THE OVERPREDICTION OF LIQUEFACTION HAZARD IN CERTAIN AREAS OF LOW TO MODERATE SEISMICITY

Transcription

1 THE OVERPREDICTION OF LIQUEFACTION HAZARD IN CERTAIN AREAS OF LOW TO MODERATE SEISMICITY Dr. Kevin Franke, Ph.D., P.E., M.ASCE Dept. of Civil and Environmental Engineering Brigham Young University April 20, 2017

2 Liquefaction Hazard Liquefaction can result in significant damage to infrastructure during earthquakes Liquefaction hazard is generally correlated with seismic hazard Some areas of low to moderate seismicity have significant liquefaction hazard, however Image: Karl V. Steinbrugge Collection, EERC, Univ of California, Berkeley

3 Overview of Simplified Empirical Method Liquefaction is usually evaluated with a factor of safety, FS L Resistance Cyclic Resistance Ratio FSL Loading Cyclic Stress Ratio CRR CSR CSR CSR site Liq No Liq Function of both a max and M w, which collectively characterize seismic loading (after Mayfield et al. 2010)

4 N req : An Alternative to CSR and CRR CSR CSR CSR site (after Mayfield et al. (2010) ) N site = (N 1 ) 60,cs,i = the actual N value for the soil layer of interest N req = the SPT resistance required to resist liquefaction at a given CSR

5 Various Approaches for Liquefaction Hazard Analysis Deterministic Approach Considers an individual seismic source and corresponding ground motions individually Usually assumes mean values for the inputs and models Pseudo-Probabilistic Approach Considers probabilistic ground motion from a single return period Usually assumes mean values for the inputs and models Probabilistic (or Performance-Based) Approach Considers probabilistic ground motions from ALL return periods Accounts for parametric and model uncertainties Results depend on desired hazard level or return period

6 Pseudo-Probabilistic Approach: How do we get a max and M w? Deaggregation Analysis Downtown San Diego

7 Conventional (i.e., pseudo-probabilistic ) Liquefaction Triggering Procedure 1. Perform PSHA with PGA and a deaggregation analysis at the specified return period of PGA (e.g., 2475-year for the MCE) 2. Obtain either the mean or modal M w from the deaggregation analysis 3. Correct the PGA value for site response using site amplification factors or a site response analysis to compute a max 4. Couple a max with the mean or modal M w to perform a scenario liquefaction triggering analysis 5. Typically define liquefaction triggering as P L 15% and FS L 1.2

8 Pseudo-Probabilistic Example.. Consider the following site in Cincinnati, Ohio:

9 Pseudo-Probabilistic Example.. Here is the corresponding 2,475-yr deaggregation from the USGS: PGA = g

10 Pseudo-Probabilistic Example.. Consider the liquefaction triggering and settlement results for a site in Cincinnati, Ohio: Conventional Approach, MCE G with Modal Magnitude Does this make sense? How likely is it that an M7.5 EQ over 450 km away produces PGA = 0.067g? <1% according to Toro et al. (1997) <2% according to Atkinson and Boore (2006)

11 Challenges with the Pseudoprobabilistic Approach It can be easy to make an incompatible (a max, M w ) pair, especially if using the modal M w PGA and M w typically are taken from a single return period, but other return periods are ignored Does not rigorously account for uncertainty in the liquefaction triggering model or the site response Contributes to inaccurate interpretations of liquefaction hazard (e.g., I used the 2,475-year PGA in my analysis, so my liquefaction results correspond to the 2,475-year return period. )

12 A Probabilistic Liquefaction Hazard Analysis (PLHA) Approach Kramer and Mayfield (2007) introduced a PLHA approach Uses probabilistic ground motions in a probabilistic manner Accounts for uncertainty in seismic loading AND the liquefaction triggering model Produces liquefaction hazard curves for each sublayer in the soil profile N N M amax FS P FSL FSL amax, m j a, m j1 i1 L i maxi j M amax P N N a, m N i req N N j1 i1 req req max j a, m (after Kramer and Mayfield 2007) maxi j

13 Back to the Cincinnati Example.. Let s use the Kramer and Mayfield (2007) PLHA approach with the Boulanger and Idriss (2012) triggering model: Conventional Approach, MCE G with Modal Magnitude PLHA Approach, Tr=2,475 years Only difference: how we considered our seismic loading and uncertainties!

