PROBABILISTIC EVALUATION OF LIQUEFACTION POTENTIAL USING MULTIVARIATE ADAPTIVE REGRESSION SPLINES.pptx

PROBABILISTIC EVALUATION OF LIQUEFACTION POTENTIAL USING MULTIVARIATE ADAPTIVE REGRESSION SPLINES

1. PROBABILISTIC EVALUATION OF LIQUEFACTION POTENTIAL USING MULTIVARIATE ADAPTIVE REGRESSION SPLINES Presented By Ranjan Kumar (Reg. No.: 2020PGCEGE05) Under the guidance of Dr. Subhadeep Metya

2. CONTENT Introduction Literature review Objective Methodology Validation Result and Discussion Conclusion Reference

3. INTRODUCTION What is Liquefaction? Why and how soil behave like liquid? Failure resulting from soil liquefaction Flow liquefaction Cyclic mobility It is been initiated by Cyclic Loading Shock Loading. Fig 1. Some effects of soil liquefaction after the  1964 Niigata earthquake

4. Earthquake intensity and duration Soil type Soil relative density Particle size distribution Plastic fine Water table location Hydraulic conductivity Aging and cementation Overburden pressure structural load Historical liquefaction FACTOR INFLUENCING LIQUEFACTION SUSCEPTIBILITY Fig 2. Limits in the gradation curve separating liquefiable soil and non liquefiable soil ( Tsuchida , 1970) INTRODUCTION…contd …

5. SIMPLIFIED PROCEDURE FOR LIQUEFACTION Most basic procedure used in engineering practice for assessment of liquefaction potential is the simplified procedure originally by Seed and Idriss (1971) subsequently updated Comparison of cyclic resistance ratio with the earthquake induced cyclic stress ratio at given depth for a specified design earthquake. CRR: Cyclic Resistance ratio. CSR: Cyclic stress ratio. Factor of Safety at a given depth: FS=   Fig 3 . Shear stress vs. depth for determination of liquefaction zone (Seed and Idriss 1971) INTRODUCTION…contd …

6. Estimation of Cyclic stress ratio : CSR = Cyclic Stress Ratio = Peak ground acceleration σ vo = Total overburden stress = Effective overburden stress r d = Stress reduction factor   Fig 4 . Determining max shear stress and the stress reduction coefficient (Seed and Idriss 1971) INTRODUCTION…contd …

7. Estimation of Cyclic Resistance ratio : CRR = CRR 7.5 (MSF) K K =correction for static shear stress K =overburden correction factor CRR 7.5 values are estimated from SPT with the help of blow count The influence of the fine content is taken care of by an equation: (N1)60c = α + β (N1)60 Where α and β depends upon fine content   Reference: IS CODE 1893 (part 1)2006 INTRODUCTION…contd …

8. AUTHOR OBJECTIVE METHODOLOGY OUTCOME Bagheripour et al et al. (2012) To calculate the probability of liquefaction potential using Genetic Algorithm (GA) Model GA and AFOSM (advance first order second moment method) to carry reliability-based analysis and optimization. Higher is the factor of safety will not always lead to a high safer state Goh and Jhang (2013) Estimation of the magnitude of liquefaction using Youd et al.(2002) Multivariate Adaptive Regression Splices (MARS) Model It is an improvement to the current(Multiple Regression Analysis) MLR model to predict the liquefaction-induced lateral displacement. Idriss and Boulanger (2014) Updated the semi-empirical formula has been developed to evaluate the liquefaction potential cohesionless soil. liquefaction correlations have been presented that are based upon a re-examination of the field data Updated semi empirical method devloped for computation of FOS LITERATURE REVIEW

9. AUTHOR OBJECTIVE METHODOLOGY OUTCOME Johari et al. (2018) To develop a probabilistic model to predict the liquefaction potential of soil using a shear wave velocity test “Truncated normal probability distribution function” while earthquake acceleration ratio and earthquake magnitude moment have been developed based on “exponential probability distribution function” V s based model gives the more conservative result than SPT based model if both are developed using (Jointly distributed random variable) JDRV model Umar et al. ( 2018) To estimate the liquefaction potential of the soil of the high seismic zone of Bihar based on SPT test data Comparison between the moduli probabilistic method and (relevance vector machine)RVM method Soil which falls under seismic zone V is more likely to get liquefied irrespective of water table and type of soil Das et al. (2018) SPT test data of borehole for computing FOS of soil liquefaction using Idriss and Boulanger (2010, 2015) A liquefaction map has been generated using the interpolation technique using(open source geographic system) QGIS software Liquefaction map has been created. LITERATURE REVIEW….contd …

