W(H)YDOC 05

2nd International Workshop of Young Doctors in Geomechanics

Paris, November 23-25, 2005
Ecole Nationale des Ponts et Chaussées

Probabilistic methods applied to

Geotechnical Engineering

Hong Kong

Dipl. Ing. Consolata Russelli Supervisor: Prof. Pieter A. Vermeer

1 Co-Advisor: Prof. A. Bárdossy
 Contents

1. Research motivation

2. Overview of the probabilistic analysis

3. The Point Estimate Method (PEM)

4. PEM application to geotechnical problems

5. Conclusions

2
 1. Research motivation

1.
Research
motivation

2.
Overview of
the
probabilistic
analysis
100%
3. The Uncertainty
Point
Estimate
Method

4. PEM
application
to
geotechnical
problems

5.
Conclusions

3
 1. Research motivation

1.
Research
motivation

2.
Overview of
the
probabilistic
analysis
100%
3. The Uncertainty
95%
Point
Estimate Uncertainty
Method

4. PEM
application
to
geotechnical
problems

5.
Conclusions

3
 1. Research motivation

1.
Research
motivation

2.
Overview of
the
probabilistic
analysis
100%
3. The Uncertainty
80%
95%
Point
Estimate Uncertainty
Uncertainty
Method

4. PEM
application
to
geotechnical
problems

5.
Conclusions

3
 1. Research motivation

1.
Research
motivation

2.
Overview of
the
probabilistic
analysis
100%
3. The Uncertainty
50%
80%
95%
Point
Estimate Uncertainty
Uncertainty
Method

4. PEM
application
to
geotechnical
problems

5.
Conclusions

3
 1. Research motivation

1.
Research
motivation

2.
Overview of
the
probabilistic
analysis
100%
0%
3. The Uncertainty
50%
80%
95%
Point Uncertainty
Estimate Uncertainty
Uncertainty
Method

4. PEM
application
to
geotechnical
problems

5.
Conclusions

3
 Geotechnical uncertainties

¾ Geological anomalies
¾ Inherent spatial variability of soil properties
1. ¾ Scarcity of representative data
Research
motivation ¾ Changing environmental conditions
¾ Unexpected failure mechanisms
2.
Overview of
¾ Simplifications and approximations adopted in geotechnical models
the
probabilistic ¾ Human mistakes in design and construction
analysis

3. The
Point
Deterministic analysis leads to extremely
Estimate
conservative design with significant failure
Method
probability :
4. PEM
application
to
geotechnical
problems
unable to account for uncertainties in material
5. and load properties.
Conclusions

4 Colorado
 How to deal with uncertainties?

“Uncertainty is inevitable”
1.
Research
motivation
Lack of perfect knowledge concerning phenomena
and processes involved in problem definition and
2.
Overview of resolution.
the
probabilistic
analysis

3. The
Point
Implementation of probabilistic analysis required
Estimate
Method
- uncertainties rationally quantified and systematically
4. PEM incorporated into the design process,
application
to
geotechnical
problems - means to evaluate uncertainties influence on the
likelihood of satisfactory performance for an
5.
Conclusions engineering system.
La Conchita, California

5
 Reluctance in adopting
probabilistic analysis

1.
Research
motivation
1. Engineers´ training in probability theory often limited to basic information during their
early years of education.
2.
Overview of
the
probabilistic
2. Less comfortable dealing with probabilities than with deterministic analysis.
analysis

3. Common misconception that it requires significantly more data, time and effort.
3. The
Point
Estimate
Method 4. Few published studies illustrate its implementation and benefits.

