SHOTCRETE MODEL

Internal report: Implementation, validation and

application of the shotcrete model

Client:
Plaxis bv
P.O. Box 572
NL - 2600 AN Delft

Dr.techn. Bert Schädlich

Ao. Univ.-Prof. Helmut F. Schweiger
Computational Geotechnics Group
Institute for Soil Mechanics and Foundation Engineering
Graz University of Technology

February 2014 / CGG_PR027_2012_4

Internal report: Shotcrete model 1

Table of contents
1 Introduction .................................................................................................... 3
2 Motivation ....................................................................................................... 4
3 Constitutive model ......................................................................................... 5
3.1 Model input parameters ............................................................................ 5
3.2 Model structure ......................................................................................... 6
3.3 Yield surfaces ............................................................................................ 6
3.4 Strain hardening and softening ................................................................. 7
3.5 Time dependent material parameters ...................................................... 11
3.5.1 Elastic stiffness ................................................................................ 11
3.5.2 Compressive and tensile strength..................................................... 11
3.5.3 Plastic deformability ........................................................................ 13
3.5.4 Fracture energy ................................................................................ 14
3.6 Creep ....................................................................................................... 15
3.7 Shrinkage ................................................................................................ 16
3.8 Safety factors........................................................................................... 16
3.9 State variables ......................................................................................... 17
4 Parameter calibration .................................................................................. 18
4.1 Time dependent strength and stiffness .................................................... 18
4.2 Fracture energy and tensile strength ....................................................... 19
4.3 Shrinkage ................................................................................................ 22
4.4 Creep ....................................................................................................... 23
4.5 Recommended parameters ...................................................................... 25
5 Application - Nivy station ............................................................................ 26
5.1 Project description................................................................................... 26
5.2 Numerical model and material parameters ............................................. 26
5.3 Calculation phases................................................................................... 28
5.4 Results ..................................................................................................... 29
5.5 Conclusions ............................................................................................. 33
6 Open issues.................................................................................................... 34
6.1 FE-convergence and structural snap back............................................... 34
6.2 Crack alignment with mesh orientation .................................................. 37
Internal report: Shotcrete model 2

7 Summary ....................................................................................................... 39
8 References ..................................................................................................... 40
Internal report: Shotcrete model 3

1 Introduction
This report summarizes the work on the shotcrete constitutive model developed
and implemented by Bert Schädlich for Plaxis b.v. in 2012-2014. The report is
based on two conference papers (Schädlich & Schweiger 2014, Schädlich et al.
2014), which have been extended by additional details on the implementation of
the model.

The constitutive model can account for time dependent strength and stiffness,
strain hardening/softening in tension and compression, creep and shrinkage. Parts
of the model are based on previous work by Schütz et al. (2011) and Meschke et
al. (1996). Primary objective of the research project was the modelling of
shotcrete behaviour for tunnelling applications, but the constitutive model can
also be used for cast concrete, jet grout and other cement-based materials.

The main part of this report deals with the formulation of the shotcrete
constitutive model. The model features and their implementation are explained,
and details of the numerical algorithm for calculating plastic strains are
presented. Calibration of model parameters is demonstrated by back-analysing
experimental results of shotcrete and concrete.

The influence of the various model features was investigated in a NATM

tunnelling example with double side drift excavation. The geometry and geology
of the example was taken from the Nivy station project in Bratislava, but has
been slightly modified for the purpose of this study.

The report concludes with discussing some issues which could not be solved
within the stress point algorithm.
Internal report: Shotcrete model 4

2 Motivation
Shotcrete is a major support element in NATM, drill & blast and conventional
tunnelling, in which the rock mass is supported by a primary shotcrete lining
directly after excavation. Because shotcrete linings are loaded at a very early age,
the influence of time dependent material properties on the deformation behaviour
and bearing capacity is much more significant than in cast concrete structures.
Notably, shotcrete strength and stiffness increase rapidly within the first few
hours after application, while ductility and creep effects decrease. Shotcrete also
exhibits plastic material behaviour before reaching the maximum strength, and
material strength reduces after the maximum strength has been mobilised.

The current engineering approach to model shotcrete linings in numerical

simulations assumes a linear elastic material with a stepwise increase of the
(artificially low, i.e. ‘hypthetical’) Young’s modulus in subsequent excavation
stages. While realistic lining deformations may be obtained with this method,
lining stresses are usually too high, in particular if the lining is subjected to
significant bending.

With the new constitutive model more realistic stress distributions can be
obtained, as the non-linearity of the material behaviour is taken into account.
Furthermore, the stability of the tunnel can be checked at all intermediate stages
without the need for additional capacity checks of the lining cross section.
Internal report: Shotcrete model 5

3 Constitutive model
3.1 Model input parameters
Table 1: Input parameters of the shotcrete model
No. parameter description unit
1 E28 Young’s modulus of cured shotcrete at thydr stress
2  Poisson’s ratio --
3 fc,28 Uniaxial compressive strength of cured stress
shotcrete at thydr
4 ft,28 Uniaxial tensile strength of cured shotcrete stress
at thydr
5  Dilatancy angle °
6 E1/E28 Time dependency of elastic stiffness --
7 fc,1/ fc,28 Time dependency of strength --
8 fc0n Normalized initially mobilised strength --
9 fcfn Normalized failure strength (compression) --
10 fcun Normalized residual strength --
(compression)
11-13 cp p
Uniaxial plastic failure strain at 1h, 8h, --
24h
14 Gc,28 Compressive fracture energy of cured force/length
shotcrete at thydr
15 ftun Ratio of residual vs. peak tensile strength --
16 Gt,28 Tensile fracture energy of cured shotcrete force/length
at thydr
17 Leq Equivalent length (if no regularization is length
used)
18 a Increase of cp with increase of p’ length
19 max Maximum friction angle °
20  cr
Ratio between creep and elastic strains --
cr
21 t50 Time for 50% of creep strains time
22 ∞ shr
Final shrinkage strain --
shr
23 t50 Time for 50% of shrinkage strains time
24 fc Safety factor for compressive strength --
25 ft Safety factor for tensile strength --
26 thydr Time for full hydration (usually 28 days) time
Internal report: Shotcrete model 6

3.2 Model structure

The model is formulated in the framework of strain hardening/softening
elastoplasticity. The total strain is decomposed into elastic strains e, plastic
strains p, creep strains cr and shrinkage strains shr.

ε  ε e  ε p  ε cr  ε shr (1)

Details on how these strain components are derived are given in subsequent
chapters. A compression negative notation is employed throughout this report.

