Professional Documents
Culture Documents
Research Paper
a r t i c l e i n f o a b s t r a c t
Article history: A model prediction using FE modeling is performed for a trial embankment on soft clay in Ballina,
Received 23 January 2017 Australia. The comparison between the prediction and the in situ measurement exhibits significant dif-
Received in revised form 13 May 2017 ferences. A detailed analysis is performed to validate the model and determine necessary improvements.
Accepted 6 June 2017
Sensitivity analyses elaborate upon the most dominant constitutive parameters. Based on site measure-
ments, inverse analyses are performed to identify the optimum parameters. The inverse analysis
approach is validated regarding its dependency on the objective function and data incorporated. A com-
Keywords:
parison between the prediction and the back-calculated parameters confirms the importance of engi-
Embankment
Soft soil
neering judgment as well as the sensitivity of the modeling strategies to the input information.
Consolidation Ó 2017 Elsevier Ltd. All rights reserved.
Finite element modeling
Parametric study
Inverse analysis
http://dx.doi.org/10.1016/j.compgeo.2017.06.002
0266-352X/Ó 2017 Elsevier Ltd. All rights reserved.
88 N. Müthing et al. / Computers and Geotechnics 93 (2018) 87–103
Geometry Physical
Settlements processes Lab data Field data
conditions
Relevant
Calculation
parameters
model
Model Inverse
validation analysis
Calibrated
parameters
Agreement? No
Parameter Uncertainty
validation analysis
Model Output
Yes response uncertainty
model response. Therefore, in a first step, a parametric study using provide a better understanding and assessment of the performed
metamodeling strategies to perform global sensitivity analyses is improvement strategy.
used to elaborate upon the most relevant constitutive parameters.
However, it has to be considered that the predominance of certain
2. Finite element model for embankment prediction
model parameters might be caused by model assumptions. A close
evaluation of this enables the assessment of the method and high-
The test embankment is simulated by both a 2D plane strain
lights what has to be considered when conducting this type of
model and a 3D Finite-Element (FE) model. Plaxis 2D (version
analysis.
2015) and Plaxis 3D (version AE) are used for conducting the 2D
With the knowledge of the relevant model parameters, inverse
and 3D simulations, respectively.
analyses using a Generic Algorithm (GA) for parameter identifica-
tion can be performed. The aim of this inverse analysis is to iden-
tify an optimum parameter set to achieve a good fit between the 2.1. 2D FE model
predicted and measured values. This can be particularly useful
when only limited experimental data on the soil properties are The geometry and mesh discretization of the 2D FE-model are
available and/or a sophisticated constitutive model requiring shown in Fig. 2. The soil deposit present prior to embankment con-
advanced laboratory tests is applied. Of course, the inverse analysis struction is simplified into 6 horizontal layers. The working plat-
itself and its output are strongly dependent on the input informa- form and horizontal drainage layers are installed prior to the
tion provided for the parameter optimization. Here, particularly, embankment construction. A total of 22 wick drains with a spacing
the influence of the chosen objective function for the optimization of 1.2 m are used in this test embankment, and the drains are
in terms of type, amount and quality of data incorporated needs to installed up to 15 m beneath the ground surface. According to
be evaluated. Thus, a step-wise inverse analysis approach consid- the in situ investigation, the ground water level is assumed to be
ering different objective functions is preferred and executed in 0.72 m below the ground surface.
the present study. This strategy allows for the assessment of the Based on the trial analysis, the mesh discretization is deter-
tool of inverse analysis regarding its dependency on the choice of mined such that model responses are mesh independent. Specifi-
objective function and data incorporated. Consequently, the cally, Fig. 3 shows the relation between the model response and
importance of engineering judgment and the sensitivity of the the number of elements. The point that is located at the center of
method to input information can be illustrated. the 2D model and 4 m below the surface is chosen to check the
After identifying an optimum parameter set, a critical examina- excess pore pressure after the embankment construction is com-
tion of the identified parameter values needs to be conducted. pleted. The model responses using 14,100 elements are taken as
Through a comparison of the initial and optimum parameter val- the reference value (42 kPa excess pore pressure) to calculate the
ues, a physical justification needs to be achieved. normalized value of the other cases. As seen, when the number
Moreover, the initial as well as final predictions are based on of elements increases from 14,100 to 18,200, the variation in the
parameter values that are not precisely known but that are mean model response is less than 1%; hence, models with 14,100 ele-
values with ranges estimated by engineering judgment. To observe ments are used in this research, and the total number of nodes is
the impact of parameter uncertainty and identify the most domi- approximately 114,000.
