See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/27343534

Service life prediction and repair of concrete structures with

spatial variability

Article · January 2007

Source: OAI

CITATIONS READS

17 305

2 authors, including:

A.C.W.M. Vrouwenvelder
TNO
79 PUBLICATIONS   985 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Reliability analysis and updating of dikes for piping View project

Explosions in tunnels View project

All content following this page was uploaded by A.C.W.M. Vrouwenvelder on 24 February 2015.

The user has requested enhancement of the downloaded file.

 Service life prediction and repair of
concrete structures with spatial
variability

Li Ying,
Royal Haskoning, China, Delft University of Technology, Delft, The Netherlands

Ton Vrouwenvelder,
Delft University of Technology, Delft, The Netherlands

Due to various mechanical, physical and chemical processes, concrete structures are subject to
deterioration such as rebar corrosion, cracking and spalling. As most parameters in those
processes are random, probability-based reliability analysis is often applied. However, in
most studies, spatial variability is not taken into account, although this phenomenon greatly
affects the behaviour of concrete structures. The aim of this paper is to show the service life
prediction procedure when spatial variability is included. Compared with the studies without
consideration of spatial variability, the results can be considered as a better simulation of
reality. Also the choice between different repair and maintenance strategies can be
underpinned in a more realistic way. A hypothetical illustrative example is shown.

Keywords: Service life, deterioration, concrete structure, spatial variability, repair criterion,
maintenance, repair strategy

1 Introduction

According to standard degradation models, the process of concrete deterioration includes

an initiation period and a propagation period. Models for the processes in these two
periods are available (Engelund and Sørensen, 1998; Duracrete, 1998; Duracrete 2000). As
most parameters in those models are random, probability based analysis is often applied.
However, the fact that many parameters also show a random spatial variability is usually
not explicitly taken into account. These variations in behaviour may be linked to
dependencies on temperature, water binder ratio, micro climate, humidity and
workmanship (Chryssanthopoulos, 2002).

HERON Vol. 52 (2007) No. 4 251

Spatial variability of physical properties includes systematic spatial variation (variation of
the mean value and standard deviation) and random spatial variation. Consider by way of
example a concrete bridge deck: the chloride content at the side-areas of the deck is
normally higher than in the centre part of the deck due to the spray of chloride by passing
vehicles. As a result the mean of the chloride content is higher at the two side areas than in
the centre, this effect is called “systematic spatial variation”. At the same time the chloride
content varies from point to point around its corresponding mean value, independent of
the area under consideration. This property is called “random spatial variation”.

The systematic spatial variation is relatively easy to handle since we know (by definition)
its variation laws. For example, the previously mentioned concrete bridge deck can be
subdivided into smaller regions or elements (e.g. two regions at the two side areas of the
deck and one region in the middle part of the deck) where it is assumed that no systematic
spatial variation takes place with each region. For random spatial variation, it is more
complex and difficult; a more advanced modelling of the random variables as well as a
more advanced calculation procedure are necessary. It will be introduced later in this
paper.

Many current studies neglect the random spatial variation within a structure or an element
and take the whole structure or element as fully correlated, that is homogeneous; the result
at one point holds for the entire element. Such a model would only suggest a total repair or
replacement of the structure or the element and local repair seems to be ruled out as a
reasonable option. It maybe clear that such a model is not suited in all cases (Duracrete,
1999a). For example, the appearance of rust stains, cracking and spalling of concrete
structures are not the same across the whole concrete surface, and the level of exposure to
aggressive agents such as chlorides is also spatially variable. One should be aware that all
such spatial variations might be important and could significantly influence the structural
and serviceability performance of structural systems. The influence of the spatial variation
of physical quantities on the reliability analysis is discussed in (Li, et al, 2003). Stewart
(2003) has found that the inclusion of spatial variability of pitting corrosion can lead to
significant decreases in structural reliability, at least for flexural reinforced concrete
structural members. Therefore, if the spatial variation is perceived to be important, it
should be modelled in such a way that the correct decisions can be derived from the
model. This paper is concerned with the service life analysis of concrete structures by using

252
 an advanced spatial variability approach, which implements the random spatial variation
of property differences across the structure. A hypothetical concrete bridge element
deteriorating due to chloride-initiated corrosion is illustrated to show the service life
prediction and its optimal repair strategy.