3.3 Yield surfaces

Plastic strains are calculated according to strain hardening/softening
elastoplasticity. The model employs a Mohr-Coulomb yield surface Fc for
deviatoric loading, which is combined with a Rankine yield surface Ft in the
tensile regime (Fig. 1). The yield functions can be formulated in terms of uniaxial
compressive and tensile yield stress, fcy and ft, as

1   3  1   3  2   rot f cy
Fc    (2)
2 2 2   rot  f cy

Ft   1  f t (3)

where 1 and 3 are the major and minor principal stresses and rot is the
intersection of the Mohr-Coulomb failure envelope and the isotropic axis. For a
given maximum inclination max of the Mohr-Coulomb envelope, rot can be
written as

fc  1 
 rot     1 . (4)
2  sin  max  

Fig. 1: Yield surfaces and failure envelope

Internal report: Shotcrete model 7

Comparison of the Mohr-Coulomb envelope with max = 37° with experimental

data on concrete strength in triaxial conditions shows good agreement up to a
stress level of ≈ -10·fc (Fig. 2). In biaxial stress conditions (3 = 0), the
experimental data are matched well on the tension side with ft = 0.1·fc, but the
material strength in compression is underestimated by ~16%. From a practical
point of view, this is still a sufficient match, as shotcrete lining behaviour is
dominated by tensile strength. The strength prediction in biaxial compression
could be improved by using more advanced failure criteria, which employ a
continuous function covering the compression and tension side. Using such
failure criteria, however, complicates the separation of compression and tension
softening, as strength reduction in tension inevitably influences the material
strength in compression and vice versa. For this reason this approach has not
been pursued further.

Fig. 2: Mohr-Coulomb failure envelope vs. experimental data in biaxial and

triaxial conditions

3.4 Strain hardening and softening

Compression
Behaviour in compression follows an approach proposed by Schütz et al. (2011).
The stress-strain curve is divided in four parts (Fig. 3): Part I - quadratic strain
hardening, part II - linear strain softening, part III - linear strain softening and
part IV - constant residual strength. Due to the time dependency of the involved
material parameters, a normalised hardening/softening parameter Hc = 3p / cpp is
used, with 3p = minor principal plastic strain (calculated from Fc) and cpp =
plastic peak strain in uniaxial compression.
Internal report: Shotcrete model 8

Fig. 3: Normalized stress - strain curve in compression

In part I, the uniaxial yield stress fcy is mobilised with Hc according to a quadratic
function:

 
f cy,I  f c  f c0n  1  f c0n   2H c  H c2  (5)

with fc0n = initial ratio of fcy / fc. During mobilization the yield surface Fc rotates
about the anchor point rot on the isotropic axis. Full mobilization of fc coincides
with Hc = 1, after which linear softening takes place, until the failure strength fcf
= fcfn·fc is reached at Hcf = cfp / cpp. Strength reduction is assumed to be caused
by the destruction of inter-particle bonds, and consequently softening is modelled
by a parallel shift of the fully mobilised Mohr-Coulomb failure envelope
(cohesion softening).

  

f cy, II  f c  1  f cfn  1

  HH c 11   (6)
  cf 

cfp is derived from the fracture energy in compression, Gc, and the characteristic
length of the finite element, Leq, which provides the necessary regularization to
avoid mesh dependent numerical results.

2  Gc
 cfp   cp
p

 
1  f cfn  f c  Leq
(7)

Leq is calculated from the size of the finite element, Ael, and the number of stress
points per element, nGP (Pölling 2000).

Ael
Leq  2 (8)
3  nGP
Internal report: Shotcrete model 9

The linear strain softening in part III is governed by the condition that the energy
in elastic unloading must not be greater than the plastic strain energy absorbed by
the crack (no snap-back of stress-strain curve on stress point level). That delivers
the plastic ultimate strain cup as


2 f c  f cfn  f cun 
 cu
p
  cfp  (9)
E

with fcun = residual strength level = fcu / fc and E = elastic Young’s modulus. The
yield stress fcy follows as

  


f cy, III  f c   f cfn  f cun  f cfn   HH c HHcf 

(10)
  cu cf 

where Hcu = cup / cpp. No further softening occurs in part IV of the stress strain
curve, which yields

f cy,IV  f c  f cun . (11)

To account for the increasing ductility with increasing confining pressure, the
total peak strain cp = cpp + cpe increases with 1, governed by the input
parameter a.

 1 
 cp   cp,UC  1  a   (12)
  f c 

Please note that in the notation of this paper, 1 is the confining pressure in a
triaxial compression test. a = 1 and 1 = fc yield a 100% increase of total peak
strain cp compared to the uniaxial compression test. Internally, the increase of cp
is translated into an increase of cpp, which is assumed to be governed by the
mean stress p = (1 + 2 + 3) / 3 according to

 p  fc / 3 
 cp
p
  cp
p

,UC  1  b 
 (13)
  f c 

fc  p f 
2 sin  max    a1  sin  max   cp  c
b
E  E
(14)
 cp  1  sin  max  / 3
p

The good agreement of model predictions based on Eq. 12 and 13 with

experimental data of cured concrete in triaxial compression (Candappa et al.
2001) is shown in Fig. 4 (a = 19, fc = 40 MPa, E = 33.3 GPa, cpp = -1.2‰, max =
43°).
Internal report: Shotcrete model 10

Fig. 4: Increase of peak strain cp with confining pressure 1 (experimental data
from Candappa et al. 2001)

Tension
The model behaviour in tension is linear elastic until the tensile strength ft is
reached (Fig. 5). Linear strain softening follows, governed by the normalized
tension softening parameter Ht = 1p / tup with 1p = major principal plastic strain
(calculated from Ft) and tup = plastic ultimate strain in uniaxial tension.

f ty  f t  1   f tun  1  H t  (15)

Similar to softening in compression, tup is derived from the fracture energy in

tension, Gt.

2  Gt
 tup  (16)
1  f tun   f t  Leq

Once the residual strength ftu = ftun·ft is reached, no further softening takes place.

Fig. 5: Tension softening

Internal report: Shotcrete model 11

3.5 Time dependent material parameters

3.5.1 Elastic stiffness
The stiffness and strength of shotcrete increases rapidly with time due to the
hydration of the cement paste. The increase of Young’s modulus E in the model
follows the recommendation of CEB-FIP model code (1990):

E t   E28  e

sstiff  1 th ydr / t  (17)

where E28 is the Young’s modulus of cured shotcrete, thydr is the time until full
curing (usually taken as 28 days), t is the time in days and sstiff is the parameter
governing stiffness evolution with time. sstiff can be related to the stiffness ratio at
1 day and thydr, E1/E28, as

ln E1 / E 28 
s stiff   (18)
t h ydr / 1d  1

The Young’s modulus is constant for t < 1h and for t > thydr. Fig. 6 compares the
increase of E with time for different values of E1/E28 with the experimental data
compiled by Chang (1994).

Fig. 6: Increase of Young’s modulus with time

3.5.2 Compressive and tensile strength

A similar approach as for E is followed for the evolution of uniaxial compressive
strength fc with time. sstrength is the parameter governing the development of
Internal report: Shotcrete model 12

strength with time and can be related to the ratio of fc at 1d and thyd. A lower limit
of fc = 0.005·fc,28 is used at very early age.
f c t   f c,28  e

sstrength 1 th ydr / t  (19)

ln  f c,1 / f c,28 
s strength   (20)
t h ydr / 1d  1

In triaxial stress space, the increase of fc corresponds to a vertical shift of the

Mohr-Coulomb envelope with time, while its inclination remains constant. The
ratio of ft / fc and the values of fcfn, fcun and ftun are assumed to be constant in
curing.