nant parameter ranges on the model output, an uncertainty analy- The bottom of the model is assumed to be fixed, with no defor-
sis needs to be executed. A comparison of the model response mation allowed. At the left and right boundaries, the horizontal
uncertainty between the initial and optimum parameters will help displacements are fixed, whereas vertical displacements are
N. Müthing et al. / Computers and Geotechnics 93 (2018) 87–103 89
1.05 cally divided into five construction steps. All construction steps
are conducted as consolidation analysis to guarantee an adequate
Normalized excess pore pressure [-]
Unit: m
30 2D cross section
the soil and drain permeability together. Using this equivalent per- on that of the undisturbed soil (kh;ps ). Here, Indraratna and Redana
meability, the variation in the degree of consolidation with time is [8] proposed the following equations:
comparable to the behavior using discrete drainage elements. The
equivalent permeability kv e in the vertical direction is determined ks;ps b
¼ ð4Þ
using the following equations proposed in [6]: kh;ps kh;ps ðlnðnÞ þ kh;ax lnðsÞ 0:75Þ a
k h;ax s ks;ax
2
!
2:5l kh
kv e ¼ 1þ kv ð1Þ 0:67ðn sÞ3
lD2e kv a¼ ð5Þ
n2 ðn 1Þ
n kh
2
2l kh
l ¼ ln þ lnðsÞ 0:75 þ p ð2Þ 2ðn 1Þðnðn s 1Þ þ 0:33ðs2 þ s þ 1ÞÞ
s kv 3qw b¼ ð6Þ
n2 ðn 1Þ
where l denotes the drainage length; De is the diameter of the unit Additionally, the permeability varied with the void ratio based
cell; kh and kv are the initial horizontal and vertical permeability for on the following equation:
the undisturbed soil, respectively; n ¼ De =dw (dw = diameter of
drain); s ¼ ds =dw (ds = diameter of smear zone); and qw is the dis- k De
log ¼ ð7Þ
charge capacity of the PVD. It should be noted that when applying k0 ck
this approach to two-way drainage conditions, the drain length
with k and k0 being the current and initial permeability, respec-
parameter l is half the thickness of the PVD-improved zone.
tively; De being the variation in the void ratio; and ck being the per-
Under plane strain conditions, the drainage is usually modeled
meability change index, which is assumed to be 1.125 for the silt
as a plane sink. Since this does not capture the axisymmetric nat-
and clay layers [5]. It should be noted that large deformations were
ure of horizontal flow toward drains, the soil permeability and
not considered in this study.
space between the drains under the plane strain conditions should
be modified. Indraratna and Redana [7] proposed three methods to
match the permeability under axisymmetric and plane strain con- 2.5. 2D and 3D unit cell analyses
ditions: (1) Geometry matching, where the spacing between the
plane strain drains is varied while keeping the permeability con- To validate the aforementioned 2D and 3D modeling
stant; (2) Permeability matching, where the permeability under approaches for the drainage, unit cell analyses were conducted
the plane strain condition is varied while keeping the spacing and are illustrated in the following section.
between the drain constant; and (3) A combination of permeability The 3D unit cell model is shown in Fig. 5(a). The length and
and geometry matching. In the present study, the permeability width of this model are identical and equal 1.2 m. The thickness
matching method is adopted in the 2D modeling. The horizontal of each layer equals the thickness in the full-scale 3D model. The
permeability of the undisturbed soil under plane strain conditions drainage element is located in the center of the unit cell model.
kh;ps is calculated according to [8]: The 2D unit cell model has the same soil layer distribution and
position of the drain as in the 3D model. The permeability values
kh;ps 0:67ðn 1Þ2 for all layers and model types are given in Table 1. The numerical
¼ 2 ð3Þ staged construction steps and ground water level in the unit cell
kh;ax n ðlnðnÞ 0:75Þ
analysis are the same as in the full-scale embankment analysis
where kh;ax denotes the horizontal permeability of the undisturbed (see Sections 2.1 and 2.3). The model bottom is fixed, and in-
soil under axisymmetric conditions. plane deformations at the vertical boundaries of the unit cell
As mentioned before, the soil permeability in the smear zone model are not allowed. Moreover, the excess pore pressure cannot
under the plane strain condition (ks;ps ) can be determined based be dissipated through the vertical or bottom boundaries.