2 Service life analysis of concrete structures with spatial variation

In probabilistic analysis, three steps can be distinguished:

1. Definition of failure modes (limit states) and corresponding models.

2. Quantification of the statistics of the random variables
3. Calculation of the desired results as for instance failure probabilities.

In this paper a durability analysis for concrete structures is carried out. Special attention is
given to the spatial variability of the random quantities. This has consequences for the
models in step 2 and, even more, for the analysis in step 3. Step 1 remains more or less
unaltered, but will be discussed for the sake of completeness.

2.1 Service life models

The service life of concrete structures is commonly modelled as a two stage process,
defined respectively as the “initiation” and the “propagation” stage (see Figure 1.) Many
studies have already been performed based on these processes over the last decades.
Physical modelling is possible by reasonable approximation, although some aspects are
still under debate. The limit state functions (critical failure modes) will be developed based
on these physical models.

Initiation Propagation

I II III IV I. Initiation of corrosion

Time II. Cracking
III. Spalling
IV. Collapse

Damage

Figure 1: Main events related to the service life of concrete structures

253
The initiation period is a period during which chloride ingress occurs into the concrete
cover until, eventually, depassiviation takes place and rebar corrosion starts. In the model
this event is associated with a critical chloride concentration at the depth of the rebars.
Once corrosion of a rebar in concrete has been initiated, phenomena may occur such as
reduced rebar cross-section, deterioration of concrete cover, cracking and spalling of
concrete, loss of steel-to concrete bond, etc. If corrosion proceeds at a sufficiently high rate,
all of these phenomena may negatively affect performance and eventually structural
capacity (Duracrete, 2000). Actually, corrosion takes place during the whole propagation
period. Initiation of rebar corrosion itself does not necessarily represent an undesirable
state, but without initiation the probability of these negative phenomena is absent. This is
why in many service life approaches, initiation is taken as an indicator of the need to carry
out maintenance; usually preventive maintenance is sufficient to secure all required levels
of performance. Models to predict corrosion initiation are simpler and more elaborated
than models to predict cross sectional loss of rebar, cracking or spalling.

2.2 Statistical quantification of the variables

The specification of a random quantity in the service life model for concrete structures
needs to include:

1. Distribution type
2. Parameters like mean value and standard deviation
3. Its fluctuation pattern in time
4. Its fluctuation pattern in space

In most studies only the first two or three items are dealt with. In this paper, we will also
deal with item 4. If a physical quantity fluctuates randomly in space, it is called a random
field. The fluctuation includes the systematic (deterministically known) spatial variation
and the random spatial variation. Systematic spatial variability may be taken into account
by e.g. different values for chloride concentrations. In order to describe the spatial random
field it is not sufficient to have only the distribution at an arbitrary point. The full
description of a field requires additional modelling with respect to its correlation structure
in space. A common type of field is the Gaussian field. Here the density function is
Gaussian for all points in space. In order to describe such a process in detail we need to
have:

254
1. The mean as a function of spatial co-ordinates
2. The standard deviation as a function of spatial co-ordinates
3. The correlation for each pair of points in space.

If the mean and the standard deviation do not depend on the spatial coordinates and the
correlation is only a function of the distance Δx between two points, the field is called
homogeneous. It is assumed here that the considered fields are homogeneous Gaussian
fields. In reality, of course, random fields are seldom homogeneous. Many properties may
differ systematically from point to point. For many applications, however, the
homogeneous Gaussian field may be a helpful tool in the statistical description of spatial
random properties. Note that it is also possible to superimpose a homogeneous field on a
deterministic field that describes the systematic variations in space.

To describe a homogeneous Gaussian field, we need a value for the mean, a value for the
standard deviation and a correlation function depending only on the distance Δx. The
correlation ρ between different pair of points at distance Δx can be described by the
following generally known expression:

Δx 2
ρ( Δx ) = ρ0 + (1 − ρ0 )exp( −( ) ) (1)
d

where ρ0 represents the common source of correlation and d the scale of fluctuation [m]. By
way of example, Figures 2 and 3 schematically show the fluctuation of the concrete cover
of a beam and its correlation in space by distance when ρ0 = 0.5 and d = 2.0 m.

c o n c re te c o v e r

Figure 2: Realisation of spatial fluctuation of concrete cover along a beam, corresponding to the
correlation function of Figure 3, where L is in the order of 20 m