Alternatively, strength evolution can be modelled according to the early strength

classes J1, J2 and J3 of EN 14487-1 (2006), which defines ranges of shotcrete
strength at different shotcrete ages up to 24h. Mean values of these ranges have
been assumed for classes J1 and J2 in the model, with class J3 lying 50% above
the boundary between classes J2 and J3 (Table 2). Between 24h and thyd, an
approach proposed by Oluokun et al. (1991) is adopted (t in days):

thydr t
f c t   f c,28   f c,1 / f c,28 thydr 1d t (21)

Fig. 7 compares the evolution of fc according to the early strength classes with
the CEB-FIP model code formulation for fc,28 = 25 MPa and thydr = 28d. It is
obvious, that the CEB-FIP equation yields very low shotcrete strength at ages <
2h.

Fig. 7: Increase of shotcrete strength with time

Internal report: Shotcrete model 13

Table 2: Mean uniaxial compressive strength of early strength classes [MPa]

time [hr] J1 J2 J3
< 0.1 0.15 0.35 0.75
0.5 0.23 0.715 1.65
12 2.0 5.5 12.0
24 3.5 12.0 28.5

3.5.3 Plastic deformability

The ability of young shotcrete to withstand large deformations is not only a result
of its low elastic modulus at this age, but also due to its high plastic ductility.
With shotcrete aging this ductile behaviour decreases. In the shotcrete model, this
behaviour is represented by a time dependent plastic peak strain cpp. Similar to
the approach proposed by Meschke (1996), a tri-linear function in time is
adopted. Input values are the plastic peak strains at t = 1h, 8h and 24h. Beyond
24h, cpp is assumed to be constant.

Fig. 8 compares the adopted function with cpp = -3%, -0.5% and -0.2% at 1h, 8h
and 24h with experimental data from uniaxial compression tests. The elastic part
of the total peak strain has been subtracted from the test results. With the
exception of the tests by Sezaki et al. (1989), there is a trend of cpp decreasing
until t ≈ 24h and staying relatively constant afterwards. Sezaki’s tests, however,
also yielded a very low Young’s modulus of E28 ≈ 5 GPa, which indicates that
the tested shotcrete was exceptionally soft.

Fig. 8: Reduction of cpp with shotcrete aging

Internal report: Shotcrete model 14

3.5.4 Fracture energy

The change of fracture energy with time is an outcome of the model, not an
explicit input. The tensile failure strain tup is derived from the tensile fracture
energy and tensile strength of the cured concrete, regardless of the current
shotcrete age. As the tensile strength ft increases with time, but tup remains
constant, the current fracture Gt increases proportionally with the increase of ft.
Compared with experimental data, this linear increase of Gt with ft is slightly
conservative, but still a good representation of real concrete behaviour (Fig. 9).

Fig. 9: Predicted stress-strain curves in uniaxial tension at different shotcrete

ages and increase of tensile fracture energy with tensile strength

The plastic failure strain in compression, cfp, is coupled to the plastic peak strain
cpp such that the ratio cfp/cpp remains constant. As cpp decreases with time
(3.5.3), also cfp and hence Gc reduce. On the other hand, the compressive
strength fc increases with time, which results in higher values of Gc. The
influence of these counteracting trends on the fracture energy Gc can be
evaluated by analysing predicted stress-strain curves at different ages. Fig. 10
shows simulated stress-strain curves in uniaxial compression at different
shotcrete ages with cpp = -20/-1.5/-1.0‰ (at t = 1h, 8h, and 24h), Gc,28 =
30 kN/m, E1/E28 = 0.5 and fc,1/fc,28 = 0.4. Evaluating the predicted fracture energy
yields rather high values at early age, a sharp drop at ~12h and a linear increase
of Gc with fc afterwards.. After 24h, Gc increases due to the increase in strength,
while cpp remains constant. While there are no experimental data to compare
with, the high fracture energy at very early age is a consequence of the desired
ductile behaviour at this stage. As very young shotcrete effectively does not fail
at all in compression, the fracture energy theoretically should be infinite.
Internal report: Shotcrete model 15

Fig. 10: Stress-strain curves in uniaxial compression at different shotcrete ages

and development of compressive fracture energy with compressive
strength

3.6 Creep
Creep is modelled with a viscoelastic approach. Creep strains cr increase linearly
with stress  and are related to elastic strains via the creep factor cr.

 cr  σ t  t 0
ε cr t    (22)
D t  t 50
cr

The evolution of creep with time t is governed by the start of loading at time t0
and the parameter t50cr. For instantaneous loading (t0 = 0), t50cr equals the time
until 50% of the creep strains have evolved. For shotcrete utilization higher than
45% of fc, non-linear creep effects are accounted for by replacing cr with the
following equation from EC 2 (2004):

kcr   cr  e1.5k 0.45 (23)

with k = c / fcm being the degree of shotcrete utilization in compression. Eq. 22

can be transformed into an incremental formulation, which is independent of the
starting time t0:

  εn
ε cr cr
n1  ε n 
ε cr  t
cr
(24)
t n  t 50
cr

with n+1cr … creep strain at the end of step n, ∞cr = cr· / D … final creep strain
for a given stress , ncr … creep strain at the beginning of step n, and t …
current time step. Due to the time dependency of the linear elastic stiffness
matrix D, the creep history is stored as normalised values of cr · E(t). The creep
history is adjusted for the stress state at first activation of the shotcrete cluster,
Internal report: Shotcrete model 16

such that no creep strains are produced by initial stresses. It should be noted,
however, that currently (Plaxis 2D 2012.01) state variables are taken over if
the previous material was also defined with the shotcrete model, in which
case also creep will continue. If a reset of state variables is desired, a nil step
with a different material (e.g. linear elastic) is required.

3.7 Shrinkage
‘Shrinkage’ in the context of this model refers to isotropic loss of volume with
time, which is independent of the stress state. Shrinkage strains shr are calculated
according to the recommendation of ACI 209-R92 (1992) as

 shr t    shr 
t
(25)
t  t 50
shr

with ∞shr being the final axial shrinkage strain and t50shr the time when 50% of
shrinkage has occurred.

3.8 Safety factors

To facilitate calculations based on design values of shotcrete strength, the
possibility to use separate safety factors fc and ft for compressive and tensile
strength, respectively, has been included in the model. In theory, design
calculations could also be carried out by using lower input values of fc,28 and ft,28
in the first place. However, this is not advisable:
1. The early age shotcrete classes J1, J2 and J3 specify shotcrete strength
values at 0.5h, 12h and 24h, which are not related to fc,28. Using a lower
value of fc,28 therefore does not affect the shotcrete strength at early age, if
the early strength classes are used.
2. If lower shotcrete strength is used due to safety considerations, also the
fracture energy should be reduced to obtain similar stress-strain curves.
3. The creep factor increases for shotcrete utilisation >0.45·fc. Using design
values for fc would therefore overestimate creep effects.

