response spectrum for a site. No near-field effects were assessed in this study. For near-field
NLRHA on bilinear single-degree-of-freedom systems of varying sites, a response modifier for increased effective damping that
postyield to initial stiffness ratios (a values) were carried out for incorporates pulse behavior, such as that from previous research by
others [Eq. (13)], could be adopted. For nonisolated structures, the 0:378 1
j EFF ¼ m j el þ 0:215 1
use of a modified rule for effective damping could be employed m 0:642
with all other aspects of the procedure identical to that presented !
here. For example, Priestley et al. (2007) include a model given by 1
Eq. (27) for thin Takeda hysteresis (with other models included as 1þ (27)
ðTEFF þ 0:824Þ6:444
well). Thin Takeda hysteresis has been proposed as applicable to
bridge columns experiencing plastic hinging. The cited reference
would prove most helpful in developing inelastic displacement The substitute-structure method for estimating inelastic dis-
spectra for hysteretic rules other than the bilinear rule used in this placement in bilinear oscillators provided more accurate results
study. compared to the AASHTO Rd method for the record sets used in
Fig. 10. Inelastic displacement spectra; a = 0.00, m = 4, Dwairi rule for j EFF
this study. For this reason, the substitute structure method is pro- magnitude earthquake, (2) a Class D subsurface profile, (3)
posed as a potential alternative for estimating inelastic displace- accelerations characteristic of the selected site, and (4) reduced
ments in bridge structures, isolated or otherwise. Specifically, the initial damping to mimic a tangent-stiffness-based damping
following conclusions are made: solution.
• For ductility values of 6 and 12, and for postyield stiffness val- • AASHTO Rd method analytical results underestimated
ues of 0.00 and 0.05, SSM inelastic displacement results com- NLRHA inelastic displacements over a significant period
puted by the guidelines presented in this study closely range for the ground-motion sets used in this study. The Rd
matched NLRHA results over a wide range of periods. method is currently used in AASHTO for ductility levels of
• SSM-based inelastic displacement spectra generally followed m = 6 and less.
the NLRHA-based spectra more closely than did the Rd-based • The SSM does, in fact, confirm the EDR, though over a limited
spectra for the specific conditions of (1) a large modal period range. The range over which the EDR applies,
for a site;
SF ¼ scale factor by which a ground motion record buildings and other structures (FEMA P-750), FEMA,
is multiplied; Washington, DC.
Field, E. H., Jordan, T. H., and Cornell, C. A. (2003). OpenSHA: A devel-
T ¼ initial period of a bilinear oscillator;
oping community - modeling environment for seismic hazard analysis.
TEFF ¼ secant-stiffness-based, effective period of a Seismol. Res. Lett., 74(4), 406–419.
bilinear oscillator; Ghosh, S. K. (2014). Significant changes from ASCE 7-05 to ASCE 7-10,
TL ¼ period at which spectral displacement reaches Part 1: Seismic design provisions. PCI J., 59(1), 60–82.
a maximum; Gulkan, P., and Sozen, M. A. (1974). Inelastic response of reinforced con-
VS30 ¼ average shear wave velocity in the upper 30 m crete structures to earthquake ground motions. J. Am. Concr. Inst.,
of a soil profile; 71(12), 604–610.
ZTOR ¼ depth to top of rupture; Katsanos, E. I., Sextos, A. G., and Manolis, G. D. (2010). “Selection of
Z2.5 ¼ depth to VS30 = 2.5 km/s; earthquake ground motion records: A state-of-the-art review from a
a ¼ ratio of postyield to initial stiffness values for a structural engineering perspective.” Soil Dyn. Earthquake Eng., 30(4),
bilinear system; 157–169.
l 1 ¼ ratio of effective period to initial period; Khose, V. N., and Singh, Y. (2014). “An anomaly in the equivalent lineari-
zation approach for the estimation of inelastic response.” Earthquake
l 2 ¼ factor used in determining the appropriate value
Spectra, 30(2), 965–972.
for initial viscous damping; NEHRP Consultants Joint Venture. (2011). “Selecting and scaling earth-
m ¼ displacement ductility = DMAX/Dy; quake ground motions.” Rep. NIST GCR 11-917-15, NIST, Redwood
m LIMIT ¼ displacement ductility at which SSM predicts City, CA.
inelastic displacements exactly equal to elastic Pacific Earthquake Engineering Research Center. (2011). “PEER
displacements in the constant velocity region of ground motion database.”hhttp://peer.berkeley.edu/peer_ground
the target spectrum; _motion_databasei.
j el ¼ initial, elastic damping; historically taken as 0.05; Pacific Earthquake Engineering Research Center. (2014). “PEER ground
j o ¼ initial, elastic component of viscous damping motion database.”hhttp://ngawest2.berkeley.edu/i.
used in the SSM to mimic a tangent-stiffness- Priestley, M. J., and Grant, D. N. (2005). “Viscous damping in seismic
based solution; design and analysis.” J. Earthquake Eng., 9(2), 229–255.
j hys ¼ added viscous damping from hysteretic behavior; Priestley, M. J., Calvi, G. M., and Kowalsky, M. J. (2007). Displacement-
based seismic design of structures, 1st Ed., IUSS Press, Pavia, Italy.
j EFF ¼ total, effective, equivalent viscous damping
SeismoSpect 1.3.0 [Computer software]. SeismoSoft, Chalkida, Greece.
used in the SSM; and Stafford, P. J., Mendis, R., and Bommer, J. J. (2008). “Dependence of
j NLRHA ¼ initial, elastic component of viscous damping damping correction factors for response spectra on duration and number
used in an initial-stiffness-based NLRHA algo- of cycles.” J. Struct. Eng., 10.1061/(ASCE)0733-9445(2008)134:
rithm to mimic a tangent-stiffness-based solution. 8(1364), 1364–1373.
Stewart, J. P., Abrahamson, N. A., Atkinson, G. M., Baker, J. W., Boore,
References D. M., Bozorgnia, Y., et al. (2011). “Representation of bidirectional
ground motions for design spectra in building codes.” Earthquake
AASHTO. (2010). Guide specifications for seismic isolation design, 3rd Spectra, 27(3), 927–937.
Ed., Washington, DC. USGS. (2015). “2008 interactive deaggregations.”hhttps://geohazards.usgs
AASHTO. (2011). Guide specifications for LRFD seismic bridge design, .gov/deaggint/2008/i.
2nd Ed., Washington, DC. Watanabe, G., and Kawashima, K. (2014). “An evaluation of the displace-
Baker, J., and Cornell, C. A. (2006). Which spectral acceleration are you ment amplification factors for seismic design of bridges.” hhttp://www
using? Earthquake Spectra, 22(2), 293–312. .cuee.titech.ac.jp/21coe/English/Events/Data/Papers/89watanabegakuho
Bispec 1.1.2 [Computer software]. Earthquake Solutions, Monrovia, CA. .pdfi (Nov. 23, 2015).
Blandon, C. (2004). Equivalent viscous damping equations for direct dis- Watson-Lamprey, J., and Boore, D. M. (2007). Beyond SA-GMRotI:
placement based design. European School of Advanced Studies in Conversion to SA-Arb, SA-SN, and SA-MaxRot. Bull. Seismol. Soc.
Reduction of Seismic Risk−ROSE School, Pavia, Italy. Am., 97(5), 1511–1524.