4. PEM
application
to
geotechnical
problems Deterministic analysis and probabilistic approach as

5.
complementary measures of acceptable design !!!
Conclusions

6
 2. Overview of the probabilistic analyses

Choice of geotechnical problems

1.
Research
motivation

2.
Overview of
Application of probabilistic methods
the
probabilistic
analysis

3. The
Point Results in terms of statistics values and
Estimate
Method probability distribution function
function of performance

4. PEM
probability density

application
µg(x) Comparison with Monte
function g(xi)

to f[g(xi)]
geotechnical Carlo Method
problems
νg(x)
5.
?
Conclusions

σg ( x ) σg ( x ) g(xi)
7
 Probabilistic methods analysed

1.
Research
motivation
¾ The First Order Second Moment Method (FOSM)
2.
Overview of
the
probabilistic ¾ The Second Order Second Moment Method (SOSM)
analysis

3. The
Point
Estimate
¾ The Point Estimate Method (PEM)
Method

4. PEM
application ¾ Monte Carlo Simulations (MC)
to
geotechnical
problems

5.
Conclusions

8
 3. The Point Estimate Method
(Rosenblueth, 1975)

Computationally straightforward technique for uncertainty analysis:

1. capable of estimating statistical values of a model output involving several stochastic variables,
Research
motivation
correlated or uncorrelated, symmetric or non-symmetric.

2.
Overview of Weighted average method similar to numerical integration formulas involving “sampling points”
the and “weighting parameters”.
probabilistic
analysis

3. The ¾ Requires little knowledge of probability concepts and applies for any probabilistic distribution.
Point
Estimate
Method
¾ Widely applied for reliability analysis and evaluation of failure probability.

4. PEM
application
to
geotechnical
problems Aim: replace probability distributions for continuous random variables with discrete equivalent
functions having the same mean value, standard deviation and skewness coefficient!
5.
Conclusions

9
 Procedure for implementing the PEM

1. Consider a relationship between performance function f(Xi) and input random variables.

2. Compute locations of sampling points (2n calculations) :

1.
Research
motivation
1
ν X i  
2 2
 ν Xi  

probability density function

ξXi + = + 1+   f(tanφ´)
2.
2   2  
Overview of  
the
probabilistic ξ Xi − = ξ Xi + − ν Xi
Ptanφ´- Ptanφ´+
analysis

3. The x i − = µ Xi − ξ X i − ⋅ σ Xi
Point
Estimate
x i + = µ Xi + ξ Xi + ⋅ σ Xi tanφ´
Method
µ tan ϕ′ − ξtan ϕ′− ⋅ σtan ϕ′ µtanφ´ µtan ϕ′ + ξtan ϕ′+ ⋅ σtan ϕ′
4. PEM
application
to
geotechnical
problems

5.
Conclusions

10
 Procedure for implementing the PEM

1. Consider a relationship between performance function f(Xi) and input random variables.

2. Compute locations of sampling points (2n calculations) :

1.
Research (e.g. friction angle)
f(X,Y)
motivation
1
ν X i  
2 2
 ν Xi  
ξXi + = + 1+  
2.
2   2  
Overview of  
the
probabilistic ξ Xi − = ξ Xi + − ν Xi
analysis

3. The x i − = µ Xi − ξ X i − ⋅ σ Xi
Point
Estimate
x i + = µ Xi + ξ Xi + ⋅ σ Xi
Method

4. PEM (e.g. cohesion)

application
to
geotechnical 3. Determine the weights Pi (probability concentrations) to obtain all the point estimates.
problems

5.
single random variable : associated weights:
Conclusions
ξxi −
PX i + = PX i − = 1 − PXi + Ps1s 2 = PXs1 ⋅ PXs 2
10 ξxi + + ξxi −
 Procedure for implementing the PEM

4. Determine the performance function value f(Xi) at each sampling point locations.

1. Sign Pi c´ tan( ϕ′ ) qf
Research ++ 0.043 14.647 kPa 0.523 653.548 kPa
motivation +- 0.457 3.003 kPa 0.409 194.171 kPa
-+ 0.043 14.647 kPa 0.409 391.703 kPa
-- 0.457 3.003 kPa 0.523 365.685 kPa
2.
Overview of
the
probabilistic
analysis

5. Determine the first three moments of the performance function:

3. The
Point 2n
Estimate
Method
µ f (X i ) = ∑ P ⋅ f (X )
i =1
i i

2n

∑ P ⋅ (f (X ) − µ ( ) )
4. PEM 2
application σ 2
f (X i ) = i i f Xi
to i =1
geotechnical
2n

∑ P ⋅ (f (X ) − µ ( ) )
problems 1 3
ν f (Xi ) = i i f Xi
σ 3
f (Xi ) i = 1
5.
Conclusions

11
 4. PEM application to geotechnical
problems

1.
Research Terzaghi´s Bearing Capacity: shallow foundation on a cohesive homogeneous soil
motivation

2.
- PEM with correlated input random variables.
Overview of
the - PEM with uncorrelated input random variables.
probabilistic
analysis
- Results comparison.

