Safety, Loads and Load Distribution on

Structures
by

Håkan Sundquist

TRITA-BKN, Report 108,

th
Structural Design & Bridges 2007, 4 Edition 2015
ISSN 1103-4289
ISRN KTH/BKN/R- -108- -SE
Lecture notes on Structural Design of Bridges

-i-
 Copyright
Div. for Structural Design & Bridges
Royal Institute of Technology (KTH)
Stockholm 2007

 ii 
Preface
This textbook is intended to be used in courses given at the division of Structural Design and
Bridges. It is based on older Swedish compendia and modern literature on safety and relia-
bility of structures and should be used together with other textbooks, issued by the div. for
Structural Design and Bridges at KTH, like
x Sundquist H., Cable Supported Structures, Report 106
x Sundquist H., Arch Structures, Report 107
x Sundquist H., Beam & Frame Structures, Report 109,
x Holmgren J., Westerberg B., Reinforced Concrete Structures, Report 115,
x Sundquist H., Infrastructure Structures, Report 116,
x Sundquist H., Torsion of Concrete Beams, Report 119,
x Sundquist H., Elastic Plate Theory for Bridge Superstructures, Report 120,
x Collin P., Johansson B., Sundquist H., Design of Steel Concrete Composite Bridges,
Report 121 and
x Sundquist H., Silfwerbrand J., Prestressed Concrete Structures, Report 122
for the design of structures according to Swedish Standards like BBK 04, BSK 99, but infor-
mation is also given for design according to the Eurocode.
Stockholm in August 2007
Håkan Sundquist

In this third edition, some corrections has been introduced and more examples have been
added.

In this 4th edition , some corrections has been introduced

Stockholm in October 2015

Håkan Sundquist

- iii -
 iv 
Contents
1. DESIGN OF LOAD-BEARING STRUCTURES ....................................... 1
1.1 Introduction ............................................................................................................................................. 1
1.2 Structures are degraded by use and by the environment .................................................................... 1
1.2.1 Commonly occurring loads .................................................................................................................. 1
1.2.2 The Environment .................................................................................................................................. 2
1.2.3 Use........................................................................................................................................................ 2
1.2.4 Accidental loads due to human activity................................................................................................ 2
1.2.5 Nature reclaims .................................................................................................................................... 3
1.2.6 Violent natural loads ............................................................................................................................ 3
1.2.7 The Richter scale for earthquakes – an empirical scale........................................................................ 4
1.3 How does a design engineer think?........................................................................................................ 5
1.4 Human errors .......................................................................................................................................... 7
1.4.1 We are all human.................................................................................................................................. 7
1.4.2 Quality .................................................................................................................................................. 8
1.4.3 Inspection and control .......................................................................................................................... 8
1.5 The design process for bridges and structures ..................................................................................... 8
1.5.1 General about analysis of risk and safety for structures ....................................................................... 8
1.5.2 Input data and tools in the design process ............................................................................................ 9
1.5.3 Models ................................................................................................................................................ 10
1.5.4 Limit states ......................................................................................................................................... 13
1.6 Safety - uncertainty ............................................................................................................................... 14
1.6.1 Reducing uncertainty to a minimum .................................................................................................. 14
1.6.2 Choice of safety margins .................................................................................................................... 15
1.7 Learning through accidents! ................................................................................................................ 16
1.7.1 The leaning tower of Pisa ................................................................................................................... 16
1.7.2 The Getå accident, the worst train accident in Sweden ...................................................................... 17
1.7.3 The worst building accident due to a structural fault in modern times – a combination of errors ..... 19
1.7.4 What has been learned about the risk of gas explosions .................................................................... 21
1.7.5 Loads due to terrorism........................................................................................................................ 22
1.7.6 Do not always trust beautiful computer calculations.......................................................................... 23
1.8 Bridges which collapse – learning from history ................................................................................. 25
1.8.1 Bridges which shake down. The engineer must also learn from history ............................................ 25
1.8.2 Fatigue failure and other phenomena in metallic materials ................................................................ 26
1.8.3 The designer’s nightmare – the unforeseen load ................................................................................ 27
1.9 Safety during construction ................................................................................................................... 28
1.10 Safety Regulations within the building sector .................................................................................... 29
1.10.1 Swedish regulations and ordinances within the building sector .................................................... 29

v
 1.10.2 Planned joint regulations in Europe ............................................................................................... 29
1.11 Concluding comments ........................................................................................................................... 30
1.11.1 What have we learned and how we shall avoid mistakes in the future? ........................................ 30
1.11.2 Questions of risk and safety are very important in all building work ............................................ 31
2. RELIABILITY ANALYSIS METHODS .................................................. 33
2.1 Preliminary remarks ............................................................................................................................. 33
2.2 The Monte-Carlo method ..................................................................................................................... 35
2.3 The problem G = R  S.......................................................................................................................... 36
2.3.1 Introduction and example ................................................................................................................... 36
2.3.2 The classical solution.......................................................................................................................... 38
2.3.3 Analysis of the function G = R – S using normal probability density functions ................................ 40
2.3.4 Representation as a joint probability density ...................................................................................... 43
2.3.5 The method of Hasofer and Lind ........................................................................................................ 44
2.4 Extensions of the Hasofer/Lind method .............................................................................................. 47
2.4.1 Linear limit state functions with many variables ................................................................................ 48
2.5 Deriving partial safety factors .............................................................................................................. 51
2.5.1 Design formats.................................................................................................................................... 51
2.5.2 Partial factors ...................................................................................................................................... 52
2.5.3 Linearization ....................................................................................................................................... 53
2.5.4 Calculation of partial factors according to codes ............................................................................... 53
2.5.5 Examples of calculation of input data ................................................................................................ 54
2.6 Idealizations of structural systems ....................................................................................................... 57
2.6.1 Series systems ..................................................................................................................................... 57
2.6.2 Parallel systems .................................................................................................................................. 60
3. LOADS ......................................................................................................... 69
3.1 General about loads on bridges............................................................................................................ 69
3.1.1 Different load types ............................................................................................................................ 69
3.1.2 Load positions .................................................................................................................................... 71
3.1.3 Load duration ..................................................................................................................................... 72
3.2 Traffic loads ........................................................................................................................................... 73
3.2.1 General ............................................................................................................................................... 73
3.2.2 Dynamic load effects .......................................................................................................................... 73
3.3 Traffic loads on roads and road bridges ............................................................................................. 75
3.3.1 Allowed loads on roads ...................................................................................................................... 75
3.3.2 Design characteristic loads for road bridges....................................................................................... 78
3.4 Loads on railways and railway bridges ............................................................................................... 81
3.4.1 General ............................................................................................................................................... 81
3.4.2 Allowed loads on railways ................................................................................................................. 81
3.4.3 Designing traffic loads for railway bridges ........................................................................................ 82