14 What About Other Cities? 10 cities selected throughout the Central and Eastern U.S (after Franke et al., 2017 [under review] )

15 What About Other Cities? 6 representative soil profiles with wide range in SPT values.. (after Franke et al., 2017 [under review] )

16 What About Other Cities? Results if assuming a Site Class D.. (after Franke et al., 2017 [under review] )

17 What About Other Cities? Results if assuming a Site Class E.. (after Franke et al., 2017 [under review] )

18 Existing Tools for PLHA Approach in Practice WSLiq ( o Developed by the U. of Washington in 2008 using VB.Net o Accounts for multiple liquefaction hazards (triggering, lateral spread, settlement, and residual strength) o Developed only for use in Washington State with 2002 USGS ground motion data, but you can trick the program for other locations o Limited control over the analysis uncertainty options and models PBLiquefY v2.0 o o o o o o V1.0 developed by BYU in 2013 using Microsoft Excel and VBA Liquefaction triggering, settlement, and Newmark slope displacement Compatible with USGS 1996, 2002, 2008, or 2014 ground motions. Can be used for any site in the U.S. Multiple model options Offers lots of control over the analysis uncertainties, including site amplification factors Neither of these tools has been used widely in design!

19 Simplified Probabilistic Liquefaction Triggering Procedure Many of us understand how the USGS NSHMP uses PSHA to develop the National Seismic Hazard Maps Gridded PSHA for Generic Rock Map Ground Motions at Targeted Return Periods Correct for Site Effects Through Amplification Factors or Site Response Analysis

20 Simplified Probabilistic Liquefaction Triggering Procedure Mayfield et al. (2010) presented a similar idea for liquefaction triggering. Gridded PB Analysis for Generic Soil Layer Map Liq Hazard at Targeted Return Periods Liquefaction Parameter Map DIFFERENT FROM a Liquefaction Hazard Map Correct for Site-Specific Soil Conditions and Stresses Depth Reduction Soil Stresses Site Amplification

21 Simplified Probabilistic Procedures for Other Liquefaction Effects In 2014, a major multi-state, multi-agency research effort was initiated to develop map-based uniform hazard analysis procedures for various liquefaction effects (settlement, lateral spread, and Newmark slope displacement).

22 Boulanger and Idriss (2012, 2014) Simplified PB Liquefaction Model Research was performed at BYU to develop a simplified procedure for the Boulanger and Idriss (2012, 2014) probabilistic triggering model. Similar to the approach introduced by Mayfield et al. (2010), but we incorporated a few changes: The quadratic equation format of the Boulanger and Idriss model requires a different and more complex approach Many engineers are still uncomfortable with the N req concept Incorporation of the (N 1 ) 60,cs -dependent MSF

23 Boulanger and Idriss (2012, 2014) Simplified PB Liquefaction Model If given a liquefaction triggering model for which CRR is defined as a function of SPT resistance N, we can see that N req is just a proxy for the seismic loading (i.e., CSR): CSR CRR N req From Boulanger and Idriss (2012, 2014): CRR N N N N exp , cs 1 60, cs 1 60, cs 1 60, cs 1 P L CRR P L N N N N , 1 60, 1 cs cs 60, cs 1 60, cs 50% CRR exp

24 Boulanger and Idriss (2012, 2014) Simplified PB Liquefaction Model By combining equations, we obtain: CSR P L Nreq Nreq Nreq Nreq 50% CSR exp So instead of developing liquefaction parameter maps for a reference N req, we can develop reference maps for the median CSR to characterize seismic loading. Engineers seem much more comfortable characterizing seismic loading with CSR than they do with N req. We have called these new maps Liquefaction Loading Parameter Maps.