N. Müthing et al. / Computers and Geotechnics 93 (2018) 87–103 91
Fig. 5. Unit cell analyses: (a) model geometry (b) variation in the vertical settlements at point A (c) variation in the pore water pressure at point B (d) comparison of the pore
water pressure at points B and C in the 3D unit cell analysis.
Table 1
Values of permeability for different scenarios (Unit: m/day).
Three numerical scenarios are considered in both the 2D and 3D of the embankment, and the dissipation of the pore water pressure
unit cell analyses: In scenario 1, only drainage is modeled, while is faster. Consequently, the soil domain in scenario 1 settles faster
the disturbance due to the drain installation and its effect on the than that in scenario 2, while the final settlement is identical. Fur-
soil permeability is not considered. In scenario 2, the impact of thermore, it is found that the pore pressure variations at points B
the smear zone is modeled using a reduced horizontal permeability and C are identical in the 3D unit cell where the equivalent perme-
in the corresponding zone. Finally, the equivalent permeability ability concept is applied (see Fig. 5(d)). This is because the equiv-
concept is applied in scenario 3. alent permeability method approximates the average pore
In the unit cell analyses, three observation points are defined. pressure in the cell. When the equivalent permeability concept is
Point A is placed at the center of the model surface, while point compared with the explicit method, where the drain and smear
B is located 8 m beneath point A. The horizontal distance between zone are modeled, a slight disagreement of the pore pressures at
points A and B is 0.34 m. Point C is defined at the corner of the unit points B and C can be observed due to the different distances
cell, and it has the same depth as point B. between the measurement points and the drain. Generally, it can
The variations in the vertical displacements and pore water be concluded that both methods, the equivalent permeability 3D
pressure with respect to time for points A, B and C are shown in model and the 2D model, provide results that compare well with
Fig. 5(b)–(d). It can be observed that without modeling the smear the refined 3D model, where both the drain and the smear zone
zone, less pore water pressure is generated during the construction were explicitly simulated.
92 N. Müthing et al. / Computers and Geotechnics 93 (2018) 87–103
0 ¼ 1 sin u .
SSC model. Since this constitutive model does not account for the friction angle as K NC 0
Table 2
Values of model parameters for blind prediction.
to be constructed (point 1), the pore water pressure generated dur- 3. Prediction validation
ing the preparation of the working platform has been fully dissi-
pated. During the embankment construction, the stress paths at Fig. 8 shows the model responses when applying the initially
all three depths of the clay layers cross the initial yield surface obtained parameters (see Tables 1 and 2) from the 2D model using
(from points 1–2). This is consistent with the aforementioned discrete drainage with smear zone and from the 3D model using
assumption that k is determined after reaching the yield stress the equivalent permeability concept. The locations of the measure-
and justifies the assumption of modeling the initial compression ment points are given in Fig. 2. Three measurement positions over
behavior with j while setting k ¼ k1 from the testing data. After the depth were selected for comparison of the time-dependent set-
1000 days (point 3), the excess pore water pressures have fully dis- tlements (SP2, Magnet 1 and Magnet 2) and pore water dissipation
sipated. Additionally, it should be noted that in the upper two clay (VWP1a, VWP1b, and VWP1c). From a comparison of the 2D/3D
layers, localized failure occurs at the 47th day (the end of embank- finite element analysis with the in situ measurement data, the fol-
ment construction is the 49th day). After that, the stress path fol- lowing conclusions can be drawn:
lows the failure line. In this process, the settlement and excess The vertical displacements and pore water pressures at the dif-
pore pressure continue increasing until the end of embankment ferent measurement points are almost identical in the 2D and 3D
construction. With the development of the consolidation process, finite element models. This is consistent with the findings from
the effective stresses increase due to the excess pore pressure dis- the unit cell analysis in Section 2.5 and confirms the model valida-
sipation and bring the stress path away from the failure line. tion regarding dimensionality.
Fig. 7. Stress path at different clay layers [point0: construction starts, point1: start of the earth fill of the embankment, point 2: end of the earth fill of the embankment, point
pffiffiffi pffiffiffi construction]; (a) depth of 2 m in clay 1 layer, (b) depth of 5 m in clay 2 layer, (c) depth of 8.9 m in clay 3 layer. Here, q represents the
3: 1000 days after embankment
deviatoric stress, m ¼ 3=ð 3 cos hl þ sin hl sin u0 Þ, with hl being the Lode angle, [17].
94 N. Müthing et al. / Computers and Geotechnics 93 (2018) 87–103
Fig. 8. Comparison of numerical results (blind prediction) with in situ measurements for (a) vertical displacements, (b) total pore water pressures.