3. The
Point
Estimate
Method
Slope Stability Analysis: fill embankment on undrained clay
4. PEM
application
- PEM with uncorrelated input random variables.
to
geotechnical
- Results comparison.
problems

5.
Conclusions

12
 Terzaghi´s bearing capacity

B=2m

qf
q q
1.
Research
1
motivation
q f = c′ ⋅ N c + q ⋅ N q + ⋅ γ ⋅ B ⋅ Nγ
2 q = 10 kPa
2.
Overview of γ = 15 kN/m3
the
probabilistic
analysis

3. The
Point Gaussian (normal) distribution Lognormal distribution
probability density function

probability density function

Estimate
Method
µ tan ϕ′ = 0.47 µ c′ = 4 kPa
f(tanφ´) σ tan ϕ′ = 0.06 f(c´) σ c′ = 3.3 kPa
4. PEM
application ν tan ϕ′ = 0 ν c′ = 3.1
to
geotechnical
problems
5% fractile 5% fractile
5.
Conclusions
0.36 0.47 0.58 tan φ´ 0 4 10 Cohesion c´ (kPa)

13
 PEM results with uncorrelated ( ρ tan ϕ′ c′ = 0 )
input variables

µ q f (kPa ) σ q f (kPa ) ν qf COVq f

305.39 110.48 1.2 0.36
1.
Research
motivation 0,0045

0,004
Probability density function of qf

2.
Overview of 0,0035
the
probabilistic
0,003
analysis

0,0025

3. The
Point 0,002
Estimate
Method 0,0015
σqf σqf
0,001
4. PEM
application
0,0005
to
geotechnical
µqf
0
problems
25
50

75

0
5
0

5
0
5
0
5
0

0
5
0
5
0
5

0
5
0
5
5

0
0
0
50
00
50

00
1

10
12
15
17

20
22
25
27
30
32

34
37
39
42
44
47

49
52
54
57
60
65

90
00

11
14
16
19
0,

5. Bearing capacity qf (kPa)

Conclusions

14
 PEM results with correlated input variables
Rosenblueth (1981) – Two random variables

 1

   ν X1 
3
   ν X2 
3
 2 
Ps1s 2 = PXs1 ⋅ PXs 2 + s1 ⋅ s 2 ⋅  ρ X1X 2 /  1 +    ⋅ 1 +    
     
1.
   2     2   
Research  
motivation

2.
Standard deviation
Overview of ρ (tanϕ′,c´) Mean value (qf)
(qf)
Skewness (qf)
the
-1,0 294.110 34.755 0.000
probabilistic
-0,9 295.279 46.193 0.614
analysis
-0,8 296.449 55.290 0.678
-0,7 297.618 63.067 0.647
-0,6 298.788 69.966 0.594
3. The
-0,5 299.957 76.225 0.537
Point
-0,4 301.127 81.991 0.483
Estimate -0,3 302.297 87.361 0.432
Method -0,2 303.466 92.406 0.385
-0,1 304.636 97.174 0.341
0,0 305.805 101.706 0.300
4. PEM 0,1 306.975 106.032 0.262
application 0,2 308.144 110.175 0.226
to 0,3 309.314 114.157 0.193
geotechnical 0,4 310.484 117.992 0.161
problems 0,5 311.653 121.695 0.131
0,6 312.823 125.278 0.102
0,7 313.992 128.751 0.075
5.
0,8 315.162 132.122 0.049
Conclusions 0,9 316.331 135.399 0.024
1,0 317.501 138.589 0.000