10. AUTHOR OBJECTIVE METHODOLOGY OUTCOME Jha and Suzuki (2018) The value of CSR and CRR is calculated using (point estimation method first order second moment) PEM-AFSM method to compute reliability Model GA and AFSM Method to carry reliability-based analysis and optimization . The result from the combined method is similar to the result from MCS so it is recommended for routine analysis Zhang et al. (2021a) GWO (grey wolf optimization) algorithm to improve the prediction and accuracy of the (support vector ) SVM model Deep neural network and shear wave velocity has been used to predict model. S oil liquefaction based on the SPT – V s and prediction parameters have been determined and analyzed according to the mathematic prediction model. Zhang et al. (2021b) This enhance the prediction, accuracy and also fasten the operating rate GWO algorithm has been used to improve the prediction and accuracy of the (support SVM model The p erformance of this GWO-SVM model is improved continuously optimizing the setting and expanding the dataset LITERATURE REVIEW…contd …

11. AUTHOR OBJECTIVE METHODOLOGY OUTCOME Zhang et al. (2021c) To calculate the probability of failure to assess the liquefaction potential using GA ELM (extreme learning machine) The prediction model based on CPT perform better than that of SPT and can be accurately predicted up to 100% for the liquefied case and overall accuracy of 87.5% Ghani and Kumari (2021) To develop a regression model using multi-linear regression analysis to predict liquefaction using important parameters such as liquid limit (LL), SPT blow count, fine content (FC), and moisture content First-order second-moment ( FOSM ) reliability analysis has been used Reliability indices and probability of liquefaction has been obtained to define the liquefaction potential in a more effective manner PL-graphs Zhao et al. (2021) Updated the semi-empirical formula has been developed to evaluate the liquefaction potential cohesionless soil. PSO KELM( practical swarm optimization kernel extreme learning machine) based methodology used for the evaluation of the potential of soil liquefaction Its performance is significantly better than the traditional algorithm. LITERATURE REVIEW contd…

12. OBJECTIVE To develop different computational codes in MATLAB environment using the methodologies proposed by Seed and Idriss (Seed and Idriss 1970), Youd et al. (Youd et al. 2001), Idriss and Boulanger (Idriss and Boulanger 2006) and IS 1893 (Part 1) (IS 1893 (Part 1) 2016) To validate the result of Youd et al. (Youd et al. 2001) and Idriss and Boulanger (Idriss and Boulanger 2004) method by site data reported in Jha and Suzuki (Jha and Suzuki 2009) A comparison between the above-mentioned methodologies will be made by reanalysing the Chi-Chi Earthquake Liquefaction Data well documented by Hwang and Yang (Hwang and Yang 2001) .  

13. OBJECTIVE contd…. To develop a comprehensive surrogate model considering all the above-mentioned six parameters (the N value, LL, PI, FC, PGA and CSR) using the multivariate adaptive regression splines (MARS), and to train and test the developed model, the Ghani and Kumari (2021) will be used. Finally, to perform probabilistic analysis using the First order reliability method (FORM) for predicting the liquefaction response of soils

14. METHODOLOGY

15. Validation of the developed computational codes Bore hole diameter correction factor C b = 1.05 Energy ratio correction Ce = 1.0 Correction for split spoons without liner Cs = 1 Correction for static shear stress, Kα = 1 From 4-14 m depth, a liquefiable sandy layer exists. The average depth of groundwater table is 3.25 m deep. The site analysis was done for PGA = 0.2 Magnitude of earthquake as M = 7 VALIDATION