For these reasons input safety factors fc and ft have been introduced. The
characteristic, time dependent compressive and tensile strengths fc and ft as well
as the corresponding fracture energies Gc. and Gt are divided by fc and ft in each
calculation step. Shotcrete utilization (StVar(15) and StVar(16)) always refers to
the characteristic values of fc and ft.
Internal report: Shotcrete model 17

3.9 State variables

State variables are internal variables used to monitor the stress-strain behaviour,
time dependent parameter evolution and creep history of each stress point. They
also provide useful information about the current utilization of the material. The
state variables are initialized at first activation of the shotcrete cluster. During
initialization, hardening parameters Hc and Ht are adjusted such that the initial
stress state is on the mobilized yield surface. As the initial material strength is
used, it is not advisable to use time dependent strength in wished-in-place
calculations (replacement of one material set with the shotcrete model without
modelling the excavation). In that case the low initial strength may result in
unrealistic mobilization of the compression yield surface due to initial stresses.

Table 3: State variables of the shotcrete model

StVar() parameter description unit
1 123 Initialization parameter ---
2 t Time at first activation time
3 E(t) Average Young’s modulus in current step stress
4 fc(t) Uniaxial compressive strength at the end stress
of the current step
5 Hc Normalized compression hardening / --
softening parameter (0-1: hardening; > 1:
softening)
6 Ht Normalized tension softening parameter --
(0: no softening; 0-1: softening; >1:
residual)
7 Leq Equivalent length length
8 fcy Current compressive yield stress stress
9-14 cr·E(t) Normalized creep history --
15 Futil,fc Shotcrete utilization in compression --
16 Futil,ft Shotcrete utilization in tension --
Internal report: Shotcrete model 18

4 Parameter calibration
With the exception of max and a, all model parameters can be derived from
uniaxial tension and compression tests at different shotcrete ages (+ one creep
test). Typically, not all of these tests will be available, in which case correlations
and published experimental data should be used (some of them already presented
in previous sections of this paper). Determination of max and a requires triaxial
tests on cured concrete. The impact of these parameters in typical tunnelling
calculations, however, is small, as at least one of the major principal stresses in
the lining is close to 0.

4.1 Time dependent strength and stiffness

Load controlled uniaxial compression tests at different shotcrete ages are
sufficient to determine the parameters fc,28, E28, fc,1/fc,28, E1/E28 and cpp (at 1h, 8h
and 24h). Determination of fcfn, fcun and Gc,28 requires a displacement controlled
uniaxial compression test on cured concrete. Fig. 11 shows the stress strain
curves of a series of such tests (Golser et al. 1991), carried out at shotcrete ages
of 8h, 24h, 72h and 168h. Model predictions are obtained on a single stresspoint
with the parameters listed in Table 4. The equivalent length has been assumed as
Leq = 0.15 m, and fc0n was taken as 0.15. Post-peak behaviour cannot be
compared with the experimental data, as this requires a complete FE-model of
the boundary value problem. It is obvious, however, that the increase of peak
strength and stiffness with time as well as the peak strains are well reproduced by
the model.

Table 4: Model input parameters

parameter value unit
E28 28.0 GPa
 0.2 --
fc,28 22.0 MPa
 0 °
E1/E28 0.6 --
fc,1/ fc,28 0.43 --
fc0n 0.15 --
fcfn 0.2 --
fcun 0.2
cpp -30 / -1.5 / -0.7 ‰
Gc,28 30.0 kN/m
max 37.0 °
thydr 28.0 d
Internal report: Shotcrete model 19

Fig. 11: Uniaxial stress strain curves at different shotcrete ages (experimental
data from Golser et al. 1991)

4.2 Fracture energy and tensile strength

The tensile strength ft,28 and the fracture energy Gt,28 can in principle be derived
from uniaxial tension test results. Due to the experimental difficulties involved in
these tests, however, indirect tests like the 4-point bend beam (EN 14488-3) test
are more common and currently the standard procedure to obtain ft,28 and Gt,28.
As a drawback of indirect testing, material parameters can only be derived by
semi-analytical correlations or inverse analysis.

Here, test results presented by Barros & Figueiras (1999) are back-analysed with
the shotcrete model. They carried out 3-point-bending tests on notched beams of
steel fibre reinforced concrete with varying fibre content and proposed a
correlation between fracture energy and fibre content Wf.

Beam dimensions and the FE-model used in the simulation are shown in Fig. 12.
Stiff plates with hinges are used to model bearings without rotational restraint.
No time dependency of material parameters is considered.

The back analysis focusses on the test series s4 with fibre contents of 30 and 60
kg/m3. Steel fibres Dramix TX60/.80 were used. Uniaxial compression tests
yielded fc ≈ 35 MPa, with slightly higher strength for the lower fibre content.
Young’s modulus and peak strain were reported as E = 30 GPa and cp = -2.2‰,
which yieldscpp = cp + fc / E = -1.0‰. The fracture energy in compression can
be estimated from the given stress-strain curve as Gc = 101 kN/m and 143 kN/m
for 30 and 60 kg/m3 of steel fibres, respectively (Gc evaluated until 0.1·fc).
Softening in compression does not play a significant role in the beam test
Internal report: Shotcrete model 20

simulation, and hence a constant value of Gc = 100 kN/m is adopted. Input

parameters common for all beam test simulations are summarised in Table 5.

Fig. 12: Beam dimensions [mm] and FE-model

Table 5: Model input parameters for steel fibre contents of 0 / 30 / 60 kg/m3

parameter value unit
E28 30.0 GPa
 0.2 --
fc,28 35.0 MPa
ft,28 3.0 / 1.5 / 3.0 MPa
 0 °
fc0n 0.15 --
cpp -1.0 ‰
Gc,28 100.0 kN/m
ftun 0.0 --
Gt,28 0.1 / 2.5 / 6.9 kN/m
max 37.0 °

Fig. 13 shows the load-displacement-curves of the beam test simulation in

comparison with the experimental data (Fy … vertical force, uy … vertical
displacement at mid span). Good match is obtained with Gt = 2.5 kN/m, ft =
1.5 MPa for 30 kg/m3 and Gt = 6.9 kN/m, ft = 3.0 MPa for 60 kg/m3 of steel fibre
content. The test data for plain concrete are matched well with Gt = 0.1 kN/m and
ft = 3.0 MPa. Strain hardening in compression does not contribute significantly to
the overall fracture energy, which is dominated by the behaviour in tension.

The Gt-values obtained in the back analysis compare very well with the
correlation proposed by Barros & Figueiras (1999)

 
Gt  1  13.159  W f 1.827  Gt 0 (26)

where Gt0 is the fracture energy of plain concrete (~0.1 kN/m) and Wf is the fibre
percentage in weight (1.25% for 30 kg/m3 and 2.5% for 60 kg/m3).
Internal report: Shotcrete model 21

It can be seen in Fig. 13 that for a fibre content of 30 kg/m3, the experimental
load-displacement curve initially follows the plain concrete, but stabilizes after a
sharp initial drop. This initial peak cannot be captured by the shotcrete model,
which employs linear post-peak softening. The practical relevance of the initial
peak, however, is small, as its contribution to the overall fracture energy is
negligible. Varying the mesh coarseness does not change load-displacement
curves significantly (Fig. 15, Fig. 14).