255
 ρ(Δx)
1
0.8
0.6
0.4
0.2
0
0 2 4 6
distance Δx [m ]

Figure 3: Correlation function according to Eq. (1) when ρo = 0.5 and d = 2.0 m

2.3 Reliability calculations

A structure can be divided into regions based on different criteria, for example based on
vulnerability to damage, environmental condition, loading or construction situation, etc.
Each region is considered as an independent random spatial field, systematic (known)
spatial variability within the region is neglected. For reasons of simplicity, this paper
focuses on one rectangular shaped region only. To enable a numerical calculation
procedure the region is divided into smaller elements. This is shown in Figure 4 for the one
dimensional and in Figure 5 for the two dimensional case. It is assumed that there is no
spatial variation within one element, so the distance between two element centres should
be in the order of half the scale of fluctuation d. Experience shows that this provides
sufficient accuracy.

The Monte Carlo Method can be used to perform the reliability analysis with respect to the
occurrence of each possible failure mode. In this paper it is carried out in the space of
independent standard normal variables. Therefore, a transformation must be carried out
for the correlated random variables due to the spatial variability (Thoft-Christensen, 1986).

1 2 i L/n

Figure 4: Schematisation of 1-dimension model (L/n < 0.5d)

256
 L

W/m
W

L/n

Figure 5: Schematisation of 2-dimension model

The Monte Carlo analysis results in a full probabilistic description of the state of the
structure as a function of time. The effect of various inspection and maintenance strategies
can easily be incorporated. An illustrative example is shown in the next section.

3 Illustrative example

3.1 Introduction
The bridge-element of the motorway over the Eastern Scheldt Storm Surge Barrier in the
southwestern part of the Netherlands, built in 1980, is chosen to illustrate the models
introduced in the previous sections for the service life prediction with spatial variation of
concrete degradation. This structure was investigated as part of a study of durability of
concrete structures along the Dutch coast (Polder & Rooij, 2005). The side view and cross
section of the bridge can be seen from Figures 6 and 7.
We focus here on the part indicated as section G. The dominant deterioration mechanism
for the concrete bridge-element is corrosion of rebar due to chlorides from the sea. The
predicted service life is compared with its design service life of 50 years.

3.2 Models for the service life prediction

It is assumed that the service life of the bridge element includes the initiation period and
the propagation period until cracking and spalling occur. The end of service life is defined
in two ways: (1) when 5% of the area shows concrete spalling; and (2) when 30% of the
area shows concrete cracking.

257
 Bridge
Upperbeam element
(splashzone) Pier (North
side)
Mean sea
levelEasternscheldt side
Sea side

Figure 6: Cross section of the storm surge barrier

5m
Bridge-element Hammen
Chloride test area
8
2002
Eastern Scheldt side

Pier
H 08

Pier
Test
H
area x 09
1989
y

Figure 7: Side view of the bridge element

The limit state function for the initiation of rebar corrosion and the chloride content in the
cover depth at year t can be modelled by the following equations (Duracrete, 1998):

g(t ) = cCl ;cr − cCl (t ) (2)

c
cCl (t ) = cCl ;s ⋅ erfc (3)
4 ⋅ DCl ⋅ t

258
where cCl;s and cCl(t) are the chloride concentration at the surface and at depth c
respectively; cCl;cr is the critical chloride concentration and DCl is the chloride diffusion
coefficient; finally c is the cover depth from the surface to the rebar and t the time. The
propagation period is controlled by the width of corrosion induced cracks in the concrete
cover, therefore the limit state function can be written as:

g(t p ) = wcr ,i − w(t p ) (4)

The actual crack width at tp (time since initation) can be estimated by (Duracrete, 1998 and
1999b):

w(t p ) = 0.05 + β ⋅ [Vcorr , a ⋅ wt ⋅ α ⋅ t p − ( a1 + a2 ⋅ c / φ + a3 ⋅ ft ,spl )] (5)

and wcr,i is the critical value of the crack width (i = 1 for cracking, i = 2 for spalling). In
these equations Vcorr,a is the corrosion rate, wt the fraction of time that corrosion is active
(wet periods), α the pitting factor, φ the bar diameter and ft,spl the concrete (splitting) tensile
strength. The parameters β and (a1, a2 , a3 ) are fitting parameters.