 vi 
3.5 Comparison between prescribed load effect on a road bridge compared to load effect on a
railway bridge ...................................................................................................................................................... 83
3.6 Load combination ................................................................................................................................. 84
4. LOAD DISTRIBUTION............................................................................. 89
4.1 General ................................................................................................................................................... 89
4.2 Concentrated loads ............................................................................................................................... 89
4.3 Concentrated load on a cantilever plate.............................................................................................. 91
4.4 Load distribution from a concentrated load on a plate structure .................................................... 97
4.4.1 General ............................................................................................................................................... 97
4.4.2 The use of influence area ................................................................................................................... 97
4.4.3 Approximate method using fictitious distribution widths ................................................................ 100
4.5 Load distribution between two beams............................................................................................... 102
4.5.1 General ............................................................................................................................................. 102
4.5.2 Lane factor not taking the restraint from the beams into consideration ........................................... 103
4.5.3 Optimal distance between beams with respect to moment in the slab.............................................. 104
4.5.4 Calculating the lane factor for a two beam bridge including the interaction between the moment in
the slab and the torsion in the main beams. .................................................................................................... 105
4.6 Load distribution between many beams ........................................................................................... 113
4.6.1 Stiff cross beams .............................................................................................................................. 113
4.6.2 Structural co-operation between main and cross beams ................................................................... 117
4.6.3 Load distribution between stiff longitudinal beams and a roadway slab ......................................... 118
5. LITERATURE ............................................................................................ 123
5.1 Literature on Risk and safety for buildings and structures ............................................................ 123
5.2 Literature on load distribution .......................................................................................................... 125
6. APPENDIX................................................................................................. 127
6.1 Designations and notations for safety analysis ................................................................................. 127
6.2 The Normal distribution N(0,1) ......................................................................................................... 127
6.3 Log normal distribution LN(OH) ...................................................................................................... 129
6.4 Rectangular distribution .................................................................................................................... 131
6.5 Combination of parameters for loads or strengths .......................................................................... 132

 vii 
 The first scientist that scientifically studied design and reliability of
structures was probably the Italian researcher Galileo Galilei, (1564 –
1642) in his famous book ’Discorsi e Dimostrazioni Matematiche, intorno a
due nuove scienze’ (1638) from which this nice picture is taken.
(Discourses and Mathematical Demonstrations Relating to Two New
Sciences)

 viii 
 1. Design of load-bearing structures
1.1 Introduction
All structures and materials are degraded and destroyed with time. The only thing which
varies is the time-scale of the degradation. A sandcastle on the beach can perhaps survive for
a few hours, whereas the breaking down of a mountain can take billions of years. In the case
of a building structure, it is important that it is designed so that it has the life-time which has
been intended. Throughout its envisaged lifetime, the structure shall also have a high level of
safety to resist the forces and loads to which it is subjected and the environment to which it is
exposed. In this chapter, we shall discuss risks and safety aspects in relation to different types
of structure and in particular the influences of nature and of human beings. In this context
building structures include everything which humans build, e.g. houses and civil engineering
structures of various kinds, such as bridges, locks, masts, tunnels etc

1.2 Structures are degraded by use and by the environment

There are many factors seeking to degrade and destroy building structures. These degrading
factors are usually divided into three categories, viz.: loads, environment and use. Use often
involves loads, so that this division is not completely logical, but it has practical advantages
in a building structure context.

1.2.1 Commonly occurring loads

The loads are usually those which the structure has been designed to support or withstand.
The walls and roof of a house are made to provide protection against e.g. snow and wind, and
the structures are therefore designed so that these loads can with great probability be borne.
That the snow load varies with the geographical location is self-evident. In the north of
Sweden will probably be greater than that in Skåne in the south of Sweden is self-evident.
Walls are designed so that the greatest wind force which occurs within a period of 50 years
can with a high degree of probability be withstood. Loads of this type are called natural loads.
Snow and wind are commonly occurring natural loads. There are also more rare natural
loads, to which we shall return later.
A bridge or a system of structural elements has been designed to carry the loads for which the
bridge or the house has been designed. We may call these loads functional loads, since it is
precisely these loads for which the structure has been made. To find the correct dimensioning
load for a bridge, it is possible to proceed in two ways. One way is to create rules for the
maximum size and weight of the vehicles which are to be allowed to use a certain system of
roads, and the second way is to measure the real loads through weighing stations or in some
other way. As in all human activities, however, there are those who do not follow the rules
and regulations and it happens that vehicles which are much heavier than the permitted limit
are in fact driven over bridges. Since such behaviour can lead to an accident, bridges are
always dimensioned so that they can indeed withstand certain overloads. However, it is
theoretically possible that a series of unfortunate events occur at the same time, and that the

1
safety margins are thereby exhausted. Nevertheless, building structures belong to those
systems in society which have the greatest margin of safety, and accidents involving finished
structures are very rare.
Injuries and accidents are in fact much more common during the constructional stages. Every
year construction workers are injured or lose their lives because, for example, the formwork
shuttering is unable to support the loads from a concrete casting. It may seem illogical that
different margins of safety apply during the building period and during the useful lifetime of
the structure, but tradition and economy have led to different safety levels for these two
situations. Similar illogical divisions can also be found within other spheres.

1.2.2 The Environment

The influence of the environment on a building structure is unavoidable. The most common
environmental load is probably corrosion in steel, but all materials are degraded in the
environment in which they act, even though their lifetimes can be extremely long. Other
environmental loads include various forms of chemical action, and erosion by wind and
water. Those human beings who use the structures also contribute to a kind of “environmental
load” as another term for “use”.

1.2.3 Use
A structure is of course intended to be used, and this means that, in addition to the various
loads mentioned, it is also subject to wear and tear. The activities which take place in the
buildings wear on the structure. Protective paint is scraped away, blows and impacts wear
away material, and small but repeated loads cause small cracks which in the long run mean
that the strength of the structure is reduced.

1.2.4 Accidental loads due to human activity

In a structural context, the concept of accidental load is used to indicate an unexpected and
rare influence on a structure. There are a great number of possible accidental loads which are
due to human activities. We can imagine cars which drive into buildings and break pillars,
explosions from gas stored in the houses, an aeroplane which crashes onto a building, or
terrorists who seek to blow up a monumental building. A few cases of this type are discussed
in the section “Learning through accidents!”. Buildings are to various extents designed and
built so that they can also withstand this type of load up to a certain level. Nuclear energy
plants situated close to airports are of course dimensioned so that nothing serious would occur
if an aircraft were to crash on them. All great buildings are designed and built so that a single
load-bearing part, e.g. a column, can be broken without any serious consequences. Very
important buildings are specially designed so that they can withstand terrorist activity etc.
Nevertheless, there are influences here which cannot be coped with within a reasonable
economic framework. The probability of the size and effect of the accidental load must be
assessed by those responsible for the design and construction of the building.