The comparison between the predicted and in situ-measured from the inherent variability to measurement inaccuracy and
settlements reveals that the numerical models adequately repro- model precision.
duce the deformation behavior for measurement point SP2 and To validate a model, it is of great importance to know which
Magnet 1 up to approximately 200 days. At the deeper measure- parameters are most significantly influencing the output or to
ment point (Magnet 2), the calculation models overestimate the which parameter variations the output is most sensitive accord-
settlements. It can be concluded that the compression behavior ingly. An appropriate method to identify the influence of different
of the clay layers below measurement point Magnet 1 in terms input parameters on a specific model output is the so-called sensi-
of the compression parameters should be identified more accu- tivity analysis.
rately. After 200 days, the calculation models predict a decay in In general, two groups of approaches for sensitivity analysis
the settlements, while the settlements in situ continue increasing. exist: local and global sensitivity analysis. In local sensitivity anal-
This holds true for all measurement points. This indicates that the yses, the parameters’ influences are examined independently from
modeling of the consolidation and/or creep behavior of the soft silt each other by a small parameter variation, comparable to the par-
and clay layer requires improvement. tial derivatives of the model. However, this method can only be
The comparison between the predicted and measured pore valid in the local region of the point in the input space or for con-
water pressures shows that the maximum pore water pressure tinuously linear models. Moreover, possible correlation effects can-
and dissipation behavior at measurement point VWP1a are ade- not be considered, as all parameter variations are performed
quately reproduced by the computational models. The in situ pore individually.
pressures at point VWP1a after approximately 300 days exhibit Since output results from complex non-linear models and large
oscillations, which may be due to seasonal groundwater table fluc- parameter ranges are anticipated in the present study, a global sen-
tuations as reported in [18] or could be caused by settlement of the sitivity analysis (GSA) is more appropriate. The employed variance-
VWP. Since water table fluctuations are not modeled and since no based GSA, introduced by Sobol [19] and further developed by Sal-
particular large displacement analysis was adopted, these effects telli [20], allows for an impact quantification of input variances on
cannot be reproduced by the FE model used within the present the output variances. The resulting first-order index Si is defined as
study. follows:
The maximum pore water pressures and consequently their dis-
sipation behavior at the other measurement points (VWP1b and ðyTA yC i Þ nðyA Þ2
Si ¼ ð8Þ
VWP1c) are significantly underestimated by the computational ðyTA yA Þ nðyA Þ2
models. This indicates a too fast predicted consolidation behavior,
which can be improved by a more accurate identification of the To additionally consider correlation effects between different
stiffness and/or permeability of soil. parameters, the total effect index ST i is employed:
For the sake of simplicity and to reduce the computational
ðyB yC i ÞT ðyB yC i Þ
effort for the later parameter improvement study, the 2D finite ele- ST i ¼ ð9Þ
ment model is used. Since in the 2D model the difference in the 2yTB yB 2nðyB Þ2
computational effort between the discrete drain modeling and
where yA ; yB and yCi denote vectors containing model outputs gener-
the equivalent permeability method is of minor importance, sce-
ated from samples all over the input space of the parameters i, and
nario 2 is given preference in this study. A and y B are the corresponding mean values.
y
According to [20], GSA evaluations require 500–10,000 model
4. Parametric study runs. More complex models require more runs to obtain reliable,
stable mean values and variance. Even with a simple 2D FE model,
4.1. Global sensitivity analysis this is significantly time consuming. A tool to overcome this type of
computational problem is provided by meta- or surrogate model-
The blind prediction is executed with parameter values that are ing approaches. Once the metamodel is established, the computa-
not precisely known. Rather than as exact values, the parameter tional costs are very low; thus, 10,000 runs could be performed for
values need to be understood as the mean values of ranges esti- the GSA in the present study.
mated by engineering judgment. The range of the parameter vari- Within metamodel approaches, computationally expensive FE
ations varies according to the precision of the parameter models are replaced by efficient mathematical regression functions
determination. However, it is always influenced by many factors, [21]. The regression function correlating the input and output data
N. Müthing et al. / Computers and Geotechnics 93 (2018) 87–103 95
set is the so-called metamodel. Each input is assumed to be an s- considering creep as represented by the creep parameter, l is also
dimensional vector that represents the different model parame- of importance. The friction angle u0 is found to have a small influ-
ters, and there are m observation points in the outputs. For the ence on the settlements. This can be anticipated since volumetric
training of a metamodel, first, 150 input parameter samples (150 deformation is dominant for the embankment behavior and since
sample points with s dimensions) are generated via the Latin shear deformation is less influential in this case.