3.3 Statistical quantification

The input for the deterministic parameters and stochastic variables for the models of the
initiation period and the propagation period are shown in Tables 1 and 2. Where relevant a
reference is given.
From the data of the chloride profile investigations (Rooij & Polder, 2002 and Polder &
Rooij, 2005) it may be concluded that the following four variables are important and also
show spatial variation:

1. The diffusion coefficient DCl (initiation phase)

2. The surface chloride concentration cCl;s (initiation phase)
3. Average corrosion rate Vcorr,a (propagation phase)
4. Fraction of time that corrosion is active wt, wetness period, (propagation phase)

The concrete quality for the bridge elements under consideration is very good (Rooij &
Polder, 2002), the concrete cover and the basic concrete splitting strength have a very
small standard deviation. Although normally these two variables probably show spatial

259
Table 1: Input data for initiation period (Rooij & Polder, 2002; Polder & Rooij, 2005)
Variable Distribution Mean Standard Deviation
cCls Normal 5.3% (by mass of cement) 1.47% (by mass of cement)
cCl;cr Normal 0.5% (by mass of cement) 0.1% (by mass of cement)
DCl Normal 8.83 E-6 m2/a 3.69 E-6 m2/a
c Normal 41.1 mm 1.4 mm

Table 2: Input data for propagation period

Variable Distribution Mean Standard Deviation. Reference
β 8.6 (top) Duracrete, 1999b
Normal 10.4 (bottom) 2.85
9.5 (average)
Vcorr;a Normal 0.003 mm/yr 0.004 mm/yr Duracrete, 1999b
wt Normal 0.5 0.12 Duracrete, 1999b
α Normal 9.28 4.04 Duracrete, 1999b
a1 Normal 74.4 mm 5.7 mm Duracrete, 1999b
a2 Normal 7.3 mm 0.06 mm Duracrete, 1999b
a3 Normal -17.4 mm/MPa 3.2 mm/MPa Duracrete, 1999b
c Normal 0.0411 m 0.0014 m Rooij & Polder, 2002
φ Deterministic 0.0010 m Rooij & Polder, 2002
tage Deterministic 20 (age at the inspection year)
ft,spl Normal 4.4 MPa 0.2 MPa Rooij & Polder, 2002
wcr,1 Normal 0.3 mm 0.06 mm Duracrete, 1999b
wcr,2 Normal 1.0 mm 0.20 mm Duracrete, 1999b

variation, their spatial variation is neglected in this example. Based on a dataset of 6

measurements in a test area of about 2 x 0.5 m2 on bridge element H8 (see figure 7), (Li,
2003) has shown that the fluctuation scale for the variables DCl and cCl;s is at least larger
than 0.5 m. Unfortunately the data did not allow much more definite conclusions. In this
example a fluctuation scale of 2 m is selected for these two variables. The same value is
taken for the other two variables. The common source of correlation (ρ0) for all the
elements is set to be zero.

260
 3.4 Service life Calculation
Region G (see shading in Figure 7) has dimensions of 5 by 2 m. The region has been
divided into n = 20 equal elements along the x direction and m = 10 equal elements in the y
direction.
The relative area with initiated rebar corrosion, cracking and spalling in the region G, as a
function of time during 80 years, is shown in the Figure 8. The lower and upper bound
limits are presented in terms of 20% and 80% fractiles respectively. Figure 9 shows the
mean results for the states of initiation of rebar corrosion, cracking and spalling.

One possible but arbitrary definition for the end of the life is that 5% of the surface suffers
from spalling. A second one could be related to cracking e.g. 30% of the surface. The first
criterion leads to an average end of life in about 35 years. At that point in time, about 55%
of the surface has undergone initiation of rebar corrosion and 18% area shows concrete
cracking. In the case of the 30% cracking criterion, we find a life of 40 years.