2
The most important and most common accidental load is fire, although this is so common that
it is often given a special heading. Practically all buildings are dimensioned so that they resist
fire for a certain time. It shall be possible for a large and important building to be exposed to
fire for a considerable time without collapsing, so that it is possible to evacuate people from
the building. Consideration is also given to the safety of the fire-fighters in attempts to ensure
that the structural framework does not collapse while the fire is being extinguished. There are
nevertheless certain cases that cannot be coped with, since it is uncertain what is stored in the
building. Here, there are also different risk levels for the people who are expected to be in the
buildings and for the firemen whose task it is to extinguish the fire.

1.2.5 Nature reclaims

The earth is not a dead body. The surface of the earth is continually changing e.g. through tectonic
movements in the earth’s crust which give rise to the violent natural loads described in the
next section. Inland ice also gives rise to movements in the earth’s crust, and rivers erode
loose soils. When these movements have occurred, the earth’s surface wants to restore the
state of equilibrium. This gives rise to earth slips in slopes, landslides and avalanches.
Geotechnicians and engineering geologists work to assess and reinforce earth and rock slopes
and to create stable rock and earth mineshafts.

1.2.6 Violent natural loads

Our earth is subject to a large number of unusual natural loads of the accidental type. A few
examples are listed here. The first that one thinks of are earthquakes, but nature also affords
many other surprises such as volcanic eruptions, typhoons, floods and meteors. Whether and
to what degree a structure is to be designed to withstand such loads is dependent upon the
probability with which one judges that such an event can occur. In areas where it is known
from experience that earthquakes occur, the structure is dimensioned for earthquake loads.
Maps are drawn up which show the expected probabilities and magnitudes of such loads and
the dislocations in the ground which can be expected.

The serious earthquake in Kobe in Japan in 1995 is however an example of the fact that one
did not have sufficiently good knowledge. The area in which the earthquake occurred had not
been classified as one of the areas at most risk, and the strength and spread of the earthquake
surprised the experts. The next section describes how earthquakes are classified.

We read about serious damage to buildings caused by typhoons, tornadoes and floods. The
fact that damage occurs is often due not to a lack of knowledge of the risks but rather to a lack
of economic resources to safeguard buildings and other installations against these rare and
large natural loads. Some people live under a constant awareness of the fact that natural loads
can cause great damage. Large areas of Holland can be seriously damaged by floods.
However, security systems have been built up over a long period of time and extensive
resources are devoted to ensuring that the security level is high. The Dutch are perhaps the
most skilful engineers in the world with respect to building in water, and the security
considerations which are adopted there have been an example for many other countries.

3
There are of course natural loads against which one cannot provide protection. The population
of Iceland, for example, live all the time with the threat that a volcanic eruption may take
place in the central part of the island where it is crossed by the Mid-Atlantic ridge.
The earth is constantly being bombarded by meteors. Large such meteors can reach as far as
the surface of the earth. The world’s largest explosion in the modern era was probably caused
when a meteor exploded at a height of between 5 and 10 km close to the Podkamennaya
Tunguska River in Siberia in 1908. The explosion was so powerful that it is considered to
have corresponded to the explosive strength of 700 atomic bombs of the size of that dropped
on Hiroshima.
There is no absolute safety, and the risk level which is provided for damage due to meteorites
is usually the level which is used as the lower limit of the acceptable risk. It is pointless to
build so that the risk level is lower than this, since a meteorite can strike at any time and cause
enormous damage. The risk that this may take place is however very small.

1.2.7 The Richter scale for earthquakes – an empirical scale

The strength of an earthquake is usually indicated by its magnitude on the Richter Scale.
Strong earthquakes have a magnitude M greater than 6 and earthquakes with M > 8.5 are very
unusual. Originally, the magnitude was related to the reading on a particular American seis-
mometer. The Richter scale was created amongst other things to provide journalists with a
simple measure of the strength of an earthquake. Later, attempts have been made to relate the
magnitude M to the energy E liberated. Figure 1.1 shows graphically an approximate rela-
tionship, where it is evident that an increase of one unit on the Richter scale corresponds to a
30-fold increase in the energy liberated.
The relationship in the figure is very uncertain, but the energy conditions in an earthquake are
so complex and varying that one must be satisfied with a rough empirical relationship
between the magnitude M and the energy liberated E.

(

(

(

(

(
E /J

(

(

( lg E /J = 1,5M + 4,4

(

(
   M   

Figure 1.1 The energy E liberated as a function of the magnitude M on the Richter scale.

4
1.3 How does a design engineer think?
Those who work with structures within the sphere of buildings and civil engineering struc-
tures have built up their own safety thinking, often with the help of regulations developed by
society. The regulations which apply in the building sphere are probably society’s most
extensive safety regulations.
The provision of safety and security in building structures, i.e. in everything which is built by
human beings, is one of the most important tasks of the design engineer and constructor.
Risks and safety act as two mutually dependent opposites in all design work. In many cases,
the building standards contain regulations which help the designer, but the design engineer
nevertheless can and should often reflect on the concepts discussed in this chapter. Matters
relating to the quality work and the quality control of the finished product are often regulated
by the designer himself. In this case, the risks and safety must also be evaluated as a part of
the control.
Issues if these kinds have occupied engineers for a long time and it is plain that in ancient
times the interaction between safety, risk and quality and also how to regulate this was well
understood. We quote freely from Hammurabi’s building code from ca. 2000 BC, shown in
Figure 1.2.
x If a master-builder builds a house for a person and makes the structure so weak that
the house collapses and causes the death of the house owner, the master-builder shall
be put to death.
x If the collapse is such that the property of the house owner is destroyed or damaged,
the own expense.

Figure 1.2 Hammurabi’s building code from ca. 2000 BC was written with cuneiform
script on clay tablets. The regulations in the code placed strict responsibility
on the master-builder for that which was built.

The engineer is of course aware that absolute safety can never be guaranteed. If too high
levels of security were applied for the strength of e.g. an aircraft wing, it would be so heavy

5
that the plane would not be able to take off. In the case of aircraft, one relies not only on a
very qualified design work, but also on constant checks and revisions of the aircraft.
In the case of bridges, the national road or railway administration can have a system whereby
the bridges are inspected regularly on an annual or biennial basis, where the inspection
intervals are dependent on the results of previous inspections.