25 Boulanger and Idriss (2012, 2014) Simplified PB Liquefaction Model BYU has recently developed the following simplified procedure for the Boulanger and Idriss (2014) model (Ulmer and Franke 2016): Step 1: Obtain the reference CSR(%) from the appropriate liquefaction loading map CSR ref CSR ref % 100

26 Boulanger and Idriss (2012, 2014) Simplified PB Liquefaction Model BYU has recently developed the following simplified procedure for the Boulanger and Idriss (2014) model (Ulmer and Franke 2016): Step 2: For every soil sublayer in your profile, compute the appropriate CSR correction factors, CSR Site Amplification: Depth Reduction: CSR F pga ln Fpga Mean magnitude from PGA deaggregation at target return period z z CSRr sin M sin d w (z in meters) Soil Stress:

27 Boulanger and Idriss (2012, 2014) Simplified PB Liquefaction Model BYU has recently developed the following simplified procedure for the Boulanger and Idriss (2014) model (Ulmer and Franke 2016): Step 2: For every soil sublayer in your profile, compute the appropriate CSR correction factors, CSR Duration: site M w 1 MSFmax exp site MSF 4 CSRMSF ln ln ref MSF ref M w 1 MSFmax exp N1 site 60, cs MSFmax MSF ref max ref ref ref ref ln CSR ln CSR ln CSR ln CSR **NOTE: if you prefer MSF from Boulanger and Idriss (2012), then CSR MSF = 0 ***

28 Boulanger and Idriss (2012, 2014) Simplified PB Liquefaction Model BYU has recently developed the following simplified procedure for the Boulanger and Idriss (2014) model (Ulmer and Franke 2016): Step 2: For every soil sublayer in your profile, compute the appropriate CSR correction factors, CSR site ' Overburden: site v 1 C ln site K P a CSRK ln ln ref ref K ref ' v 1 C ln P a site 1 C N 1 60, cs site 1 C ref ref ref ref ln CSR 4.918ln CSR 1.762ln CSR 5.473ln CSR 33.65

29 Boulanger and Idriss (2012, 2014) Simplified PB Liquefaction Model BYU has recently developed the following simplified procedure for the Boulanger and Idriss (2014) model (Ulmer and Franke 2016): Step 3: For every soil sublayer in your profile, compute the site-specific CSR corresponding to the targeted return period Total Correction for layer i: CSR CSR CSR CSR CSR CSR pga d i, i F, i r, i MSF, i K, i Site Specific CSR for layer i: CSRi CSR CSR i ref exp ln

30 Boulanger and Idriss (2012, 2014) Simplified PB Liquefaction Model BYU has recently developed the following simplified procedure for the Boulanger and Idriss (2014) model (Ulmer and Franke 2016): Step 4: For each soil sublayer in your profile, characterize liquefaction triggering hazard using whichever metric you prefer Factor of Safety: FS L i CRR CSR i i N 1 N 60, 1 N cs 60, cs 1 N 60, cs 1 i i i 60, cs i exp CSR i Probability of Liquefaction: P L i ln CRR i CSR ln FSL i i *Note that these equations account for both parametric uncertainty (e.g., (N 1 ) 60,cs ) and model uncertainty, and are only to be used with the Boulanger and Idriss (2014) procedure.

31 Boulanger and Idriss (2012, 2014) Simplified PB Liquefaction Model Does the simplified procedure actually work? Here are some comparisons from 10 different cities, 5 different soil profiles, and 3 different return periods (Ulmer and Franke 2016):

32 Example Demonstration San Diego, CA CSR ref years CSR ref - 2,475 years ref CSR CSR ref = (from Franke et al. 2016)

33 Example Demonstration San Diego, CA Depth, z (m) Soil Type Thickness (m) (blows/0.3 meter) Fines (%) Unit Weight (kn/m 3 ) 0.1 Hydraulic Fill Hydraulic Fill Poorly Graded Sand with Silt Poorly Graded Sand with Silt Silty Sand Poorly Graded Sand with Silt Groundwater at z = 2.0 meters Free-face case, W = 10% (blows/0.3 meter) 6.1 Poorly Graded Sand with Silt Silty Sand Silty Sand Silty Sand Silty Sand Silty Sand Silty Sand Computed using Idriss and Boulanger [2008, 2010] (from Franke et al. 2016) 1 60 D50 15 = 0.5mm N 1 60,cs N