16. Sl. No. Depth(m) γ (KN/m 3 ) N FC (%) FOS (Jha and Suzuki, 2009) FOS obtained in present study using Youd et al. (2001) method Idriss and Boulanger (2006) method 1 1.5 20.9 6 87 - 1.71 1.4 2 3 21.5 10 12 - 1.47 1.37 3 4.5 20.9 16 4 1.56 1.46 1.47 4 6 19 14 3 1.17 1.17 1.14 5 7.5 19.5 14 8 1.05 1.09 1.05 6 9 21.7 14 6 0.89 0.93 0.94 7 10.5 20 15 3 0.87 0.93 0.93 8 12 21.4 12 2 0.68 0.73 0.76 9 13.5 20 15 17 1.01 1.26 1.06 10 15 16.6 13 97 - 1.20 1.03 11 16.5 16.1 13 98 - 1.21 1.02 12 18 17 13 98 - 1.22 1.01 Table 1 . Validation of developed computing codes for the Youd et al. (2001) method and Idriss and Boulanger (2004) method VALIDATION…… Cont …...

17. RESULTS AND DISCUSSIONS COMPARATIVE STUDY OF DIFFERENT DETERMINSTIC METHODS Sl. No. Methods Maximum error (%) Minimum error (%) Mean Absolute Error (MAE) Coefficient of determination (R 2 ) 1 Seed and Idriss Method 2.79 0.025 0.005 0.996 2 Youd et al. Method 1.45 0.13 0.02 0.979 3 Idriss and Boulanger Method 2.3 0.33 0.003 0.998 4 IS 1893 (Part 1) 2016 Method 8.51 0.85 0.004 0.996 Table 2 . Summary of results of different statistical parameters

18.   Sl. No. Depth (m)  ( kN /m 3 ) SPT N FC Water table (m) Values of F L using different methods Youd et al. Method Idriss and Boulanger Method IS 1983 (Part 1) 2016 Method 1 5 19.8 9 20 4 0.487 0.494 0.297 2 5.8 19.0 7 25 5 0.436 0.474 0.298 3 8.3 19.6 12 13 2.8 0.398 0.386 0.195 4 6.3 22.0 16 15 1.2 0.517 0.368 0.241 5 2.8 18.5 6 22 0.7 0.381 0.612 0.256 6 7.3 20.5 11 21 5.0 0.459 0.406 0.274 7 3.0 20.0 5 24 2.4 0.443 0.715 0.316 8 7.5 18.5 12 55 2.8 0.453 0.332 0.287 9 5.8 19 4 35 2.8 0.317 0.707 0.245 10 5.8 18.3 10 30 1.5 0.408 0.614 0.265 Table 3 . Summary of F L Based on Different Methods RESULTS AND DISCUSSIONS….contd ….

19. Table 4. Result of liquefaction analysis for site data reported by Jha and Suzuki (2009) using Idriss and Boulanger (2006) method. Sl. No. PGA r d (N 1 ) 60 MSF CSR CRR F L β P L 1 0.2 0.99 15.254 1.224 0.129 0.158 1.4 1.394 0.082 2 0.2 0.97 14.511 1.224 0.127 0.152 1.369 1.348 0.089 3 0.2 0.95 18.014 1.224 0.142 0.184 1.473 1.47 0.071 4 0.2 0.93 14.895 1.224 0.155 0.155 1.144 0.689 0.245 5 0.2 0.94 14.541 1.224 0.162 0.152 1.05 0.27 0.394 6 0.2 0.88 13.426 1.224 0.164 0.143 0.944 0.039 0.484 7 0.2 0.85 13.771 1.224 0.166 0.146 0.933 0.009 0.496 8 0.2 0.83 10.565 1.224 0.164 0.112 0.766 0.725 0.234 9 0.2 0.79 16.65 1.224 0.163 0.171 1.061 0.434 0.332 10 0.2 0.77 16.341 1.224 0.162 0.168 1.035 0.328 0.371 11 0.2 0.75 16.144 1.224 0.161 0.166 1.021 0.255 0.399 12 0.2 0.72 15.937 1.224 0.159 0.164 1.01 0.243 0.404 Reliability analysis of soil liquefaction using FORM RESULTS AND DISCUSSIONS….contd ….