Thus, the work of the design engineer includes the task of achieving a suitable balance
between risk and safety. In this analysis, the factors to be considered will depend on whether
the task involves
x economic risks, or
x risks to human beings
and the results may be entirely different. In both cases, it is a question of trying to identify
and study the risks, and this can be fairly difficult. The risks can be classified in different
categories:
x Risks which are known about in advance and for which there is statistical documen-
tation based on earlier studies.
x Risks indicated by previous accidents, as a result of which it is possible, through some
form of check list, to evaluate the possible risks which exist.
x Risks which are shown to be possible on the basis of separate and independent studies.
When working with very dangerous systems (e.g. nuclear energy), it is not possible to
rely solely on the experience gained from previous accidents. Here, it is essential to try
to take into account all the potential deficiencies and faults in the systems. For obvious
reasons, this type of risk assessment is extremely difficult and requires very qualified
experts.

The various measures available for handling risks can be divided into the following
categories:
x self-evident measures,
x Standards and Norms,
x the application of previous experience,
x the design engineer’s own assessments, and
x risk analysis.
Within the building sphere, the regulations in Building Codes and Standards are primarily
directed towards the risks for the health and safety of human beings, where the risk is the
probability that the structure will collapse. Modern Standards refer to the consideration of the
“ultimate limit state” under normal loads, of which there is knowledge and experience. For
instance, measurements are made of e.g. how much snow falls within a certain area, and the
structure is then normally designed for a snow load which statistically is not likely to occur
more often than once in 50 years.
For the design of dams and structures in watercourses, accurate measurements of the water
flow are made over as long a time period as possible. On the basis of these measurements, a

6
statistical calculation is then carried out to assess how great the probability is that the flow
rate is of a given magnitude once in e.g. 100 years. Usually, the data are available for much
shorter measurement periods than would be needed for making a certain statement about the
probability, and the weather can have periodic variations with a repeat period of e.g. 100
years.
Certain checks of the failure safety with respect to so-called “accidental loads” are also
carried out. These should perhaps be referred to as “disaster loads” in accordance with the
terminology introduced above. Such loads are often very schematic and are intended, at a
certain (often not unambiguously determined) level, to take into consideration the risk that
e.g. a lorry is driven into a house or that an explosion occurs because LP-gas has been kept in
the apartments.
However, there are many risks that are ignored, such as for example the risk that an aeroplane
may crash or a meteorite strikes the house. The reason why this type of risk is ignored is, of
course, that the probability of such an occurrence is deemed to be extremely small. However,
the designer is himself free, through a suitable design and other measures, to build in a safety
also against this type of risk. In the case of a nuclear energy plant, some of these are in fact
also taken into consideration. Naturally, there remain risks which cannot be eliminated, e.g.
that a meteorite falls and hits the plant, but consideration is e.g. given to the risk that a small
aeroplane might crash into it.
The Building Standards also contain regulations which take economic and hygienic risks into
consideration, e.g. that a building structure develops deformations so great that it is
unpleasant to live in, or that cracks arise which mean that there is a risk of heat and moisture
leakage. Such a consideration is usually referred to as a check on the “serviceability limit
state”.

1.4 Human errors

In all design and construction work, one must include “the human factor” in the calculations.
Human beings are all liable to make errors, and a structure must to a certain extent be
designed so that such errors do not lead to too great risks.

1.4.1 We are all human

An error can lead to serious accidents and of course economic losses. Methods for eliminating
errors have therefore attracted great interest. A certain variation in the design and quality of
the products must always be expected, and the Codes and Standards recognise this by
demanding e.g. that the material properties shall with a certain probability fulfil the
requirements strived for. However, the material manufacturers are usually so anxious to
obtain good test results that a result below the lowest acceptable level is extremely rare. If
such a result is obtained, it is usually because a gross error has occurred, e.g. there has been a
mix-up of materials. Thus, it is important to create systems so that gross errors do not occur.

7
1.4.2 Quality
By quality, we mean that the products shall have the properties which are required and
expected by those who buy and use the product. To deliver higher quality than is required and
expected is poor quality, since this can cause additional costs for which the supplier will not
be paid. Systems for achieving quality have always attracted considerable interest.
Practically all quality systems seek to delegate the responsibility for quality down to those
who work with the products or systems. In this way all those concerned feel such a
responsibility that no errors arise. However, the so-called human factor cannot be entirely
eliminated and the available knowledge is not always sufficient to meet all the problems
which can arise in the processes.

1.4.3 Inspection and control

An alternative way to that described above for supervising that the products and structures do
indeed have the quality aimed at is through control and checks. The old systems where special
supervisors were employed to oversee, measure and test the products have practically
disappeared from the construction industry, since the method was not sufficiently effective.
The efficient way is to delegate responsibility and quality awareness to as many as possible.
All errors can still not, however, be avoided.

1.5 The design process for bridges and structures

The design process for bridges is like that for other types of buildings and structures, even
though the process for bridges requires a much deeper analysis and more extensive
calculations than that for buildings. The interaction with society is also much more
pronounced since roads and bridges are usually constructed on ground owned by the state or
by the local municipal authority. The technical process is usually also much more regulated
with more standards and regulations issued by the road or rail authorities. These regulations
must usually be strictly followed by the bridge designer.

The formal design process for bridges is described in greater detail in the textbook
Infrastructure Structures.

1.5.1 General about analysis of risk and safety for structures

Considerations and discussions about risk and safety are vitally important in the design
process for important structures like bridges. In this and the following chapters, design tools
and processes will be discussed in some detail for structures in general. More specific issues
regarding the use of specific materials will be discussed in textbooks devoted to these mate-
rials. Examples of such textbooks issued by the Division for Structural Design and Bridges
are: Cable Supported Structures, Reinforced Concrete Structures, Prestressed Concrete
Structures, Design of Steel Concrete Composite bridges and Torsion of Concrete Beams.
These textbooks also describe and discuss the design tools and methodologies specific for the

8
particular types of structures constructed in steel or concrete or in combinations of these
materials.

1.5.2 Input data and tools in the design process

The design of houses, bridges and many other types of structure is based on simplified
models. These models can be created by the designer based on experience or can be found in
standards and handbooks. In practice, the models used are often a combination of these
possibilities.
The models can be deduced theoretically or can be based on empirical knowledge. Different
types of models are often used depending on the type of problem to be solved. If the problem
is coupled to a serviceability criterion such as the risk of cracking or excessive deformation, a
model for solving that problem has to be used. If the task is to check whether there is a risk of
total failure of the structure, another type of model should be used. In the first case, a model
based on the theory of elasticity can be used, whereas in the second case the phenomenon of
plasticity must also be considered. Since a total knowledge of the behaviour of a structure at
different stages is in practice not available, the structural designer has to use models that are a
combination of different approximate theories, acquired by tests or experience.
To prepare a correct design of a structure, the designer must have access to or create a set of
basic input conditions and data based on factors such as the following:
x functional demands,
x input data for the quantities needed for the design,
x a model that describes the structural behaviour satisfactorily
x a method for verification of the analysis, and
x a qualitative assurance methodology for the analysis.