Initiation of rebar corrosion Development of Cracking

100% 100%

80% 80% low er bound

average
upper bound
Failed area

Failed area

60% 60%

40% low er bound 40%

average
upper bound
20% 20%

0% 0%
0
1980 20
2000 40
2020 60
2040 80 (Year)
2060 0
1980 20
2000 40
2020 60
2040 80 (Year)
2060
Year Year
Development of Spalling
100%

80% low er bound

average
upper bound
Failed area

60%

40%

20%

0%
0
1980 20
2000 40
2020 60
2040 80
2060 (Year)
Year

Figure 8: Development of initiation of rebar corrosion, cracking and spalling during 80 years

261
 Average damage development
100%
initiation of corrosion
80% cracking
spalling
Service life target II:
Failed area

60%
30% area with cracking
40%
Service life target I:
20%
5% area with spalling
0%
0
1980 20
2000 402020 602040 80 (Year)
2060
Year

Figure 9: Average damage development of initiation, corrosion and spalling

3.5 Optimal maintenance strategy

The assessment of concrete structures is primarily based on visual inspections that are
aimed at recording the signs of deterioration. The criterion for the onset of repair and
maintenance in practice is based on a given part of the surface of the structure that shows
signs of corrosion such as rust stains, cracks, spalling or delamination of concrete. The
previous results under the given example conditions show that the average structure does
not meet the design service life of 50 years. If we want (at least) to meet the design service
life of 50 years, interventions are necessary. Three repair strategies are discussed in the
next section to extend the service life of the structure based on the repair criterion of 5%
visible area with spalling (Figure 10).

Area of spalling
Repair target
5%

Time
T1 T2

Figure 10: Criterion for repair strategy

The optimal maintenance strategy is defined as the one giving minimum expected
maintenance cost E[C] (Frangopol, 1999; Gasser, 1999; Li 2004). Human safety is not
relevant. In this paper the value of E[C] is calculated from:

E[C ] = E[Cinsp ] + E[Crep ] (6)

262
 mi C insp , k
E[C insp ] = ∑ tk
(7)
k =1 ( 1 + r )

nr Crep , j
E[Crep ] = ∑ tj
(8)
j =1 ( 1 + r )

C rep ; j = C0 + Cu ⋅ ∑ Arep ;i (9)

Where Cinsp is the cost of one inspection (50 €), mi is the number of inspections, tk is time of
inspection (in years), r the interest rate (0.02/yr), C rep the repair costs, nr the number of
repair cycles, tj the time of repair (in years), C0 the initial cost for each of block of repairs
(5000 €), Cu the unit repair cost (2000 €/m2) and ∑ Arep ;i the total repaired area. The cases
of renewal at the end of the life time are neglected. These costs will be roughly the same for
all alternatives.
A repair strategy is defined as a set of predefined and related measures with the aim to
enlarge the service life of the structure. Three different repair strategies are compared:

a Strategy 1: Repair is carried out for the elements showing spalling (Figure 11).
b Strategy 2: Repair is carried out for the elements showing spalling and the adjacent
elements in the rebar direction (Figure 12), that is, horizontal.
c Strategy 3: Repair is carried out for the area showing spalling and the surrounding
area (Figure 13) (horizontal & vertical).

In all cases the 5% spalling criterion is adopted.

Fai l ed el ement

Figure 11: Strategy 1: Only failed (spalling) elements will be repaired

263
 Failed element Horizontal adjacent
element
Main rebar

Figure 12: Strategy 2: Failed (spalling) elements and its horizontally adjacent elements will be
repaired

Failed element

Figure 13: Strategy 3: Besides failed elements, all its surrounding elements will be repaired

It is assumed that repair is carried out only within the design service life of 50 years and
repaired elements will recover to the state of new.

The effects of different strategies are shown in Figure 14. All three strategies can satisfy
that within the period of 50 years the spalling of concrete never exceeds the 5% limit. As
there is no repair after 50 years, all curves in Figure 14 show turn points around year 50.
Expected yearly repair costs (including annual inspection costs) for different strategies are
shown in Figure 14. All the costs are discounted to the value in year 0. It shows that the
expected repair costs are relatively high from year 30 to 40. The total expected costs are
compared in Figure 15, where strategy 3 has the minimum total cost.

4 Conclusions

In this paper we have developed a realistic approach for evaluating the effect of spatial
variation on deterioration and optimising the repair strategy for concrete structures. It is
based on commonly used corrosion models and probabilistic-based reliability methods.
Additionally it takes into account the spatial variability of concrete properties that has a

264
significant impact on design and maintenance decisions for structures. This approach
reflects the actual situation more realistically and has the flexibility to implement spatial
differences of the structural properties. It can provide useful information to back up the
repair or maintenance strategy for concrete structures. Decision making for the optimal
maintenance or repair strategy is based on the maintenance cost-based optimisation
method. The approach has been demonstrated successfully in a practical case.