20. Fig 5 . Relationship between sample size and potential of soil liquefaction Sl. No. Parameters Mean COV Probability Distribution 1 (N 1 ) 60 f(z) 0.25 lognormal 2 MSF 1.22 0.05 lognormal 3 PGA 0.2 0.2 lognormal 4 r d f(z) 0.1 lognormal Table 5 . Statistical properties of the random parameters RESULT AND DISCUSSIONS….contd …. Probability Density Function of CSR and CRR using MCS

21. Fig 6 . Probability density function of CSR   Fig 7 . Probability density function of CRR Probability Density Function of CSR and CRR RESULT AND DISCUSSIONS….contd ….

22. Fig 8 . Illustration of observed and predicted values of factor of safety DEVELOPMENT OF MARS MODEL Importance of adding plasticity properties of soil Parameters Considered [(N 1 ) 60 , FC, PGA, CSR, LL and PI)] Initially 100 datasets have used for the training the MARS model ( Ghani and Kumari 2021) 13 datasets have then been employed for testing/ verifying the developed MARS model RESULTS AND DISCUSSIONS….contd ….

23. Sl. No. Item RMSE R 2 1 Training 0.088 0.968 2 Testing 0.133 0.893 3 Liquefied Cases 0.117 0.749 4 Non-liquefied Cases 0.092 0.957 Table 6 . Statistical performance detail for the developed MARS model. The performance of the developed MARS model in predicting liquefied and non-liquefied soil RESULTS AND DISCUSSIONS….contd ….

24. Fig 9. Variation of FOS with plasticity index, PI Fig 10 . Variation of FOS with liquid limit, LL   Effect of Soil Plasticity and liquid limit against Liquefaction RESULTS AND DISCUSSIONS….contd ….

25. RELIABILITY ANALYSIS BASED ON THE DEVELOPED MARS MODEL Table 7 . The COV values for random parameters. Sl. No. Parameters COV Probability Distribution 1 (N 1 ) 60 0.4 lognormal 2 FC (%) 0.35 lognormal 3 PGA 0.1 lognormal 4 LL (%) 0.25 lognormal 5 PI (%) 0.2 lognormal 6 CSR 0.15 lognormal Table 8. Summary of reliability results using FORM coupled with MARS model Sl. No. FOS Against Liquefaction, F L Reliability Index, β Liquefaction Probability, P L P L (%) 1 0.54 1.95 0.03 2.56 2 0.84 0.67 0.25 25.21 3 1.13 0.09 0.46 46.26 4 1.46 1.15 0.13 12.60 5 1.39 0.54 0.29 29.40 6 1.31 0.51 0.31 30.59 7 1.59 0.95 0.17 17.17 8 0.97 0.26 0.40 39.64 9 1.17 0.34 0.37 36.56 10 1.19 0.39 0.35 34.82 11 1.76 1.64 0.05 5.07 12 1.19 0.45 0.33 32.79 13 2.04 1.75 0.04 3.99 RESULTS AND DISCUSSIONS….contd ….

26. Fig 11. Relationship between the FOS against liquefaction (F L ) and liquefaction probability (P L ) RESULTS AND DISCUSSIONS….contd ….

27. A comparative study between the sensitivity indexes of the random parameters has also been done for all the 13 sites Fig 12. Sensitivity analysis study based on the FORM method RESULTS AND DISCUSSIONS….contd ….

28. Based on various studies made in this dissertation, the following conclusions have been drawn In this present study, it has been observed that the cyclic strength ratio (CSR) obtained using the developed computing codes based on Seed and Idriss (1971) method, Youd et al. (2001) method, Idriss and Boulanger (2006) method and IS 1893 (Part 1): 2016 method are quite similar those reported by Hwang and Yang (2001) using the Chi-Chi earthquake site data. Comparing the Chi-Chi earthquake site data, it is observed that the mean absolute error (MAE) associated with Idriss and Boulanger method is the minimum and the coefficient of determination value is the maximum amongst the four methods considered in this study. Therefore, it can be concluded that Idriss and Boulanger method is the most accurate method. CONCLUSIONS