Input data can be subdivided in the following categorises;

x geometrical conditions,
x functional loads1,
x environmental actions and
x material properties.

The geometrical conditions are often the most important and most difficult since they depend
on many factors that the designer cannot control. The geometrical conditions are in turn also
dependent on the design process itself. Examples of geometrical conditions are section mea-
surements, heights, lengths, slopes and restrictions. Geometrical conditions also include
deviations in the measurements due to tolerances, unintentional eccentricities and curvatures
etc.

1Actions is often used as a generic term to describe different types of influences on a structure like loads, im-
posed deformations and environmental burdens like wear, salt, scour etc.

9
Actions is a generic concept embracing all the functional loads, imposed loads, imposed
deformations and environmental influences on a structure. The functional loads are the loads
which the structure is designed to withstand, i.e. traffic loads and imposed loads from people
and furniture in a building. Environmental actions are unintended (but often unavoidable)
influences important for the structural behaviour and degradation of a structure such as acidic
rain, thaw salt etc.
In the following chapters, different actions, loads and load combinations will be discussed.
The properties of the materials which build up the structure are decisive for its strength and
stiffness. No material has unambiguously defined properties, so the designer is obliged to use
statistical models to decide which material properties are applicable for use in the design
calculations.

1.5.3 Models
A model for the particular problem concerned is always a simplified model of reality. The
model must have a certain set of fundamental properties to enable it to be used as a basis for
the design. The following set of properties is an example of what may be required
x the structural behaviour must be described satisfactory,
x the actions must be modelled,
x important functional demands must be paid attention to,
x material properties must be included and also
x geometrical conditions.
In addition, it is an advantage if the model
x can be controlled against other models, and
x is so pedagogic in its description that other designers can understand and use it.
In the following chapters, different models will be presented and other textbooks will describe
many other models for both the structural and material behaviour. It should, however, always
be remembered that any model is always a simplification and an approximation to reality.
Figure 1.3 shows schematically how a structural model can be designed and tested against
reality. In the next phase, the model must be checked against the safety and functional
demands. This is done using a design model. An example of such a model is schematically
depicted in Figure 1.4.

 10 
Figure 1.3 Schematic description of how a calculation model is built up and tested against
reality.

The load model indicates the types of load which are assumed to arise and how these affect
the structure. This often involves considerable simplification, e.g. the assumption of a
uniformly distributed load on a beam as an approximation for the true load which may be due
to furniture and human beings acting as point loads and moving in a random pattern. The
geometrical assumptions can include placing in bearings as an approximation for a
complicated support condition in reality. Other geometrical simplifications are the assumption
that an angular support and complicated combinations of measurements can be replaced by a
simply supported beam.

 11 
Figure 1.4 Schematic description of the design process. In the calculation model, “R” is the
load-bearing capacity and “S” is the effect of loads or other actions.
It is common practice and usually appropriate to choose the model “on the safe side”, so that
either the load-bearing capacity is under-estimated and/or the actions are over-estimated. In
order for the design to be carried out, the actions must be expressed in the same units as the
load-bearing capacity. A comparison may, for instance, be made between the calculated
moment which can arise in a given section and the moment which a beam can support in the
same section. In other models, forces and stresses may be compared. In summary, we thus
usually compare a load effect designated S with a load-bearing capacity designated R, see
Figure 1.4.
When we have managed to obtain values for S and R, we must verify that R with a sufficiently
high degree of certainty is greater than S. In Chapter 2 statistical models will be discussed.
The following chapters gives examples of how large the variations in loads, material
properties and geometries can be.
In the design of a structure, whether it be a new project or the control of the load-bearing
capacity of an existing structure, it is usual (cf. BKR 2003) to establish an equation in the
form:

RtS (1-1)

where R is the ability of the structure to withstand different types of actions S.

The actions can be of many different types. They can be due to ordinary loads, accidental
loads, temperature, environmental influences etc, or a combination of these various types.

 12 
The load-bearing ability is in turn dependent on a large number of factors. It may depend on
the material’s strength and ability to be deformed and also on very important factors such as
how the loads and stresses are propagated within the structure.
Eq. (1-1) involves a large simplification, however, because the actions and the load-bearing
capacity are interdependent. There are, for instance, dynamic effects which depend upon the
interaction between these two factors. There are other factors where S and R are not indepen-
dent, such as the frequency with which the loads are repeated, the duration of the loads and
the dynamic augmentation of the effect of the loads.
This textbook considers primarily the manner in which traffic loads affect the structures. The
factor S representing the effect of the load will be discussed in terms of moments, axial forces
and shear forces. In another textbooks, design factors taking into account these forces and
moments will be discussed.
Bridges are designed according to more straightforward structural principles than conven-
tional buildings. Clear structural principles are usually adopted based on the introduction of
bearings and joints. This means that the calculations for bridges are more accurate than those
for ordinary houses. One should nevertheless be conscious of the fact that it is often necessary
to choose a design model that contains large approximations. In particular there is consi-
derable uncertainty in the interaction between the soil and e.g. slab frame bridges.

1.5.4 Limit states

A structural component has a certain strength R(s), and it is exposed to load and stress effects
S(s). Both the load and the strength can be expressed in terms of a mutual quantity such as a
stress, a deformation, an environmental effect or the like s. In a wide sense, we can imagine
various types of influence which can lead to a fault so that the structural component ceases to
function as intended.

The state when the structural component passes from a situation when it has approved
properties to another situation when it no longer functions as intended is called the limit state.
Depending on the functional requirements, it is possible to define an arbitrary number of limit
states. Within the building sector two limits are normally defined, viz.:
x the serviceability limit state, and
x the ultimate limit state.
The serviceability limit state is defined as the limit at which a structure no longer functions in
a fully satisfactory manner with respect to its use. This can be due to defects such as
x comfort and health,
x crack formation,
x aesthetic problems,
x deformations which lead to difficulties, or
x vibrations or oscillations which cause discomfort.

 13 
The ultimate limit state is defined as the state under which the structure will fail under the
axial loads which are assumed in its design.
In addition to these limit states, other limit states can be defined, e.g.:
x the accident load limit state, and
x limit state due to fire.

