1 Asia Managing Geotechnical Risk Learning from the Failures “Issues related to the use of Numerical Modelling in Design of Deep Excavations in Soft Clay” Andy Pickles of GCG (Asia) Ltd.

2 Asia Content of Presentation Describe the Method A/B Problem Describe the Method A/B Problem Comment on Cam Clay model in routine design Comment on Cam Clay model in routine design Highlight Difficulty of modelling piles in 2D Analyses Highlight Difficulty of modelling piles in 2D Analyses Comments on modelling of JGP Comments on modelling of JGP

3 Asia Simplified Soil Behaviour Most engineers are familiar with E and υ Most engineers are familiar with E and υ Preferable to adopt Shear Modulus (G) and Bulk Modulus (K) Preferable to adopt Shear Modulus (G) and Bulk Modulus (K) Shear strains due to changes in shear stress are proportional to 1/G Shear strains due to changes in shear stress are proportional to 1/G Volume strains due to changes in mean stress are proportional to 1/K Volume strains due to changes in mean stress are proportional to 1/K Water has zero G and very high K Water has zero G and very high K w For drained and undrained conditions G is the same For drained and undrained conditions G is the same For drained conditions K is K for soil For drained conditions K is K for soil For undrained conditions K becomes very high (i.e. is K) For undrained conditions K becomes very high (i.e. is K w )

4 Asia Mohr Coulomb Model and Method A/B Most analyses adopt simple Mohr Coulomb model with no dilation Most analyses adopt simple Mohr Coulomb model with no dilation For undrained condition no volume change For undrained condition no volume change Soil particles are only affected by changes in effective stress Soil particles are only affected by changes in effective stress No volume change means no change in mean effective stress (p’) in soil No volume change means no change in mean effective stress (p’) in soil Soil is constrained to constant p’ stress path Soil is constrained to constant p’ stress path Soil will fail where constant p’ crosses failure line Soil will fail where constant p’ crosses failure line Method A/B refers only to choice of strength criteria in undrained analyses using Mohr Coulomb model Method A/B refers only to choice of strength criteria in undrained analyses using Mohr Coulomb model Method A uses c φand Method B uses Cu Method A uses c φand Method B uses Cu

5 Asia Method A C, phi Method B Cu Normally Consolidated Clay Undrained Loading FE Model Constant p’ Zero dilatancy Cam Clay Soil is contractive

6 Asia Over-consolidated Clay Ko Consolidated Clay

7 Asia Method A at Nicoll Highway M3 Section ● Method A/B problem is not unique to Plaxis ● Method A was in widespread use in Singapore (and is widely adopted internationally) ● Method A was adopted for design of C824 ● Method A (and other methods) should be compared with design Cu profile ● Excavations at C824 were deepest ever in Singapore

8 Asia Nicoll Highway M3 Design Section MC Lower MC Upper EC Soft Clay 40 m

9 Asia Effect of Method A on Cu Profile Method A, Ko = 1 Method A, Ko = 0.6 Design Cu Profile

10 Asia Method A on Net Pressure Profile Excavation for 6 th Strut 5 th Strut Excavation Level Upper JGP Layer Design Cu Profile Method A Ko = 0.6 Net Pressure +ve Pa > Pp 15m Span

11 Asia Effect of Method A on Wall Displacement Method AMethod B

12 Asia Effect of Method A on Bending Moments Method AMethod B

13 Asia Effect of Method A on Strut Loads Strut RowPredicted Strut Load Using Method B Design Strut Load Using Method A Ratio Method B to Design Strut Load % % % % % % % % % Design Strut Load may be controlled by backfilling process

14 Asia Mohr Coulomb and Cam Clay Type Models ● For deep excavations Method A can under-estimate wall displacement and BM ● For shallow excavations Method A will over-estimate wall displacement and BM ● Method B matches the design undrained strength profile and is preferable ● Neither Method A or B model the real behaviour of soft clay ● Post collapse recommendation to use Cam Clay type models