Repair strategy 1
1200
1000
repair cost (euro)
Expected yearly

800
600
400
200
0
0 10 20 30 40 50
year

Re pair strategy 2
1200
1000
repair cost (euro)
Expected yearly

800
600
400

200
0
0 10 20 30 40 50
year
Repair strate gy 3
1200
1000
repair cost (euro)
Expected yearly

800
600
400
200
0
0 10 20 30 40 50
ye ar
Figure 14: Expected yearly repair cost of strategy 1-3

265
From a practical perspective, the spatial variability of deterioration is a fact of life. From a
theoretical perspective, prediction of the spatial variability presents a challenge. No
standardization to a specific format has been used until now. For the scope of this research
the availability of data about the actual variability are assumed to be present. Some
practical issues in this area are still far from being resolved.

The deterioration models for concrete structures by the present studies in the world are
still not very satisfying. There is a large amount of uncertainty that comes from inherent
uncertain factors or from basic lack of knowledge.

The approach is worth of more research and development in the future.

Total m aintenance costs

23000 22446

22000
21000
(euro)

19830
20000
18821
19000
18000
17000
Strategy 1 Strategy 2 Strategy 3

Figure 15: Expected total maintenance (repair + inspection) costs of different strategies

Acknowledgements
The financial support of the Ministry of Transport, Public Works and Water Management
(RWS), Netherlands School for Advanced Studies in Construction as well as the
Netherlands Organization for Applied Scientific Research (TNO) is greatly appreciated.
Appreciation also goes to Rob B. Polder and Mario de Rooij from TNO for sharing the data
from the Eastern Scheldt study and their experience in practice.

266
References:

Chryssanthopoulos, M. K & Sterrit, G. (2002), “ Integration of Deterioration Modelling and

Reliability Assessment for Reinforced Concrete Structures”, ASRANet International
Colloquium 2002.
Duracrete (1998), Modelling of Degradation, The European Union – Brite Euram 111, 1998.
Duracrete (1999a), Probability Methods for Durability Design, The European Union – Brite
Euram III, 1999.
Duracrete (1999b), Statistical quantification of the variables in limit state functions,
Document BE95-1347/R9, CUR, Gouda
Duracrete (2000), Final Technical Report, Probabilistic performance based durability design
of concrete structures, The European Union-Brite EuRam III, May 2000.
Frangopol, D.M., and Estes, A.C. (1999), Optimum Lifetime Planning of Bridge Inspection
and Repair Programs, Structural Engineering International, 3/99.
Gasser, M. and Schuëller, G.I. (1999), “On optimising maintenance schedules by reliability
based optimisation,” Case studies in optimal design and maintenance planning of civil
infrastructures systems, Frangopol, D.M. ed., 102-114.
Li, Y., Vrouwenvelder, A. & Wijnants, G. H. (2003), “Spatial variability of concrete
Degradation”, Proceedings for the Joint event JCSS & LCC Workshops in Swiss Federal
Institute of Technology – EPFL, Lausanne, Switzerland, March 24-26, 2003.
Li, Y. (2004), Effect of Spatial Variability on Maintenance and Repair Decisions for Concrete
Structures, PhD thesis, Delft University Press, 2004.
Polder & Rooij (2005), Durability of marine concrete structures – field investigations and
modelling, HERON, Vol. 50 (3), 133-143
Rooij & Polder, (2002) , Investigation of the concrete structure of the Eastern Scheldt Barrier
after 20 years of exposure to marine environment, TNO-Report 2002-CI-R2118-3 final
version, Jan 2003; Appendices, Investigation of the influence of different e exposure
zones on concrete parts of the Eastern Scheldt Storm Surge Barrier, TNO report 2002-CI-
R2118-1A, July 2002.
Stewart, M.G. (2003), Effect of Spatial Variability Of Pitting Corrosion On Structural
Fragility And Reliability Of RC Beams In Flexure, Research Report No. 232.02.2003,
Discipline of Civil, Surveying and Environmental Engineering, The University of
Newcastle, Australia.
Thoft-Christensen P., Murotsu Y., “Application of Structural Systems Reliability Theory”,
Springer-Verlag, 1986.

267
 268

View publication stats