The accident load limit state is intended to describe the state when a structure fails as the
result of a sudden unusual load of a magnitude which could not normally be anticipated when
the structure was being designed, and which it is not reasonable to expect that the structure
will be able completely to withstand without damage (although it is normally expected that
the extent of the damage will be limited). Examples of accidental loads are:
x collision by an aircraft, vehicle or shipping vessel,
x gas explosions, and
x terrorist activities.
The term “accident load limit state” is not commonly used in Standards, but it can be logical
to use it, and also the term “fire limit state” which is intended to describe the ultimate limit
state which applies in the event of a fire. In certain cases, other limit states can be importance,
e.g. the earthquake limit state in regions where such risks may arise.

In the following sections, an attempt is made to describe how theories have been developed
relating to the task of design with reference to the different limit states. Since the ultimate
limit state is probably the most important and is also the easiest to describe, this is the limit
state which is here illustrated in the first place. The same principles apply however for all the
other limit states.

1.6 Safety - uncertainty

1.6.1 Reducing uncertainty to a minimum
Like all human activities, building operations involve uncertainty. This uncertainty includes
both the load effect and the load-bearing ability of the structure and is due to the fact that both
these parameters are associated with a certain degree of scatter. If, for instance, one measures
how large a load is acting on a plate structure in a house, it is found that there is a large scatter
in the values obtained. In the same way, any measurement of the strength of e.g. concrete
shows a considerable variation in the results.
In spite of this, one cannot feel safe unless the measured values are representative. It might,
for example, happen that a tenant in our building example decides to use his apartment for
some purpose other than living, e.g. as a storeroom for steel plating. It might also happen that
the constructor, either through carelessness or intentionally to save money, departs from the
dimensions or material grades given in the design drawings. It is impossible to guard against
such eventualities by any reasonable safety margin. The same is true of the calculations them-
selves; it is surprisingly easy – especially for an untrained designer – to misplace a decimal
point so that the load is under-estimated or the strength over-estimated by a factor of ten.

 14 
For this reason, it is necessary to distinguish between the uncertainty associated with
temporary random deviations from a given value (load or strength) and the uncertainty due to
gross errors or systematic deviations from the given value. The random deviations can be
taken into account by applying a safety margin; the gross errors or systematic deviations can
be avoided only by education and checks. See Table 1.1.

Uncertainty Counteracting measure

Gross deviations Education and checks

Systematic deviations

Temporary or random deviations Safety margin

Table 1.1 Possible measures to counteract deviations.

It is worth noting that systematic methods have been developed for the assessment of the
necessary safety margins, but that corresponding methods for taking into account the need for
education and checks are largely non-existent. This is particularly remarkable considering that
the majority of the cases of failure described in the literature have been due to a gross or
systematic error. The possibilities available to the designer include demanding the use of
control programmes and the authorisation of those companies which are to carry out the
various tasks. In the case of the designer’s own work, a control programme as described in
ISO 9000 can be implemented. In certain contexts, it has been suggested that a safety system
should also take into account the competence of the designer.

1.6.2 Choice of safety margins

The factors which govern the choice of a safety margin are
x Uncertainties in the load estimate,
x Uncertainties in theories and calculations,
x Variations in material strength,
x Discrepancies in dimensions and imperfections,
x Constructional errors, and
x Damage consequences.
In order to reach the final comparison between strength and load-bearing capacity and there
from calculate the failure risk, it is also necessary to take into consideration that both the load
and the strength are always dependent on a number of factors.
If the uncertainty in the load assumptions, theories etc are expressed in statistical terms, i.e.
standard deviations, it is possible to calculate the uncertainties in the load and strength and
thereafter calculate the failure risk according to Chapter 2.

 15 
If the failure risk is specified, e.g. H 105 it is thus possible to design the structure in a con-
sistent manner taking into consideration in a correct manner the uncertainties in the input
factors. This creates conditions for a fair comparison between different structures.

This procedure also makes it possible to calculate the optimal failure risk from an economic
viewpoint if only economic damage is at stake. By designing the structure for different failure
risks H it is possible to develop a relationship between the building costs A(H ) and the failure
risk. In addition, it is possible to calculate the risk costs, i.e. the costs for repair or rebuilding
after failure has occurred K(H ) multiplied by the failure probability H. The total cost of the
structure during the estimated lifetime will then be A + HK.
The fundamental relationship between the building costs and the failure risk is shown in
Figure 1.5, which shows that the building costs decrease with increasing failure risk. The
figure also shows how the risk costs HK increase with increasing failure risk, somewhat more
than linearly since the extent of the damage and thus the repair costs increase with increasing
failure risk. Summation of the two curves gives the total cost curve, which has a minimum at
a certain H-value, which thus represents the optimal failure risk.

Construction cost
Failure probabilty*Failure cost
Sum
Optimal risk
Cost

Probability of failure (H )

Figure 1.5 Calculation of the optimal failure risk.

1.7 Learning through accidents!

The skilled designer starts his/her work by finding out not only what others have done
correctly, but also what mistakes have previously been made. History is the principal source
from which to be able to learn and gain understanding for the future.

1.7.1 The leaning tower of Pisa

The world’s perhaps most well-known faulty structure is the leaning tower of Pisa. Already
during the building process, the ground below the tower started to settle, and to settle
unevenly. Nevertheless, the construction work continued, while attempts were continually

 16 
made to correct the error. The result is that the tower is “banana-shaped”. To date, the tower
has sunk ca. 2 meters, and the inclination is now ca. 5,3°, as shown in Figure 1.6. The faulty
structure has become a tourist attraction, but experts have been working persistently to seek
methods to reduce the increase in inclination. These efforts have earlier not been completely
successful, but recently applied counteraction methods have probably stopped the increase in
inclination and the safety of the tower is considered to be satisfactory.
+58 m

2,5 m
North South
12,0 m

15,5 m
o HW
5 26 ' +3 m
Sea
0
Silty
sand 19,58 m
-10
Pancone clay
-20
Clay
Sand
-30
Clay

-40
-37 m

-50 Firmly stratified sand

-60

Figure 1.6 The leaning tower of Pisa. The figure shows that the tower has sunk more than
2 metres into the clay.