15 Asia FE Model Constant p’ Cam Clay or real Soil Idealised behaviour of soil using Cam Clay type models

16 Asia Actual behaviour of Singapore Marine Clay ● Real behaviour of Marine Clay determined from high quality lab tests ● Sampling carried out using thin wall with 5 degree cutting angle ● Samples anisotropically re-consolidated to in situ stresses prior to testing ● Testing carried out undrained in extension and compression

17 Asia Real Behaviour

18 Asia Parameters for Upper Marine Clay Cu Peak68 kPa φ at Peak undrained 25º Cu Large Strain 52 kPaφLarge Strain34º % Change 25% reduction % Change35% Increase Design φ adopted in Singapore is 22º (NSF calcs?) To obtain correct design Cu profile with modified Cam Clay model, φ = 17º is required

19 Asia Mohr Coulomb v Modified Cam Clay Modified Cam Clay model includes features of soft clay behaviour Modified Cam Clay model includes features of soft clay behaviour Some natural soft clays differ from Modified Cam Clay Some natural soft clays differ from Modified Cam Clay Physically unrealistic values may be required to match undrained strength profile Physically unrealistic values may be required to match undrained strength profile For managing risk care must be taken to understand implication of differences For managing risk care must be taken to understand implication of differences Possibly simpler to adopt Mohr Coulomb with Method B Possibly simpler to adopt Mohr Coulomb with Method B

20 Asia Modelling Piles in 2 D Analyses ● Structures constructed in deep excavations in Singapore are often founded above soft clay on piles ● Piles are often constructed after installation of JGP layers but before commencement of excavation ● Piles will be bonded to the JGP ● Heave of ground during excavation results in tension in piles ● Presence of piles will restrain heave and also restrict wall movements

21 Asia Comments on modelling of Piles ● Modelling piles in 2D analyses as walls connected to the ground can severely restrict the predicted wall movement ● Wall displacements will be under-predicted and wall bending moments also under-predicted ● If 3D modelling is not available then it may be preferable to carry out sensitivity studies without piles and with piles modelled as “anchors” not connected to the soil mesh ● For managing risk you must understand the limitations implicit in simple 2D models – sensitivity analyses

22 Asia Modelling JGP Numerical models for design typically adopt Mohr Coulomb type model Numerical models for design typically adopt Mohr Coulomb type model E = 150MPa, C u = 300kPa (minimum UCS is 900kPa) E = 150MPa, C u = 300kPa (minimum UCS is 900kPa) JGP strength is a factored value used in analyses where soil strength is unfactored JGP strength is a factored value used in analyses where soil strength is unfactored How are design values justified? How are design values justified?

23 Asia Design Value 150 MPa USC Results Design 900kPa Average 2000kPa E 50 from UCS Tests Average 500 MPa

24 Asia Axial strain at failure in UCS tests on JGP Average 0.8%

25 Asia Summary of JGP Properties Model CuCuCuCuE Fail Strain % LaboratoryUCS> DesignM-C Back Analyzed Real *1 >2 Advanced Analysis Brittle? 500 *2 / *2 *1 – Non linear response *2 – Peak to residual at 20% plastic strain

26 Asia Modelling of JGP Actual mass characteristics of JGP not well understood Actual mass characteristics of JGP not well understood No direct relationship between lab and field performance No direct relationship between lab and field performance Parameters and model presently used for design are probably incorrect and may be unsafe Parameters and model presently used for design are probably incorrect and may be unsafe JGP is probably a brittle material whereas Mohr Coulomb is elastic/perfect plastic JGP is probably a brittle material whereas Mohr Coulomb is elastic/perfect plastic Sensitivity analyses with high and low strength and stiffness values are essential Sensitivity analyses with high and low strength and stiffness values are essential

27 Asia Concluding Remarks Numerical modelling has an important role in design Numerical modelling has an important role in design Numerical modelling requires specialist knowledge Numerical modelling requires specialist knowledge For managing risk make sure that the limitations of the model are well understood (investigated) For managing risk make sure that the limitations of the model are well understood (investigated) Do not rely on preciseness of results Do not rely on preciseness of results Sensitivity/ trends in behaviour more important Sensitivity/ trends in behaviour more important Always perform sanity checks by alternative means Always perform sanity checks by alternative means

28 Asia End of Presentation