Hypercube sampling method in the pre-defined ranges of the cor- Generally, GSA can be performed using any type of output infor-
responding parameters. Then, the corresponding output values mation (e.g., deformation as well as pore water pressure) gained at
(150 outputs for m observation points) can be obtained via FE sim- any position and time point. The choice of the selected reference
ulation. Within this framework, the metamodel is the regression output certainly has a major influence on the calculated sensitivity
function mathematically correlating the inputs (s dimensions) with indices, as the hydraulic or mechanical model responses will have
the output at each observation point. The type of regression func- different sensitivities toward the different constitutive parameters.
tion, e.g., polynomial regression or radial basis function, must be In the present study, the focus was set on the adequate modeling of
chosen according to the complexity of the computational problem. the displacement and consolidation behaviors. Thus, the vertical
The metamodel should be validated before its employment by displacements in the center line of the embankment at the final
comparison of independent test samples from the FE and meta- stage of the measurements were selected as the most relevant
model. This validation ensures the accuracy of the metamodel. model output and were chosen as the reference output for the
Using a metamodeling approach, the total calculation time for a GSA. Of course, if other aspects come into focus, additional GSA
GSA can be reduced to a fraction of the time needed without this should be considered.
approach. The application of GSA to geotechnical problems is
demonstrated in [22–25]. 4.2. Inverse analysis
In the present study, a GSA is performed with a metamodeling
approach using proper orthogonal decomposition in combination To obtain a parameter set that gives an optimum fit between
with radial basis functions (POD-RBF); see [26]. the predicted and measured values, inverse analysis is applied
To reduce the dimension of the metamodel, only the clay layers for parameter identification [27–29]. In the inverse analysis, first,
assumed to dominate the compression of the subsoil are consid- an objective function is defined as the normalized sum of normal-
ered in the GSA. Moreover, a coupled behavior of the three clay ized differences between calculation and measurement:
sub-layers is assumed. This means that all three clay sub-layers " 2 #
experience identical relative changes for each parameter. Six input 1X M
ycalc ðXÞ ymeas
f ðXÞ ¼ wi i i
ð10Þ
parameters are introduced for each clay sub-layer: the permeabil- M i¼1 ymeas
i
ity k, over-consolidation ratio OCR, modified compression index k ,
modified recompression index j , modified creep index l , and where X is the vector of the input parameters of the constitutive
friction angle u0 . The parameter space is assumed through an accu- model to be identified; ycalci ðXÞ and ymeasi are the calculated result
racy estimation of the parameter assessment from laboratory and and field measurements, respectively; M is the number of measure-
field test data and by engineering judgment: the permeability k ments; and wi is an optional weighting factor for the corresponding
is varied by a factor of 10 in both directions, and the other param- measurement points, assumed to be 1 in the present study.
eters are varied by ±20%. As output, the settlements after The aim of the inverse analysis is to identify the set of input
1000 days at the ground surface (SP2) are evaluated. parameters for which a minimization of the objective function is
Fig. 9 shows the results of two GSAs performed for the embank- reached. The minimum of the objective function may serve as a
ment model. The GSA including the OCR as an input parameter measure for comparison of the quality of different inverse analy-
illustrates the clear dominance of the OCR on the settlement out- ses: a smaller f ðXÞ indicates a better optimization result. This can
put. Since the OCR defines in which stress range j lies and in be found using evolutionary algorithms, e.g., Genetic Algorithms
which stress state k is applied, it has an augmented influence on (GA) or Particle Swarm Optimization (PSO) [30,31]. For the present
the final settlement value. When neglecting the OCR as a variable study, a genetic algorithm (GA) is chosen for solving the optimiza-
for the GSA, a clearer picture of the settlement sensitivity toward tion problem. This type of algorithm is inspired by Darwin’s evolu-
the other parameters can be derived. The compression index k tion theory. Details of applying GA algorithms to similar problems
and permeability k exhibit the highest sensitivity, while the recom- can be found in [25]. It should be noted that all inverse analyses in
pression index j is of minor influence. This seems plausible, as the this study were performed using the metamodel to identify the
settlements are determined in a consolidation analysis whereby uncertain model parameters; then, the identified parameters were
stiffness and permeability are expected to have a dominating influ- adopted in the FE model to evaluate the model responses and com-
ence. As anticipated in the assumptions for the constitutive model, pared with the real measurements.