29. It has also been observed that factor of safety against liquefaction determined using the Idriss and Boulanger Method is more close to one than the other methods in those site is susceptible to liquefaction. This is quite similar to the observation made in Chi-Chi earthquake site data reported in Hwang and Yang (2001). For probabilistic liquefaction analysis, first order reliability method (FORM) has been used and the probability of soil liquefaction have been calculated (PL). From a given bore log data ( Jha and Suzuki 2009), it has been observed that the probability soil liquefaction is maximum at depth 10.5 m whereas it is minimum at 4.5 m. It is also observed that in some depth, even though the factor of safety is greater than one but still some chances of liquefaction is there. Although it is well accepted that plasticity of soil has a dominant role in predicting liquefaction potential of soil, very few studies have been reported in this direction. In this study, a MARS (Multivariate Adaptive Regression Splines) based surrogate model have been developed including plasticity properties of soils (namely, LL and PI) in addition to other significant parameters [(N1)60, FC, PGA, CSR] to predict the factor of safety against liquefaction. CONCLUSIONS….contd ….

30. From the studies on effect of plasticity on liquefaction, it has been observed in the present study that fine grained soil with LL > 25% or PI < 8% has more chances of failure. From the probabilistic evaluation of liquefaction potential based on the developed MARS model, it is observed from that the factor of safety against liquefaction is not the consistent measure risk particularly when it indicates liquefied case. From the probabilistic sensitivity analysis, it is observed that for most of the sites, LL (%) is found out to be second most significant parameters and in some site, it is the most dominated parameters amongst the six parameters considered in this study. Therefore, it may be concluded that for fine grained soil, the plasticity properties of soil have a significant contribution in liquefaction assessment. CONCLUSIONS….contd ….

31. REFERENCE Andrus, R. D., and Stokoe , K. H. (1998). “Liquefaction Resistance of Soils from Shear-Wave Velocity.” Proc., NCEER Workshop on Evaluation of Liquefaction Resistance of Soils, Tech. Rep. NCEER-97-0022, T. L. Youd and I. M. Idriss, eds., National Center for Earthquake Engineering Research, Buffalo , 89–128. Bagheripour , M. H., Shooshpasha , I., and Afzalirad , M. (2012). “A genetic algorithm approach for assessing soil liquefaction potential based on reliability method.” Journal of Earth System Science , 121(1), 45–62. Boulanger, R., and Idriss, I. (2004). “Evaluating the potential for liquefaction or cyclic failure of silts and clays.” Neuroscience Letters , 339(December), 123–126. Bray, J. . D., and Sancio , R. B. (2006). “Assessment of the Liquefaction Susceptibility of Fine-Grained Soils.” Journal of Geotechnical and Geoenvironmental Engineering , 132(9), 1165–1177. Chung , J., and Rogers, J. (2017). “Deterministic and Probabilistic Assessment of Liquefaction Hazards Using the Liquefaction Potential Index and Liquefaction Reduction Number.” Journal of Geotechnical and Geoenvironmental Engineering , 143(10), 04017073. Jha, S. K., and Suzuki, K. (2009). “Reliability analysis of soil liquefaction based on standard penetration test.” Computers and Geotechnics , 36(4), 589–596 .

32. REFERENCE….contd…. 7) Idriss , I. M., and Boulanger, R. W. (2004). “Semi-empirical procedures for evaluating liquefaction potential during earthquakes/Proceedings of 11-th International Conference on Soil Dynamics and Earthquake Engineering and the 3-d International Conference on Earthquake Geotechnical Engineering 32–56. 8) Idriss , I. M., and Boulanger, R. W. (2006). “Semi-empirical procedures for evaluating liquefaction potential during earthquakes.” Soil Dynamics and Earthquake Engineering , 26(2-4 SPEC. ISS.), 115–130. 9) Idriss , I. M., Seed, H. B., and ASCE, A. M. (1968). “Seismic Response of Horizontal Soil Layers.” Journal of the Soil Mechanics and Foundations Division , 94, 1003–1031 . 10) H. B., and Idriss, I. . (1970 ). “A simplified Procedure for Evaluating soil liquefaction potential”. 11) Seed , H. B., and Lee, K. L. (1966). “Liquefaction of saturated sands during cyclic loading.” Journal of Soil Mechanics and Foundation Division, 92(SM6), 105–134.

33. THANK YOU