Workshop on Nonlinear modelling of geotechnical problems: from theory to practice Johns Hopkins University, Maryland, 3-4 November 2005
Nonlinear modelling of soils David Muir Wood University of Bristol, UK
Summary 1. Recent work on constitutive modelling - hierarchical extensions of Mohr Coulomb, Cam clay - multiaxial testing driving/informing modelling 2. Promotion of use of advanced numerical models in practice
Hierarchical extensions of Mohr Coulomb, Cam clay build on familiar foundations advantage in using well known models as basis – check implementation – acceptability relatively straightforward to add extra features to a soil model extra features imply additional soil parameters and additional calibration tests seek adequate complexity in modelling – match complexity of model to availability of data and needs of application
Mohr-Coulomb model with strength dependent on state variable
influence of density softening dilatancy simplicity
Mohr-Coulomb model with strength dependent on state variable
define state variable ψ function of density and stress level requires location of critical state line mathematical definition not important linear semi-logarithmic? (simple)
Mohr-Coulomb model with strength dependent on state variable
current available strength depends on current value of state variable ψ ψ varies during test, stress history, etc simple linear relationship? (data collected by Been & Jefferies)
Mohr-Coulomb model with strength dependent on state variable
distortional hardening monotonic increase of ratio of mobilised to available strength (η/ηp) with distortional strain εqp hyperbolic hardening law: simple but available strength is not constant
Mohr-Coulomb model with strength dependent on state variable
flow rule links dilatancy with mobilised strength η so density changes during shearing linear relationship? (simple)
current peak strength
conventional drained triaxial compression tests different initial density (state variable) volume change accompanies shearing hence change in state variable hence change in available strength model automatically homes in on critical state softening emerges without being described mathematically peak strength is moving target reached at infinite distortional strain – then identical with critical state strength
Severn-Trent sand add kinematic hardening: elastic region of high stiffness carried round with recent stress history boundary of elastic region is the yield surface use bounding surface plasticity: plastic hardening stiffness depends on separation of the yield surface and bounding surface kinematic hardening Mohr-Coulomb: strength dependent on state variable: hierarchical development
Severn-Trent sand calibrated against triaxial test data for Hostun sand effect of different density/stress level automatically described (ignore practical problem of maintaining homogeneity within softening sample)
Gajo & Muir Wood, 1999
Severn-Trent sand cyclic undrained test Hostun sand
Stress response envelopes: Hostun sand: small-medium strain Stress response envelope for samples with stress histories ABC:
Stress response envelope for samples with stress histories AB: σz
σz 250
250
150
150
Distortional strain: (%)
Distortional strain: (%) C
B
B
0.05% 0.2%
50
0.05% 0.2%
50
0.4% -250
ABC
-150
A
-50
qx: kPa 50
150
0.4%
0.6% -250 0.8%
250
-50
1.0%
qx: kPa -150
A
-50
50
σy
-150
two corners
σx
qz: kPa
-250
150
Distortional strain: (%)
AB qz: kPa
1.2%
radial shearing
distortional strain
0.05% 0.2%
50
0.4%
A -250
A
-150
-50
qx: kPa 50
150
250
-50
0.6% 0.8% 1.0% 1.2%
σx
σy
-150
isotropic compression -250
qz: kPa
0.8%
σy
multiaxial testing
σz 250
0.6% 1.0%
-150
-250
Stress response envelope for initilally isotropically compressed samples :
250
-50
1.2% σx
150
0.05%: history recalled 1%: history ‘forgotten’
Modelling framework: kinematic elasto-plasticity Sz/p
learn from multiaxial experiments Bounding surface
defining hardening rule A
Loading Surface Target Surface Sy/p
exploring stress-dilatancy rule importance of escaping from axial symmetry Sx/p
Cam clay
elastic-hardening plastic model volumetric hardening associated flow – normality
kinematic hardening extension yield locus carried around with stress state – 'bubble' – strongly influenced by recent history stiffness falls as yield 'bubble' approaches bounding surface – controlled by distance b when loading with 'bubble' in contact with bounding surface model is identical to Cam clay
volumetric strain
η
kaolin
η
constant p' cycles hysteresis experiment
simulation
η η
distortional strain
build up of volumetric strain
natural soils often contain structure: bonding between particles: destroyed with mechanical or chemical damage… …or developed with (geological) time
ratio of sizes of structure surface and reference surface gives indication of current degree of structure
design model: yield surface has increased size as result of bonding with plastic straining (or chemical weathering) yield surface shrinks to the yield surface, for remoulded, structureless material extension of 'bubble' kinematic extension of Cam clay all features of 'bubble' model retained
Norrköping clay – calibration tests
Rouainia & Muir Wood (2000)
Norrköping clay – undrained – isotropic overconsolidation Rouainia & Muir Wood (2000)
simulation
experiment
Bothkennar clay results normalised by Hvorslev equivalent consolidation pressure p'e for structureless soil
Gajo & Muir Wood, 2001