As described above, the inverse analysis itself and its output are
strongly dependent on the input information provided for the
parameter optimization, namely, on the optimized input parame-
ters and the output parameters incorporated in the objective func-
tion. In particular, the chosen output employed in the objective
function, in terms of type, amount and quality of data incorporated,
may have a significant influence on the quality of the inverse anal-
ysis results. Thus, in the present study, a hybrid inverse analysis
approach considering different optimized input parameters as well
as different output parameters for the objective function is per-
formed. The input parameters were defined according to the rele-
vant parameters identified in the sensitivity analysis. Specifically,
three inverse analysis scenarios are studied:
Scenario I: Optimized input parameters: relevant parameters
Fig. 9. Sensitivity indices from GSA with respect to settlements at the ground for the clay layer (OCR; k ; j ; l , and k) assuming a coupled behav-
surface (point 0 in Fig. 2) after 1000 days. ior of the three sub-layers such that any relative change in one
96 N. Müthing et al. / Computers and Geotechnics 93 (2018) 87–103
parameter is assumed to be identical for this parameter in all three scenario Ia, a clear improvement in the model response compared
sub-layers (see Section 4.1). to the blind prediction (see Fig. 8) can be achieved.
output parameters for objective function: Scenario I(a) – total Fig. 10(c) and (d) presents the results obtained when using the
settlements at 3 positions over depth (SP2, Magnet 1 and Magnet optimized parameters derived based on the measured pore water
2; see Fig. 2); Scenario I(b) – pore water pressure at 3 positions over pressure (scenario Ib). Although the prediction of the maximum
depth (VWP1a-c; see Fig. 2); Scenario I(c) – total settlements at 3 pore water pressure at point VWP1c is improved, the overall pore
positions over depth (SP2, Magnet 1 and Magnet 2) and pore water water pressure dissipation is modeled even less accurately com-
pressure at 3 positions over depth (VWP1a-c). pared to scenario Ia. Moreover, the predicted soil deformations sig-
Scenario II: Optimized input parameters: relevant parameters nificantly differ from the in situ measurement data. Fig. 10(e) and
for the first and third clay sub-layers as well as the silt layer (f) shows the results using the optimized parameters derived based
(OCR; k ; j ; l , and k) were varied independently from each other. on the vertical displacements and the pore water pressure (sce-
output parameters for objective function: layer-wise settle- nario Ic). Here, both the settlements as well as the pore water dis-
ments calculated from 5 measurement points over depth (Points sipation prediction exhibit strong differences to the in situ
1–5; see Fig. 2) for the considered layers (1st and 3rd clay sub- measurements. This may be attributed to the fact that seasonal
layers and silt layer). groundwater fluctuations as reported in [18] or possible displace-
For the objective function in each case and for each measure- ment of the VWP are not modeled in the FEA nor is any large dis-
ment response (three settlements and/or three pore water pres- placement analysis adopted. Since these effects cannot be captured
sure measurements), 6 time points equally distributed over the by the constitutive parameters alone, an inverse analysis using this
logarithmic time variation were chosen. pore pressure measurement is less accurate than an inverse analy-
In scenario I and II, the complete time span of the available sis based on settlements.
measurement data (up to 1000 days) is used for the inverse analy- By comparison of scenarios Ia to Ic, it can be illustrated that the
sis. However, inverse analysis is usually performed not to fit exist- choice of data to form the objective function has a significant influ-
ing data but rather to calibrate a model with data from a limited ence on the inverse analysis. It becomes obvious that an increase in
period of time, where measurement data are available. Then, the the amount of data considered in the evaluation of the objective
calibrated model is used to predict the future behavior of the sys- function does not automatically improve the result of the inverse
tem. Therefore, in scenario III, measurement data from shorter analysis. Since there are more uncertainties involved in the objec-
time spans are used to calibrate the model. Then, the model predic- tive function when incorporating more data, the overall prediction
tion for up to 1000 days is compared to the monitoring data to might also be of less quality compared to optimization based on a
evaluate the influence of the time span on the prediction quality smaller amount of measurement data. According to the above dis-
and to analyze the ability of the model to capture the real response. cussion, inverse analysis based on soil displacements provides a
Scenario III: Optimized input parameters: as defined in sce- good agreement between numerical results and in situ measure-
nario II. ments. Therefore, the optimized parameters obtained via scenario
output parameters for objective function: layer-wise settle- Ia are used for the further inverse analysis study.