Fig. 13: Load-displacement curves of bend beam test for different steel fibre
content

Fig. 14: Load-displacement curves of bend beam simulations with different FE-
meshes
Internal report: Shotcrete model 22

Fig. 15: FE-meshes for bend beam test simulation

4.3 Shrinkage
Shrinkage of concrete and other cement-based materials is considered as a load-
independent loss of volume due to drying and hydration. There is a strong
influence of environmental conditions and water-cement-ratio, such that low air
humidity and high water-cement ratios amplify concrete shrinkage. Due to less
aggregate content and higher water-cement ratio, shrinkage of shotcrete is more
pronounced than for conventional cast concrete (Austin & Robins, 1995).
Eurocode 2 recommends final shrinkage strains for cast concrete of -0.2‰ to -
0.6‰ depending on air humidity, concrete class and the effective size of the
structural element.

Fig. 16 shows a compilation of shotcrete shrinkage data in comparison with the

approach employed in the shotcrete model (after Schütz et al. 2011). Typical
final shrinkage strains are about -1.0 … -1.5‰. Shotcrete additives can
significantly increase shrinkage effects, as demonstrated in Fig. 17 for polymer
modified shotcrete (Galler et al. 2009).
Internal report: Shotcrete model 23

Fig. 16: Shrinkage test results on regular shotcrete and model prediction (after
Schütz et al. 2011)

∞shr = -2.5‰
t50shr = 45d

Fig. 17: Shrinkage test results on polymer modified shotcrete and model
predictions

4.4 Creep
Creep properties of shotcrete are usually derived from uniaxial multistage creep
tests, in which the load level is increased in a stepwise manner and held constant
for a certain time. In the creep test shown in Fig. 18, loading started at t = 8h with
3 = -2.5 MPa, with subsequent load steps by -2.5 MPa at 24h, 36h and 48h
(Aldrian 1991). The sample was unloaded at t = 240h.

Deriving creep properties from such a test requires additional information about
strength, stiffness and ductility development with time or - in the absence of such
tests - a reasonable estimate of these parameters, as plastic and elastic strains
Internal report: Shotcrete model 24

make up a large portion of the measured total deformation. The model

parameters listed in Table 6 are very similar to those calibrated from the uniaxial
compression tests in 4.1, with some minor adaptation due to values given in the
original publication of Aldrian (1991). The shotcrete model slightly
overestimates creep in unloading due to its viscoelastic formulation, which is
however of limited practical relevance.

Table 6: Model input parameters for creep test simulation

parameter value unit
E28 28.0 GPa
 0.2 --
fc,28 25.0 MPa
E1/E28 0.65 --
fc,1/ fc,28 0.43 --
fc0n 0.15 --
cpp -30 / -0.7 / -0.7 ‰
max 37.0 °
cr 2.6 --
t50cr 1.5 d
∞shr -0.5 ‰
t50shr 28.0 d
thydr 28.0 d

Fig. 18: Creep test simulation (experimental data from Aldrian 1991)
Internal report: Shotcrete model 25

4.5 Recommended parameters

The following parameter recommendations are based on back analysis of
published experimental data of shotcrete and concrete. The ‘default’ values are
used in the constitutive model if the user input value for this parameter is 0. It
should be noted that it is the responsibility of the user to check the validity of
these material parameters in a specific project. General project settings must be
[kN] for forces, [m] for length and [d] for time.

Table 7: Recommended model parameters for shotcrete / concrete

parameter recommended values default
E28 25 … 30 GPa --
 0.15 … 0.25 --
fc,28 depending on strength class --
ft,28 0.05 … 0.1 · fc,28 --
 0 … 10° --
E1/E28 1 … no time dependency
1.0
0.5 … 0.7 for shotcrete
fc,1/ fc,28 1 … no time dependency
0.2 … 0.3 for cast concrete 1.0
for shotcrete classes: -1 (J1), -2 (J2), -3 (J3)
fc0n 0.1 … 0.25 --
fcfn 0.1 (1 for no softening) 0.05
fcun 0.1 (1 for no softening) 0.05
cpp at 1h: -0.01 … -0.03
at 8h: -0.001 … -0.0015 --
after 24h: -0.0007 … -0.0012
Gc,28 30 … 70 kN/m --
ftun 0.0 (1 for no softening) --
Gt,28 0.05 … 0.15 kN/m for plain shotcrete,
--
for SFRC see 4.2
Leq 0 in FE-calculations (determined automatically),
--
in Soiltest > 0 based on average element size
a 16 … 20 --
max 35 … 43° depending on stress level 37°
cr 2.0 … 3.0 for tunnel linings, for cast concrete
--
see Eurocode 2
t50cr 1.0d … 5d --
∞
shr
-0.0005 … -0.0015 --
t50shr 28d … 100d --
fc depending on design standard 1.0
ft depending on design standard 1.0
thydr 28d 28d
Internal report: Shotcrete model 26

5 Application - Nivy station

5.1 Project description
The Nivy railway station project in Bratislava (Marcher et al. 2012) is analysed
here but has been slightly modified for the purpose of this study. The maximum
overburden is approx. 24 m above the tunnel crown. The station consists of two
platform tunnels with approx. 120 m² cross section each. The axial distance
between the platform tunnels is approx. 30.5 m. The length of the platform
(station) tunnels is 150 m. For simplicity for this study the analysis has been
performed with only one of the two platform tunnels under drained conditions,
the water table being approximately 5.5 m below the surface.

The geological environment consists of Quaternary capping sediments and

Neogene strata of varying thickness. Sand sediments consist of fine-grained
additions to loamy sands, which are altered by cohesive soils of clayey character
with medium plasticity and sandy clays of stiff consistency. At the depth of the
Station Nivy mostly sand fractions are present, while the upper parts of the
stratum exhibits cohesive soils of clayey character with lower and medium
plasticity and sandy clays of stiff consistency.

5.2 Numerical model and material parameters

Computations have been performed with 2D plane-strain finite element models
using PLAXIS 2D 2012 (Brinkgreve et al. 2012). The geometric layout, the finite
element mesh (using 15-noded triangles), the soil profile and excavation
sequence follow from Fig. 19 and Fig. 20.

Fig. 19: Finite element model (dimensions in m)

Internal report: Shotcrete model 27

Fig. 20: Close-up of FE-model (dimensions in m)

A simple Mohr-Coulomb failure criterion (parameters in Table 8) for the soil is

employed here because the emphasis of this study is on the constitutive model for
the shotcrete. Shotcrete parameters are given in Table 9. Tensile strength
parameters ft,28 and Gt,28 represent a steel fibre content of 60 kg/m3. The
remaining parameters have been calibrated with experimental data from uniaxial
compression tests at different shotcrete ages and multistage creep tests (see
previous sections).