15
 Influence of the correlation coefficient
on PEM results of the bearing capacity

1.
Research 0,012
ρ tan ϕ′c′ = − 1.0 If ρ tan ϕ′c′ increseas then p.d.f . is wider
Probability density function
motivation
0,01 and probability values sligthly decrease.
2.
Overview of 0,008 If ρ tan ϕ′c′ decreases then p.d.f . is narrower
the
probabilistic
ρ tan ϕ′c′ = 0 and probability values increase.
analysis 0,006

0,004
3. The
Point ρ tan ϕ′c′ = 1.0
Estimate 0,002
Method

0
4. PEM
0

0
30

60

90
1

00

00

00

00
12

15

18

21

24

27

30

32

35

38

41

44

47

50

53

57

60

70
00

application

10

13

16

19
0,

to
geotechnical Bearing capacity (kPa)
problems c.c.= 0 c.c. = 0.1 c.c. = 0.2 c.c. = 0.3 c.c. = 0.4 c.c. = 0.5 c.c. = 0.6 c.c. = 0.7
c.c. = 0.8 c.c. = 0.9 c.c. = 1.0 c.c. = -0.1 c.c. = -0.2 c.c. = -0.3 c.c. = -0.4 c.c. = -0.5
5. c.c. = -0.6 c.c. = -0.7 c.c. = -0.8 c.c. = -0.9 c.c. -1.0
Conclusions

16
 Comparison of PEM and MC results

Method µqf(kPa) σqf (kPa) νqfCOVqf

Monte Carlo ρ = 0 304.72 115.94 1. 9 0.38
1. PEM ρ=0 305.39 110.48 1. 2 0.36
Research
motivation PEM ρ = − 0.6 298.79 69.97 0.59 0.23

2.
Overview of 0,007
the
PEM ρ = − 0.6
Probability density function

probabilistic
0,006
analysis
FS=2
0,005
3. The
Monte Carlo ρ = 0
Point 0,004
Estimate
Method
PEM ρ = 0
0,003

4. PEM 0,002
application deterministic mean value
to 0,001
geotechnical
problems 287.6 kPa
0
30
60
90

0
0
0
0
0
0
0
5
5
5
5
5
5
5
5
0
0
0
00
00
00
00
1

12
15
18
21
24
27
30
32
35
38
41
44
47
50
53
57
60
70
00

5.

10
13
16
19
0,

Conclusions
Bearing capacity qf (kPa)

17
 Slope stability analysis
Bishop´s simplified method of slices

n
1
1.
Research
∑ [ci ⋅ bi
i =1
+ Wi ⋅ tan ϕi ]⋅
m α (i )
motivation FS = n

∑ W ⋅ sin α
i =1
i i
2.
Overview of
the O (37.5 m, 19 m)
probabilistic
analysis

3. The R = 37.5 m
Point
Estimate
Method
6m
FILL
4. PEM
application 4m CRUST
to
geotechnical
problems 8m MARINE CLAY

5.
Conclusions LACUSTRINE CLAY
6,5 m

18 Source: J.T. Christian, C.C. Ladd, G.B. Baecher; „Realiability and Probability in Slope Stability Analysis“
 Input uncorrelated soil parameters

Layer Mean Value Standard deviation Coefficient of Variation

γ = 20 kN / m 3 1 0.071
Fill
ϕ´u = 30 ° 3 0.1
1. γ = 18 . 81 kN / m 3 0.94 0.049
Crust
Research c u = 40 kPa 10 0.25
motivation γ = 18 . 81 kN / m 3 0.94 0.049
Marine Clay
c u = 34.5 kPa 6 89 0.2
γ = 20 . 31 kN / m 3 0.99 0.049
2. Lacustrine Clay
c u = 31.2 kPa 9 98 0.32
Overview of
the
probabilistic
analysis probability density function
probability density function

probability density function

probability density function

of marine clay unit weight

of marine clay cohesion

of fill friction angle

3. The
of fill unit weight

f(φ) γ
f(φ) γ
f(φ) f(φ)
cu
Point
Estimate
Method

27° 30° 33° 19 20 21 17.9 18.81 19.8 27.5 34.5 41.5

4. PEM Marine clay cohesion (kPa)
Fill friction angle (°) Fill unit weight (kN/m3) Marine clay unit weight (kN/m3)
application
to

of lacustrine clay unit weight

probability density function
probability density function

probability density function

probability density function
of lacustrine clay cohesion
geotechnical
of crust unit weight
of crust cohesion

problems f(φ)
cu γ
f(φ) f(φ)
cu γ
f(φ)