ments as defined in scenario II. Scenario III(a) – for 6 time points The parameters found by inverse analysis scenario Ia enable a
in the first 243 days; Scenario III(b) – for 6 time points in the first rather good reproduction of the total settlements and pore water
510 days; dissipation. However, several soil layers are involved in this case
All inverse analyses were performed using the metamodeling study and contribute to the total settlement. The aim of the inverse
approach. analysis is to not only reproduce the final total settlements but also
identify parameter sets characterizing the soil behavior of each
layer adequately. Therefore, the layer-wise displacements are a
good indicator for the validation of the inverse analysis. Fig. 11
4.3. Inverse analysis scenario I – Parameter optimization of three clay shows the total and layer settlements using the initially obtained
sub-layers based on total settlements and pore water pressure parameters for the blind prediction. The layer displacements are
calculated by the difference in the adjacent measurement points
Fig. 10 displays the results of the soil settlements and pore (see Fig. 2). It can be observed that there is a large disagreement
water pressure dissipation using the optimized parameters between the prediction and in situ measurements for all layers.
attained from inverse analysis scenario I. To study the quality of inverse analysis scenario Ia, Fig. 12
Fig. 10(a) and (b) compares the in situ measurements with the shows the same comparison between the total and layer displace-
model responses calculated using the parameter values identified ments calculated with the optimized parameter values. The com-
by inverse analysis based on the measured total displacements parison reveals that the layer settlement prediction for the
(scenario Ia). The numerical simulation of the settlement paths middle clay sub-layer and the sand layer (including a thin part of
agrees with the in situ measurements overall. In particular, the clay sub-layer 3) are improved under inverse analysis scenario Ia.
final settlements fit much better compared to the blind prediction. However, the layer settlement predictions for the silt and the
Moreover, the maximum pore water pressure and dissipation upper and lower clay sub-layers still do not adequately match
behavior at point VWP1b are modeled properly. However, some the in situ measurements. To follow the aim of identifying param-
differences between the prediction results and measured data eter values that reproduce the physical behavior of the soil, more
remain. The consolidation settlements for Magnet 2 start too early, effort needs to be spent to improve the displacement prediction
which indicates a drainage that is too fast. In addition, the maxi- for these layers. Therefore, a second inverse analysis (scenario II)
mum pore water pressures and dissipation behavior at VWP1a is conducted in the following.
and VWP1c require improvement. Here, it is noteworthy that the
maximum pore water pressure is overestimated at VWP1a, 4.4. Inverse analysis scenario II – Parameter optimization for silt layer
whereas it is underestimated at VWP1c. This indicates that a uni- and first and third clay sub-layers based on layer displacements
form change in the parameters for the three clay sub-layers within
the inverse analysis might not be adequate. Therefore, an indepen- Inverse analysis scenario II is performed based on the parame-
dent optimization is needed for these three sub-layers. However, in ter values identified in scenario Ia. The silt layer and the first and
general, it can be stated that using parameter values identified in third clay sub-layer are selected for further parameter identifica-
N. Müthing et al. / Computers and Geotechnics 93 (2018) 87–103 97
Fig. 10. Results of inverse analysis scenario I: (a) vertical displacements using scenario Ia, (b) pore water pressure variation using scenario Ia, (c) vertical displacements using
scenario Ib, (d) pore water pressure variation using scenario Ib, (e) vertical displacements using scenario Ic, (f) pore water pressure variation using scenario Ic.
tion in terms of their relevant parameters (OCR; k ; j ; l , and k). To in the improved modeling of the layer-wise settlements for the silt
form the objective function, the layer-wise settlements calculated and upper clay layers and consequently in the improved total set-
from 5 measurement points over depth (see Fig. 2) for the consid- tlements. This can be quantified according to the minimum value
ered layers are used. of the objective function for the two scenarios: 0.0679 for scenario
Fig. 13 shows the model response compared to in situ measure- Ia and 0.0657 for scenario II. The layer-wise settlements for the
ments from this study. It can be stated that the calibrated numer- lower clay layer are improved regarding its shape with an
ical model is able to well predict the total as well as layer-wise unchanged final value. The remaining differences between the
settlements. Additionally, the pore water pressures displayed in measurements and the model predictions could be eliminated by
Fig. 14 are precisely reproduced in terms of the maximum total further improvement of the constitutive soil model in describing
pore water pressure and dissipation behavior. the anisotropy and destructuration behavior.