Table 8: Soil parameters

parameter Q1 Q5 N1 N5 unit
E 9 500 140 000 52 000 85 000 kN/m2
 0.3 0.3 0.3 0.3 --
’ 20 34 25.5 27 °
c’ 16 0 21.0 5 kN/m2
 0 0 0 0 °
K0’ 0.66 0.44 0.57 0.55 --
 20 20 20 20 kN/m3

Table 9: Shotcrete parameters

parameter value unit
E28 30·106 kN/m2
 0.2 --
fc,28 35 000 kN/m2
ft,28 3 000 kN/m2
 0 °
Internal report: Shotcrete model 28

parameter value unit

E1/E28 0.6 --
fc,1/ fc,28 class J2 --
fc0n 0.15 --
fcfn 0.10 --
fcun 0.10
ftun 0.0
cpp -30 / -1.0 / -1.0 ‰
Gc,28 100.0 kN/m
Gt,28 6.9 kN/m
cr 2.6 --
t50cr 1.5 d
∞shr -0.5 ‰
t50shr 28 d
max 37.0 °
thydr 28 d

5.3 Calculation phases

Excavation of each section is modelled by a stress release phase (to account for
3D-effects in the 2D simulation), followed by activation of the shotcrete lining.
The stress release factors listed below were based on experience from projects
under similar conditions. The time assigned to each phase represents the tunnel
advance rate.

1. Initial stress state K0 = 1 - sin'

2. Dewatering around tunnel excavation
3. Stress release of th right – 15% (1 day)
4. Excavation of th right – 50% (1 day)
5. Stress release of bench right – 15% (1 day)
6. Excavation of bench right part – 100% (1 day)
7. Stress release of th left – 15% (5 days)
8. Excavation of th left – 50% (1 day)
9. Stress release of bench left– 15% (1 day)
10. Excavation of bench left – 100% (1 day)
11. Stress release of th centre – 30% - (5 days)
12. Excavation of th centre – 70% (1 day)
13. Stress release of bench centre – 15% (1 day)
14. Excavation of bench centre – 90% (1 day)
15. Deactivate side drift walls – 100% (1 day)
16. Standstill (100 days)
17. Reset groundwater (300 days)
Internal report: Shotcrete model 29

5.4 Results
Not all aspects of the shotcrete behaviour are equally important in a given
project. Therefore the model features have been activated separately and their
influence is compared with regard to a reference elastic calculation, in which a
stepwise increase of E from 5 to 15 GPa in subsequent excavation phases is
assumed, which is common practice at present.

Fig. 21 compares the vertical lining displacements along the circumference after
phase 15 (end of excavation) for 3 elastic cases. The largest deformations occur
at the tunnel crown (L = 19 m). Less surprisingly, assuming a constant E =
30 GPa yields the smallest deformation, but also the model with time dependent
stiffness is stiffer than the approach with E = 5/15 GPa.

Fig. 21: Vertical lining deformation after phase 15 - elastic lining

Slightly larger deformations are obtained with the shotcrete model (Fig. 22).
Softening in compression and tension, however, does not have a significant
influence on lining deformations. Considering creep and shrinkage increases
crown displacements by ~5 mm, with a more pronounced effect after phase 16
and 17 (100 and 400 days after excavation, not shown here).

Apparently, tensile strength and tension softening do not play a major role in this
example, but drawing such a conclusion would be premature. Neglecting the
shotcrete tensile strength completely results in collapse in the first excavation
phase. Also the assumption of ft,28 = 1.5 MPa, Gt,28 = 2.5 kN/m (which is
representative for a steel fibre content of 30 kg/m3) results in failure during
excavation of the central pillar. This demonstrates that the tensile strength of the
shotcrete is essential for the tunnel stability, even though there is little influence
of tension softening on overall lining deformations.
Internal report: Shotcrete model 30

Fig. 22: Vertical lining deformation after phase 15 - shotcrete model

This becomes more obvious if stresses in the side drift walls are considered,
which are subjected to significant bending. Fig. 23 shows the tensile principal
stresses for different assumptions for the shotcrete after centre bench excavation.
The elastic calculation with E = 5/15 GPa yields tensile stresses of ~15 MPa,
which are far beyond the tensile strength of the material. Applying the shotcrete
model without the softening part limits tensile stresses to the tensile strength of
3.0 MPa, but only with the softening part of the model the tension cracking and
subsequent strength reduction can be taken into account. Assuming a lower
tensile strength and fracture energy facilitates an additional hinge forming at mid
span of the side drift walls, which leads to subsequent collapse of the tunnel.
Adding creep strains reduces stresses slightly, but does not prevent the
occurrence of tensile cracks.

Fig. 24 compares bending moments M and normal forces N of the left side drift
wall after phase 14. The internal forces have been obtained by numerical
integration of the lining stresses. Accounting for plasticity and softening does not
only reduce maximum bending moments, but also normal forces, as the side drift
wall becomes softer relative to the outer lining. Creep, however, has a negligible
effect on both N and M at this stage.

The effect of creep is more pronounced for the outer lining at 100 days after
excavation (Fig. 25, Fig. 26). Bending moments decrease by up to 60% if creep
is taken into account, whereas normal forces decrease only by ~10% (Fig. 27).
Internal report: Shotcrete model 31

Fig. 23: Major principal stresses (tension) at side drift - outer lining intersection
after phase 14 (scale from 0 to 3.0 MPa)

Fig. 24: Internal forces of left side drift wall after phase 14
Internal report: Shotcrete model 32

Fig. 25: Bending moments of outer lining after phase 14

Fig. 26: Bending moments of outer lining after phase 16

Internal report: Shotcrete model 33

Fig. 27: Normal forces of outer lining after phase 16

5.5 Conclusions
The following conclusions can be drawn from this study:
1. Tension softening occurred at the connection between side drift walls and
the outer tunnel lining, but did not have a notable effect on overall lining
deformations and internal forces. However, tunnel stability during
excavation was very sensitive to the chosen tensile strength and fracture
energy.
2. The compressive strength of the shotcrete was only reached in very few
stress points. Compression softening did not play any role.
3. The effect of creep on internal forces was not significant during
excavation, which appeared to be primarily load controlled. At later
stages, however, considering creep effects resulted in a notable reduction
of bending moments. Normal forces were less affected by creep.
Internal report: Shotcrete model 34

6 Open issues
6.1 FE-convergence and structural snap back
Problem
In some cases involving tension softening with low fracture energy, the FE-
calculation does not converge (resulting in ‘soil body seems to collapse’), even
though the model itself can never fail physically. Closer examination reveals that
with crack initiation the global error increases massively (Fig. 28), even though
the step size is gradually reduced by the global iteration procedure. After the step
size is halved ~240 times, the calculation kernel detects a step size which is
almost zero and the calculation is stopped for numerical reasons.