1.7.2 The Getå accident, the worst train accident in Sweden

Roads and railway embankments are structures which human beings designed and constructed
so that they will bear the loads associated with traffic. They are designed by soil mechanics
engineers who, on the basis of geological information and the results of drillings, either
accept that the existing ground can carry the loads from the embankment and traffic or design
various reinforcing structures. Recently, a large number of effective methods have been

 17 
developed to reinforce soils and rocks which in themselves have a poor ultimate bearing
strength. As in other disciplines, it has also been necessary to learn from history within this
sphere.
In the afternoon of 1 October 1918, the train between Norrköping and Stockholm with an
engine and seven carriages fell down a slope by Getå on the Bråviken shore. The three last
carriages in the train remained on the embankment. Of the approximately 125 passengers and
personnel on the train, 41 persons died.
Just before the train passed the stretch in question, a landslide had taken place. As shown in
Figure 4, the embankment and the main road parallel to the embankment had been built up on
sloping ground.
Shoreline before
Shoreline after

CL- Railtrack
Embankment
landslide

landslide

before slide
Natural
Roadway ground contour
after slide Soil replaced
after slide
Fallen ballast
Groundwater level
Natural ground in moraine
contour beneath
embankment

30 20 30 0 10 20 30 40 50

Figure 1.7 A section through the ground at the site of the Getå landslide in 1918. The
figure shows the surface contour before and after the landslide and also the
approximate earth-layer sequence.

The embankments themselves gave rise to an increased load on the relatively loose earth
which consisted of clay and silt, on top of which material which had slid down from the rock
above had settled in layers. Owing to a lot of rain in the period before the accident, it was
stated that the weight of the embankment had increased and that an overpressure in the soil
layers beneath the embankment had led to a deterioration in the load carrying strength.
The geotechnical commission of the Swedish State Railways had already been formed to
investigate the safety of embankments etc. The Getå accident speeded up research and deve-
lopment work within the geotechnical field and Sweden has made good contributions to the
understanding of how to provide protection against landslides. One can never be completely
certain, and every year landslides of a greater or lesser magnitude occur, often along rivers
and streams which flow through clay areas. It is very difficult to provide complete protection
against this type of phenomenon, unless one can choose to build only on soils with good
bearing-strength properties. As emphasized in the introduction to this chapter, nature will
sooner or later also break down that which we consider to be solid ground.

 18 
1.7.3 The worst building accident due to a structural fault in modern
times – a combination of errors
On Friday 17 July 1981, the atrium in Kansas City’s newest hotel, the Hyatt Regency, was
filled with festively-dressed people who were celebrating the weekend with dancing and
listening to a big band. The Premises had been built as a high glass-roofed well with a ceiling
height of 18 metres. From the ceiling, walkways were suspended at the third and fifth floor
levels, to provide convenient passage between different parts of the building. These walkways
were also filled with people beating time to the music with their feet.
Suddenly, the disaster occurred. Two of the walkways collapsed, casting all those who were
on the walkways down onto those who were on the floor. 114 people were killed and almost
200 were injured. This accident is the worst which has happened to an American building
structure due to structural mistake. No such large accident in peacetime caused by a building
collapse has occurred in Europe in recent times.
An investigation into the cause of the accident rapidly ascertained that the original drawings
of the suspension of the walkways had not been adhered to, and that a change had been made
during the actual building process. The reader is encouraged to study Figure 1.8 carefully.

Figure 1.8 A detail of the building design for the by steel bars supported hanging
walkways which caused the greatest accident due to a structural fault in the
West in recent years.
a) This is how it was conceived and drawn. The force acting on the nut and
washer (B) is here 1F, where F is the load coming from one walkway.
b) This is how the part was built. The force acting on the nut and washer is
here 2F.

The walkways were supported on welded box girders (A) which, with a nut and washer (B in
the original design and C after the change), were suspended in long threaded vertical rods (D)
from the ceiling. The figure shows the suspension of the upper of two walkways, one above

 19 
the other, which were suspended in the same rods. At the time of the accident, the box girders
gave way so that the nuts with their washers were pulled through the girders which were
badly deformed.
What was the reason for this design change? In the original design, the vertical stays were
14 metres long and they had to be threaded all the way up. (Alternatively, the rods could have
had a greater diameter at the threaded sections, but this would have been an expensive
solution). It was totally impossible to handle such long objects and to get the bridges in place.
Two walkways were suspended above each other, and the upper was to be held by nuts which
were to be rotated 9 metres along the thread on the stays. In other words, the design was a
complete impossibility. Somebody had the idea that one could use several short stays which
would then only need to be threaded closest to the ends, and the modified design shown in
Figure 1.8 b) was chosen.
The investigation showed that the original design in itself was under-dimensioned and had
only 60 % of the strength prescribed in the Building Standards. The Standards presuppose a
certain safety factor, and it may be assumed that the Standards include a dynamic addition to
take into account resonance effects caused by dancing people, like the well-known problem of
a marching army. In this case, however, the safety factor was only about 1, i.e. there was no
margin whatsoever. This meant that a fracture in one place led immediately to an overload in
other places which could not even support the load until all the people had been able to reach
safety. In other words, the structure was not “fail safe” and contained no “redundancies”, i.e.
it was not robust and tolerant of partial failure in the event an overload. The result was an
immediate total collapse.
This was nevertheless not the principal reason for the accident. The reader is encouraged to
look again at the figure. In the original design, the nut (B) takes up the load from a single
footbridge. In the modified design, the nut (C) takes up the load both from the upper
footbridge and from that which is suspended below it. This nut is thus overloaded by a factor
of 2. The result was that a poor structure was replaced by a highly dangerous one.
One may ask why nobody questioned the design. There was good reason to do so, when it
was found necessary to make a change. Warning signals were also present. The building
workers noticed that the walkways swayed and they therefore made detours with their
wheelbarrows to avoid crossing them. With the obvious safety risks, the structural engineer
should have reconsidered the design of the structure.
There were other alternative solutions. If the threaded rods were too long, it would have been
possible to join several shorter ones together with threaded sleeves, and thus have retained
their load-bearing ability.

When the walkways were rebuilt, they were supported by columns from the floor. In this way,
the impression of the free volume of the hall was of course lost, but the walkways were much
more stable.

This example has been taken from a thought-provoking book: “To engineer is human” by the
American professor Henry Petroski. His theme in the book is that one must take risks in order
to be able to make progress. It is not strange that one sometimes fails, but one must learn from
experience.

 20 
1.7.4 What has been learned about the risk of gas explosions
Another case which has been very important for the development of Standards and Codes for
buildings is the collapse in May 1968 of a part of a residential block of apartments at Ronan
Point outside London, as shown in Figure 1.9.

Figure 1.9 The collapse of a part of a prefabricated constructed block of flats at Ronan
Point outside London. The collapse began as the result of a gas explosion in a
flat on the 18th floor and then spread through almost all the building.

The collapse was initiated by a gas explosion in one of the corner flats on the 18th floor. The
house was a structure based on prefabricated elements and several walls fell out. Owing to a
poor connection between the elements, more elements both above and below the accident
floor collapsed like a house of cards, and many people were killed or injured.