Comparing the results of scenarios I and II, a further improve- Table 3 gives the initially chosen parameters (blind prediction)
ment in the prediction can be found. This improvement is observed and the optimized parameters (scenario II). According to the com-
98 N. Müthing et al. / Computers and Geotechnics 93 (2018) 87–103
Fig. 11. Model responses using the initial prediction parameter set; (a) total displacements, (b) layer displacements.
Fig. 12. Model responses of inverse analysis scenario Ia: (a) total displacements, (b) layer displacements.
Fig. 13. Model responses of inverse analysis scenario II: (a) total displacements, (b) layer displacements.
parison between these parameters, the optimized soil properties results in a stiffer compression behavior. The transition from
are physically reasonable. This is also supported by the sortation over- to normally consolidated soil was set too late. The impact
of the parameter magnitude, as the OCR and the permeability of this effect was strengthened by the underestimation of the
become smaller with increasing depth. compression and recompression indices. This particularly holds
Additionally, it can be concluded that all initial OCR values true for j values of the clay layers. Moreover, Fig. 14 shows
for the clay were over-estimated by approx. 10–20%. This may the model validation regarding the pore water pressure
be attributed to the fact that rate effects in the CRS test were generation and dissipation. The accurate permeability values
not considered in the initial determination of the OCR [1]. Con- enable the model to capture the in situ pore water dissipation
sequently, the initial yield stress is over-estimated, which process.
N. Müthing et al. / Computers and Geotechnics 93 (2018) 87–103 99
values are normalized with respect to the value of the best sce-
nario: scenario II. It can be observed that the permeability in the
initial prediction and scenario IIIa significantly differs from the
final value. The same holds true for the OCR value of clay layer 2.
The deviation of these sensitive parameters may explain the higher
error in these two prediction scenarios.
5. Probabilistic evaluation
Table 3
Values of optimized parameters from initial and optimum parameter sets.
Fig. 15. Model responses of inverse analysis scenario IIIa: (a) total displacements, (b) layer displacements.
Fig. 16. Model responses of inverse analysis scenario IIIb: (a) total displacements, (b) layer displacements.
Fig. 18. Histograms of the probability distribution of the settlements at point 0 (a) and the pore water pressure at point VWP1a (b).
that caused by the soil heterogeneity in the field or the variability desirable. Precise measurements derived by design of experiment
of conditions in different cross sections, and the model uncertainty (DOE) methods are essential for an adequate numerical simulation.
itself.
An alternative approach would be a Bayesian updating as 6. Conclusion and outlook
described in [35,34]. This approach would allow for a reduction
in the input uncertainty based on the measured outputs. Since In the framework of this study, a blind prediction was per-
probabilistic evaluation is not the focus of this study, the applica- formed for an embankment constructed on soft, structured clay.
tion of this approach is beyond the scope of this paper. A numerical model was implemented based on assumptions and
Summarizing from a methodical point of view, it can be stated simplifications regarding the geometrical conditions and staged
that only improved knowledge of the parameter value ranges, i.e., a construction of the embankment, the initial and boundary condi-
smaller standard deviation r mathematically speaking, reduces the tions as well as the constitutive modeling. The calibration of the
uncertainty. A shift in the mean value from improved knowledge of numerical model was conducted based on in situ measurement
this characteristic does not decrease the uncertainty of the model data to identify the corresponding model parameters. Different
response. Consequently, improved knowledge of the measurement methods for parameter optimization were investigated. The main
accuracy and that of the homogeneity of the in situ conditions are conclusions can be stated as follows:
102 N. Müthing et al. / Computers and Geotechnics 93 (2018) 87–103
[30] Kennedy J, Eberhart R. Particle swarm optimization. Proceedings of IEEE [33] Ching J, Phoon KK. Constructing multivariate distributions for soil parameters.
international conference on neural networks, vol. 4. p. 1942–8. In: Ching J, Phoon KK, editors. Risk and reliability in geotechnical
[31] Shi Y, Eberhart R. A modified particle swarm optimizer. In: Proceedings of IEEE engineering. Boca Raton: CRC Press; 2014. p. 3–77. chap. 1.
congress on evolutionary computation. p. 69–73. [34] Miro S, König M, Hartmann D, Schanz T. A probabilistic analysis of subsoil
[32] Phoon KK, Kulhawy F. Characterization of geotechnical variability. Can parameters uncertainty impacts on tunnel-induced ground movements with a
Geotech J 1999;36:612–24. back-analysis study. Comput Geotech 2015;68:38–53.
[35] Kelly R, Huang J. Bayesian updating for one-dimensional consolidation
measurements. Can Geotech J 2015;52(9):1318–30.