Fig. 28: FE-model and global error over steps

The global error or out-of-balance-force (obf) can be understood as the difference

between the elastic prediction and the real solution according to the constitutive
model. In the example in Fig. 28 the sharp increase of the obf coincides with the
occurrence of the first tension cracks in the jet grout column. As soon as the first
tension cracks appear, the force associated with the current displacement reduces.
This results in unloading of the surrounding (elastic) soil mass and a movement
of the column to the left, as the column is effectively pushed back towards its
initial position (snap-back). This load redistribution within the FE-model (the
prescribed displacement remains constant) leads to further strain localization and
softening within the crack, as the crack is now the weakest part of the column. So
even though the applied prescribed displacement remains constant, softening and
crack opening continues due to the forces released by the unloading soil mass. To
reach static equilibrium within that process not only would the force need to
drop, but also the applied displacement would need to decrease. As Plaxis cannot
decrease the applied displacement, the global error increases sharply at first and
is converted into external forces afterwards (while further scaling down the step
size).
Internal report: Shotcrete model 35

Mechanical background
Such effects are not uncommon in numerical analysis of brittle structures and
have been investigated since the 1980ies (e.g. de Borst 1987). Whether a
structure made of strain softening material behaves in a brittle or ductile manner
not only depends on the material behaviour formulated at stress point level, but
also on the size of the structure, with the response becoming ever more brittle the
larger the structure is (Fig. 29, Fig. 30) . This is due to the increase of energy
released by the unloading part of the structure compared to the fracture energy
dissipated in the crack. If the energy in unloading is larger than the fracture
energy of the crack, both forces and displacements need to decrease in order to
reach equilibrium. In a displacement-controlled problem that is usually not
possible, and instead of following the true load-displacement curve the forces
drop vertically at constant displacement

Fig. 29: Structural response of tension bar depending on its length (Karihaloo
2003)

Fig. 30: Ductile-brittle dimensional transition in the 3-point bending test

(Karihaloo 2003)
Internal report: Shotcrete model 36

Importance from a practical point of view

It should be noted that this problem leads to conservative numerical results: If the
structure indeed fails due to tension cracking, the global error will continue to
increase as no equilibrium between external and internal forces is found. Plaxis
might terminate the calculation prematurely due to the downscaling of the step
size, but it will not overlook failure.

However, if large models and/or low values of fracture energy are used, Plaxis
may overlook a possible solution based on the residual strength of the material. If
that happens in models which cannot fail (based on engineering experience),
potential users will become concerned about the validity of the constitutive
model, and subsequently may hesitate to use it. Therefore it is desirable to find a
robust and feasible solution for this issue.

Solutions and workarounds

1. Allowing for more iterations before reducing step size: The most obvious
workaround is to increase the number of iterations before the applied load step is
reduced (increasing ‘desired max. number of iteration’). In the cases where the
snap back problem occurred, allowing for 30 iterations before reducing the step
size solved this problem. It can happen, though, that using 30 iterations before
reducing the step size results in a non-converged last step of the calculation
phase.
2. Using a constant step size: By setting ‘desired max. number of iterations’ to
60 any reduction of the step size can be excluded. The applied load is increased
after 60 iterations, regardless of the magnitude of the global error, and the obf of
the previous step is added to the external loads. This approach is not feasible for
most cases, as usually the obf in the last step of the calculation phase is too large,
and the calculation is terminated.
3. Inserting nil steps: As the cutting of the step size is the reason for the
eventual termination of the calculation, inserting nil steps without any increase of
Mstage should solve this issue. After the obf has increased in e.g. three
subsequent steps despite reduction of the step size, only the obf is used to drive
further iterations. Once the convergence criterion is met, nil stepping is stopped
and the standard load advancement procedure takes over.
4. Adding viscous damping: Artificial viscous damping is added to the
constitutive equations, which should be large enough to avoid snap back of the
load-displacement curve, but small enough not to influence the structural
response. These goals, however, appear to be mutually exclusive. Introducing
artificial viscous damping effectively adds fracture energy to the system and
hence changes the material behaviour. As demonstrated in the tunnelling
example, the stability of some concrete structures is very sensitive to changes of
the tensile fracture energy. Increasing this material parameter for numerical
reasons can lead to unsafe results.
Internal report: Shotcrete model 37

5. Performing dynamic calculations: The obf in dynamic calculations is

converted into accelerations, which invoke inertia forces counteracting said
acceleration. The sudden release of energy in concrete cracking is therefore
delayed/damped which can be sufficient to stabilize the calculation. While this
approach represents the true mechanical response, there are also several
drawbacks: Dynamic boundaries are required, and time steps must be in the
range of seconds to invoke relevant inertia forces. It is not clear how these small
time steps are compatible with the time scale in shotcrete hardening, which is in
the range of hours to days.
6. Using a tangent stiffness matrix: The large obf in softening is a result of the
difference between the anticipated elastic behaviour and the true softening
response. That difference could be eliminated by using a tangent stiffness matrix
for the FE-iterations, combined with arc-length-control. The tangent stiffness
matrix for the current stress point can be obtained by adding small strain
variations to the current strain increment and examine the resulting change stress
increment. It is not clear, however, how unloading situations can be detected, in
which the material response will be much stiffer than predicted by the tangent
stiffness.

Until a fully implemented solution is available, option 1 offers a workaround

for more experienced Plaxis Users.

6.2 Crack alignment with mesh orientation

The regularization in tension and compression softening is based on assigning
equivalent lengths Leq to each stress point and adjusting local stress-strain curves
accordingly. All stress points within a finite element have the same Leq, as Leq is
calculated from the area Ael of the finite element and the number of stress points
per finite element, nGP.

Ael
Leq  2 (27)
3  nGP

Consequently, all stress points within a finite element have the same softening
behaviour and should be equally prone to strain localization. Numerical examples
show, however, that there is a strong tendency of strains to localize along
element boundaries (Fig. 31). This is the case for both tension and compression
softening, and for both 15-noded and 6-noded elements.

The consequence of this phenomenon is more severe for compression softening,

as shown in Fig. 32. If the crack can evolve along element boundaries, strains
indeed localize in a single stress point, as assumed in the fracture energy
regularization approach. However, if the crack has to cross a finite element,
strains tend to localize in about 2 to 3 stress points, which increases the overall
fracture energy of the structure.
Internal report: Shotcrete model 38

Fig. 31: Horizontal strain distribution in oedometric shrinkage test with tension
softening, 15-noded elements

coarse coarse
diag

Fig. 32: Influence of mesh alignment on load-displacement curves and shear

strain distributions in uniaxial compression

Unequal weight factors of the stress points due to the shape function of the
higher order 15-noded elements cannot be the reason for this behaviour, as the
same behaviour is observed with 6-noded elements (3 stress points with equal
weight factor per element). A non-local strain regularization approach may solve
this issue, but the high computational demand of this approach complicates its
application in large FE-models. In most engineering concrete/shotcrete structures
compression softening does not play an important role as the structural behaviour
is governed by tension softening. An exception may be heavily reinforced
concrete beams, but the application of the shotcrete model to such structures is
questionable in any case.
Internal report: Shotcrete model 39

7 Summary
Constitutive model
The shotcrete model implemented as a UDSM in Plaxis can account for:
- limited strength in compression (Mohr-Coulomb failure criterion) and
tension (Rankine tension cut-off)
- time-dependent strength, stiffness and ductility
- strain hardening in compression
- strain softening in tension and compression
- stress-dependent and stiffness dependent creep strains
- stress-independent shrinkage strains

All of the input parameters have a physical meaning and most of them can be
obtained from standard uniaxial tension and compression tests. While the fracture
energy in tension could be derived analytically from direct tension tests, inverse
analysis of indirect tests like the bend beam tests is also feasible.
Modelling undrained behaviour is possible (bulk modulus of water is derived
from the final shotcrete stiffness E28 and  = 0.495), but assuming such a
behaviour for shotcrete/concrete is questionable from a mechanical point of view.
The use of a non-porous drainage type is recommended. Please note that, in order
to view stresses in the output program, the corresponding option needs to be
selected in the Settings window (View > Settings > Results > Show stress for
nonporous material).