34 Example Demonstration San Diego, CA Liquefaction Triggering Results (Ulmer and Franke 2016) (N 1 ) 60,cs (blows/0.3 meter) CSR CSR Fpga CSR rd 1 60 N CSRK CSR (Eqn A1) 475 (2,475) CRR 1 N 1 60,cs FS Liq Depth, z (m) USCS Eqn A2) Eqn A3) (Eqn A4) Eqn A5) (Eqn A7) 475 (2,475) 1.5 SP-SM (0.07) 0.08 (0.07) (0.203) >0.6 >2 (>2) 2.1 SP-SM (0.07) 0.07 (0.06) (0.233) >0.6 >2 (>2) 3 SM (0.07) 0.05 (0.05) (0.278) >0.6 >2 (>2) 4.6 SP-SM (0.07) 0.03 (0.03) (0.368) (0.44) 6.1 SP-SM (0.07) 0.00 (0.00) (0.390) (0.43) 7.6 SM (0.07) (-0.03) (0.386) >0.6 >2 (>2) 9.1 SM (0.07) (-0.07) (0.402) >0.6 >2 (>2) 10.7 SM (0.07) (-0.10) (0.413) >0.6 >2 (>2) 12.2 SM (0.07) (-0.14) (0.420) >0.6 >2 (>2) 13.7 SM (0.07) (-0.18) (0.423) >0.6 >2 (>2) 15.2 SM (0.07) (-0.22) (0.425) >0.6 >2 (>2) 1 Computed with Boulanger and Idriss [58], PL=50% 2 Computed with Kramer and Mayfield [5] procedure using the Boulanger and Idriss [58] model (from 1 Franke et al. 2016)

35 Conclusions The conventional pseudo-probabilistic approach can overpredict liquefaction hazard in areas of low to moderate seismicity Especially where the selected M w 7.5 and is located more than 200 km away from the site Current seismic design provisions (e.g., IBC, ASCE, AASHTO) serve to propagate the overprediction of liquefaction Probabilistic approaches can help solve the problem, but are not easy to apply without special tools New simplified approximation methods can give you the benefits of the probabilistic approach with the convenience of the conventional approach Reference parameter maps and online tools to use them are currently being developed for Utah, Idaho, Oregon, Montana, Alaska, South Carolina, and Connecticut

36 References Boulanger, R.W. and Idriss, I.M. (2012). Probabilistic standard penetration test-based liquefaction-triggering procedure. J. Geotech. Geoenviron. Eng. 138(10), Franke, K.W. and Kramer, S.L. Procedure for the empirical evaluation of lateral spread displacement hazard curves. J. Geotech. Geoenvir. Eng., 140(1), Franke, K.W., Ulmer, K.J., Ekstrom, L.T., and Meneses, J. (2016). Clarifying the differences between traditional liquefaction hazard maps and performance-based liquefaction reference parameter maps. Soil Dynamics and Earthquake Engineering, Elsevier, 90(2016), Franke, K.W., Lingwall, B.N., and Youd, T.L. (2017, under review). Overestimation of liquefaction hazard in areas of low to moderate seismicity due to improper characterization of probabilistic seismic loading. Submitted to Soil Dynamics and Earthquake Engineering, Elsevier. Idriss, I.M. and Boulanger, R.W. (2008). Soil Liquefaction During Earthquakes, EERI Monograph MNO-12. Oakland, CA. Idriss, I.M. and Boulanger, R.W. Semi-Empirical Procedures for Evaluating Liquefaction Potential During Earthquakes. Soil Dynamics and Earthquake Engineering, Elsevier, Vol. 26, 2006, pp Ishihara, K., & Yoshimine, M. (1992). Evaluation of settlements in sand deposits following liquefaction during earthquakes. Soils Found., 32, Kramer, S.L., and Mayfield, R.T. (2007). Return period of soil liquefaction. J. Geotech. Geoenviron. Eng. 133(7), Mayfield, R.T., Kramer, S.L., and Huang, Y.-M. (2010). Simplified approximation procedure for performance-based evaluation of liquefaction potential. J. Geotech. Geoenviron. Eng. 136(1), Ulmer, K.J. and Franke, K.W. (2015). Modified performance-based liquefaction triggering procedure using liquefaction loading parameter maps. J. Geotech. Geoenvir. Eng , DOI: /(ASCE)GT Youd, T.L., Hansen, C.M., and Bartlett, S.F. (2002). Revised multilinear regression equations for prediction of lateral spread displacements. J. Geotech. Geoenvir. Eng., 128(12),

37 THE OVERPREDICTION OF LIQUEFACTION HAZARD IN CERTAIN AREAS OF LOW TO MODERATE SEISMICITY Dr. Kevin Franke, Ph.D., P.E., M.ASCE Dept. of Civil and Environmental Engineering Brigham Young University April 20, 2017