Hierarchical extensions of Mohr Coulomb, Cam clay build on familiar foundations advantage in using well known models as basis – check implementation – acceptability relatively straightforward to add extra features to a soil model extra features imply additional soil parameters and additional calibration tests seek adequate complexity in modelling – match complexity of model to availability of data and needs of application
Promotion of use of advanced numerical models in practice • • • • •
Education, education, education! Keep it simple Build on familiar foundations Unification not disintegration Develop respect
Education, education, education! • has anything from the past 50 years of research in soil mechanics entered the undergraduate curriculum? • is the answer to that challenge close to zero? • what about concepts of critical state soil mechanics (≠ Cam clay)? • is anything approaching even a simple complete soil model introduced into a typical undergraduate degree programme? • greater understanding and appreciation of soil models can only be obtained by wide appropriate introduction at least into graduate degree programmes
SBCSSM
GM
Education, education, education! • are these two books (Soil behaviour and critical state soil mechanics, (1990) CUP; Geotechnical modelling, (2004) Spon) helpful? • defining a syllabus for educating engineers into the possibilities and problems of soil modelling • also useful for continuing professional development courses for practising engineers – harder to convince them • target the young!
Keep it simple • adequate complexity in geotechnical modelling • user of modelling should have some idea of phenomena expected to be important • ensure that these phenomena are included in the modelling • which aspects of soil response are first order or second order for performance of a geotechnical system? • experience? • careful parametric study 3
0.25
Disp (m) displacement (m) 0.20
no strengthening No Strengthening
Mohr-Coulomb model (regulators) 0.15
0.10 Anchors groundGround anchors 0.05
Counterforts & Piles 0.00 0
Mair and Muir Wood (2001)
1
2
time (s)
3 Time (s) 4
5
counterforts and piles
6
Build on familiar foundations • engineers more likely to use models which are incrementally different from ones with which they have some familiarity • (than models which adopt a completely different language) • certain models (with minor variations) generally available in geotechnical numerical analysis programs (EPP Mohr Coulomb, Cam clay) • some models can be readily developed from the teaching on soil strength – part of every undergraduate programme (EPP Mohr Coulomb) • (undergraduate background for Cam clay less ubiquitous) • hence: develop hierarchical models from elastic-perfectly plastic Mohr-Coulomb and Cam clay
Unification not disintegration • separation in companies between structural and geotechnical divisions • sparing in information in requests for parameters: control • plead guilty in universities too! • from first year of typical civil engineering degree: separate units in structures, soil mechanics, hydraulics … • do we make enough effort to introduce unifying units requiring combined appreciation of two or three of these subjects? • soil-structure interaction: obvious vehicle for unification/integration • learn by doing • numerical analysis programs (black boxes?) not essential
integral bridge abutment – remove need for bearings between deck and supports but what are stresses on abutments?
thermal expansion of bridge deck abutments move towards backfill passive loading – governed by strength of backfill? abutments can be flexible – vertical support for deck
numerical studies of actual prototype show that strength of backfill has no effect on horizontal stresses on abutment! relative stiffness of abutment and backfill is important
look at stress paths from numerical analysis elastic-perfectly plastic Mohr-Coulomb model dominant effect: increase in mean stress with little change in shear stress not heading towards failure
Develop respect • difficulties with numerical modelling (Potts, 2003): - because there is no standard numerical strategy for implementation of nonlinear models - because some constitutive models seem to be unable to give reasonable predictions - because, even for apparently simple problems, the results of numerical modelling can be very dependent on the decisions made by the user
What happened at Nicholl Highway was in part down to over-reliance by engineers on computerised soil analysis programs, he believes. These are "far more sophisticated than the people using them", Davies asserts. "What matters is how you put the data in to start with. You need to look at the overall problem." He maintains that use of inappropriate data in modelling soil behaviour skewed temporary works design in the wrong direction – steelwork was found to be under-strength. Engineers increasingly "do things to a set routine rather than thinking for themselves".
PLAXIS analyses same problem – different modellers spread of predictions (Schweiger, 2003)
Develop respect • restrict use to experts? (Potts…) • or educate students to explore, discover, understand? • need for communication between computer modellers and designers • education in respective languages • always support advanced modelling with back-ofthe-envelope estimates
Promotion of use of advanced numerical models in practice • • • • •
Education, education, education! Keep it simple Build on familiar foundations Unification not disintegration Develop respect