Application
The influence of the various model features was investigated in a NATM
tunnelling example with double side drift excavation. This excavation method
relies on the capacity of the temporary side drift walls before the outer lining is
completed. The structural behaviour of the tunnel was clearly governed by the
material behaviour in tension. Assuming properties of steel fibre reinforced
shotcrete with a fibre content of 60 kg/m3, tension softening occurred in the
connection between side drift walls and outer lining, but remained a local effect
which did not influence overall deformation. However, tunnel collapse occurred
in the simulation with tensile strength and fracture energy representative of
30kg/m3 of fibre content, which highlights the importance of correct calibration
of tensile strength and fracture energy.

Open issues
The following issues related to the shotcrete model could not be solved within
the constitutive model:
- lack of convergence in some FE-models with low fracture energy
- preferred crack alignment along element boundaries
Internal report: Shotcrete model 40

8 References
ACI 209R-92 1992. Prediction of creep, shrinkage and temperature effects in
concrete structures. American Concrete Institute, Committee 209.
Aldrian, W. 1991. Beitrag zum Materialverhalten von früh belastetem
Spritzbeton. PhD thesis, Montanuniversität, Leoben.
Austin, S.A. & Robins, P.J. 1995. Sprayed Concrete: Properties, Design and
Installation. Whittles Publishing
Balmer, G. G. 1949. Shearing strength of concrete under high triaxial stress -
computation of Mohr's envelope as curve. SP-23 Structural Research Lab., US
Bureau of Reclamation, Denver, Co., Report.
Barros, J. A. O. & Figueiras, J. A. 1999. Flexural behaviour of SFRC: testing and
modelling. J. Mat. Civ. Eng. ASCE 11: 331-339.
de Borst, R. 1987. Computation of post-bifiurcation and post-failure behavior of
strain softening solids. Computers & Structures 25(2): 211-224.
Brameshuber, W. & Hilsdorf H. 1987. Development of strength and deformation
of very young concrete. Proc. SEM/RILEM Intern. Conf. Frac. Concr. Rock.
(Shah, Swartz, eds.), Houston, Texas, USA. p. 409 - 421.
Brinkgreve, R.B.J., Engin, E. & Swolfs, W.M. 2012. Finite element code for soil
and rock analyses. Users Manual. Plaxis bv, The Netherlands.
Candappa, D., Sanjayan, J. & Setunge, S. 2001. Complete Triaxial Stress-Strain
Curves of High-Strength Concrete. J. Mater. Civ. Eng. 13(3): 209–215
CEB-FIP model code 1990. Design code – comite Euro-international du Beton.
London: Thomas Telford.
Chang, Y. 1994. Tunnel support with shotcrete in weak rock – a rock mechanics
study. PhD thesis, Royal Institute of Technology, Stockholm.
EN 1992-1-1 2004. Eurocode 2: Design of concrete structures. European
Committee for Standardization.
EN 1992-1 2004. Eurocode 2: Design of concrete structures. European
Committee for Standardization.
EN 14487 2006. Sprayed concrete. European Committee for Standardization.
EN 14488 2006. Testing sprayed concrete. European Committee for
Standardization.
Gabet, T., Malecot, Y. & Daudeville, L. 2008. Triaxial behavior of concrete
under high stresses: Influence of the loading path on compaction and limit states,
Cement and Concrete Research 38(3): 403-412.
Golser, J., Rabensteiner, K., Sigl, 0., Aldrian, W., Wedenig, H. Brand1, J. &
Maier, C. 1991. Materialgesetz für Spritzbeton. Technical Report FV 696,
Straßenforschung.
Karihaloo, B. 2003. Failure of Concrete, Comprehensive Structural Integrity (2),
edited by I. Milne, R.O. Ritchie & B. Karihaloo, Pergamon, Oxford, 477-548
Internal report: Shotcrete model 41

Kupfer, H. B. & Gerstle, K. H. 1973. Behaviour of concrete under biaxial

stresses. J. Eng. Mech. Div. ASCE 99(4): 853–866.
Marcher, T., Jiricny, F., Schweiger, H.F. and Nasekhian, A. 2010. Nonlinear FE-
analyses of tunnel structures based on EC7 design approaches – a case study
based on the Nivy station, Bratislava. Proc. 14th Dan.-Eur. Conf. Geot. Eng.
(Frankovska, Hulla, Ondrasik, Turcek, eds.), Bratislava, 2.-4. June 2010, CD-
Volume.
Meschke, G., Kropik, C & Mang H.A 1996. Numerical analysis of tunnel linings
by means of a viscoplastic material model for shotcrete. Int. J. Num. Meth. Eng.
39: 3145–62.
Pölling, R. 2000. Eine praxisnahe, schädigungsorientierte Materialbeschreibung
von Stahlbeton für Strukturanalysen. PhD thesis, Ruhr-Universität, Bochum.
Oluokun, F. A., Burdette, E. G. & Deatherage, J. H. 1991. Splitting tensile
strength and compressive strength relationship at early ages. ACI Mater. J. 88(2):
115-121.
Østergaard, L. 2000. Early-Age Fracture Mechanics and Cracking of Concrete.
PhD thesis, Technical University of Denmark.
Schädlich, B. & Schweiger, H.F. 2014. A new constitutive model for shotcrete.
Proc. 8th Eur. Conf. Num. Meth. Geot. Eng., accepted for publication.
Schädlich, B., Schweiger, H.F., Marcher, T. & Saurer, E. 2014. Application of a
novel constitutive shotcrete model to tunnelling. Proc. 2014 ISRM Eur. Rock
Mech. Symp., accepted for publication.
Schütz, R., Potts, D.M. & Zdravkovic, L. 2011. Advanced constitutive modelling
of shotcrete: Model formulation and calibration. Computers and Geotechnics 38
(6): 834–845.
Sezaki, M., Kibe, T., Ichikawa, Y. & Kawamoto, T. 1989. An experimental study
on the mechanical properties of shotcrete. J. Soc. Mater. 38: 106–110.
Wierig, H.-J. 1971. Einige Beziehungen zwischen den Eigenschaften von
„grünen“ und „jungen“ Betonen und denen des Festbetons. Betontechnische
Berichte, Beton-Verlag, Düsseldorf.