5.
Conclusions

30 40 50 17.9 18.81 19.8 21.5 31.2 40.9 19.3 20.31 21.3

Crust cohesion (kPa) Crust unit weight (kN/m3) Lacustrine clay cohesion (kPa) Lacustrine clay unit weight (kN/m3)
19
 Comparison of PEM and MC results

Method applied µ FS σ FS COVFS ν FS

Monte Carlo 1.462 0.279 0.191 0.012
1. PEM 1.535 0.286 0.186 0.00004
Research
motivation

1,6
2.
Overview of
1,4
Probability density function

the
probabilistic
1,2
analysis
deterministic mean value
1
3. The
Point 0,8
Estimate
Method 0,6

0,4
Gaussian fit
4. PEM
application 0,2
to
1.532
geotechnical 0
problems
0,11
0,22
0,33
0,44
0,55
0,66
0,77
0,88
0,99

1,21
1,32
1,43
1,54
1,65
1,76
1,87
1,98
2,09

2,31
2,42
2,53
2,64
2,75
2,86
2,97
0

1,1

2,2
5. Factor of safety
Conclusions
Monte Carlo PEM (Rosenblueth, 1975)

20
 5. Conclusions
PEM advantages vs. MC simulations

Reasonably robust and satisfactorily accurate for a wide range of

1.
Research practical problems !
motivation

2.
Overview of
the
probabilistic
analysis
¾ Results as reliable and accurate as MC simulations.

3. The
Point ¾ Smaller computational effort for a comparable degree of accuracy
Estimate
Method

4. PEM ¾ No need of knowledge of p.d.f. shape of input random variables.

application
to
geotechnical
problems ¾ Behaviour of non-linear function well captured.
5.
Conclusions

21
 5. Conclusions
How to cope with PEM drawbacks

¾ Performance function p.d.f. to be assumed, thus introducing uncertainty.

In Soil Mechanics normal and lognormal distributions frequently result as
1.
Research
output of probabilistic analysis.
motivation

2.
Overview of
the
¾ If more accuracy required than larger number of input variables necessary, i.e. number
probabilistic
analysis
of required evaluations too high to be implemented practically.
Rosenblueth approximation method (1981) for Gaussian distributed uncorrelated input
3. The
Point
variables.
Estimate
Method

4. PEM
application
¾ Results poor and not accurate for discontinuous functions or functions having
to
geotechnical
discontinuous first derivatives and for large COVs of input variables.
problems
Typical geotechnical problems described by continuous functions, whose “non-linearity”
5. not difficult to be treated by PEM. Small COV values in geotechnical literature, frequently
Conclusions
lower than the unity.
22
 5. Conclusions
Final observations

¾ Different probabilistic methods applied to geotechnical problems to find out

the most suitable one for the geotechnical field and for a further analysis.
1.
Research
motivation
¾ PEM gave as reliable and accurate results as MC with less computational effort.
2.
Overview of
the ¾ It was easily applied to a multivariate problem.
probabilistic
analysis

3. The ¾ Output as a lower bound for the evaluation of failure probability, because effects of
Point
Estimate other factors not included in the analysis.
Method

4. PEM ¾ Spatial variability ignored to simplify the calculations assuming perfect

application
to autocorrelation, obtaining more conservative results.
geotechnical
problems

5. PEM: simple, but powerful technique for uncertainty analysis.

Conclusions
Its use in geotechnical reliability analysis justified by experience and theory.
23
 Thanks for
your attention !

For further clarifications and discussion,

please contact
russelli@igs.uni-stuttgart.de

Bainbridge Island, WA USA

24