 21 
In most countries, this accident has led to the requirement that a building shall have a certain
protection against so-called progressive structural failure. In most cases, the protection means
a requirement regarding various connecting devices between the pre-fabricated elements.

1.7.5 Loads due to terrorism

A disaster which will influence the design and building of all buildings for a long time is the
flying of aircraft into the World Trade Center (WTC) and the Pentagon in the USA in 2001.
Already in 1945, a small bomber flew into the then highest building in the USA, the Empire
State Building in New York. The building was damaged only locally and the number of
people who died was small. A fire also started, but the amount of fuel was small and the fire
was rapidly extinguished.

Designers thus became aware of the risk of aircraft flying into tall buildings and, in the design
of special buildings, the risk of an aircraft flying into the building has to a certain extent been
taken into consideration. To build so that no damage arises is probably almost impossible, but
to construct so that the possible damage would not be much worse than in the accident in
1945 has been the target in some cases. The intention was thus to design the World Trade
Center towers so that they would resist being flown into by a large air liner which had got off
course by mistake, without the building collapsing. In the event, the combination of the local
damage caused by the aeroplane crashing into the tower and the subsequent extremely strong
fire led to the total collapse of the towers, as indicated in Figure 1.10:
x The collision between the aircraft and the building seriously damaged the outer load-
bearing shell, Figure 1.10 a) and Figure 1.10 b),
x The fire softened the internal supporting columns, since a large part of the fire pro-
tection had been damaged by the explosion which accompanied the actual collision,
see Figure 1.10 c),
x The framework remaining in the damaged region was no longer able to support the
load due to the weight of the whole upper part of the building, Figure 1.10 d), and
x When collapse had actually started, the kinetic energy of the collapse was so great that
nothing could prevent it from continuing down through the whole building,
Figure 1.10 e).
At the present time, there is an intensive debate among designers as to what can be done to
reduce the risk of total collapse of tall buildings in a manner similar to that which took place
in the World Trade Center. It is unavoidable that there is great damage and that many people
are killed if an aircraft flies into a building, but specialists are agreed that measures can be
taken to improve the safety of important buildings in the face of terrorist attacks. These
measures involve chiefly better fire protection and designs which ensure that different
structural parts are held together better if some parts fail.

 22 
 a) b) c) d) e)

Figure 1.10 Schematic illustration of the collapse of one of the towers of the World Trade
Center in New York on 11 September 2001. a) and b), the collision between the
aircraft and the building caused an explosion and seriously damaged the outer
load-bearing shell; c) the fire softened the internal supporting columns since a
large part of the fire protection had been damaged in the explosion, d) the
remaining framework in the damaged part was unable to support the weight of
the whole upper part of the building, and e) when the collapse had begun, the
kinetic energy was so great that nothing could prevent it from continuing down
through the whole building,

1.7.6 Do not always trust beautiful computer calculations

Another very interesting example is the collapse of the Sleipner A oil rig in 1991. This type of
platform is a Norwegian speciality and many have been built, and several intended for even
greater depths of water are being planned. These gigantic structures have been built in a
special way, utilising the conditions existing along the Norwegian coast with great depths of
water. The design of the platform support is indicated in Figure 8.

 23 
 a) 110 m CL
97,5 m

32 m

c)

m
24

Deformation d)

pi
0
55
Detail A
Ø 25 c170

Spalling

pj Crack

Spalling and deformed

Ø 25 c170 reinforcement

800

Figure 1.11 The Sleipner A drilling platform. a) the dimensions of the platform, b) the
FEM-model created to calculate the stresses and deformation in the structure,
c) the probable fracture surface in the structure, and d) the approximate
fracture surface found in the tests subsequently carried out by the Swedish
Government Testing Institute.

The structure shown in Figure 1.11 was towed out and partially filled with water, the inten-
tion being that it would sink to a depth which would make it possible to float out the super-
structure (the actual platform) and place it on the towers. While being submersed, the whole
structure suddenly collapsed because a wall between the internal water-filled triangular
spaces was pressed in. This caused the platform to start to sink and several walls collapsed. In
a few minutes, the structure lost its buoyancy and the whole structure, worth ca. 150 million
USD at the time of the collapse, sank to the bottom with such a force that the impact was

 24 
sufficient for readings to be recorded by seismologic stations. All the people on the structure
were able to get to safety, so the whole event was only an economic disaster.
What then was the cause of this disaster? Two reasons have been put forward:
x Incorrect calculations caused by too coarse and incorrectly evaluated Finite Element
Method (FEM) analyses. An FEM-analysis is today the most widely used method for
carrying out an approximate analysis of a complicated structure.
x Unsuitable reinforcement design (see Figure 1.11).

How can this occur? For this type of structure, very sophisticated control systems are applied,
and it should not be possible for any fault to arise. One type of check – which had been made
– involved, for example, the execution of several different independent FEM-analyses. These
were however based on the same theoretical documentation and a similar coarse mesh
division, and no real independent check was therefore involved. The calculations also showed
that it should be possible to use the unsuitable reinforcement design shown in Figure 8. With
a better understanding of the mode of action of reinforcement in a concrete structure, the
reinforcement should however have been given a much better anchoring. The problem is
actually so elementary that the whole situation could have been assessed with calculations
carried out on a paper-napkin, if one had built up a simple model of the type taught at every
Civil Engineering University.

1.8 Bridges which collapse – learning from history

Nearly always when one speaks of the development of the study of the strength of materials,
examples are taken from the sphere of bridges. This is natural; since bridges are spectacular
structures which if they collapse arouse interest and give rise to considerable anxiety. Today,
we experience this type of debate in the case of present-day accidents with car ferries. In the
earliest development of all types of structures, there were no theories to follow. One had to
use the “trial and error method”, i.e. one had to feel one’s way forward. If the building failed,
the master-builder, who at that time was often the same person as the designer, was replaced
and a new concept was tested. Thus, a bridge functioned as a kind of full-scale experiment in
the field of the strength of materials.
Today, bridge structures are among the technical systems where the safety is the greatest.
Even though accidents occasionally occur, the number of accidents should be compared with
the enormous number of bridges in the world, and the large amounts of traffic which pass
over these bridges. Compared with the risk of meeting with a traffic accident, the risk that the
bridge will collapse beneath the car is extremely small.

1.8.1 Bridges which shake down. The engineer must also learn from
history
The case of the bridge at Tacoma Narrows, which collapsed in 1940 after it had been set in
oscillation by a relatively moderate wind, is very well known. Many people have seen the
film of the collapse and have been surprised at the force of the wind, but those who designed
the bridge had not studied their history books properly, for this type of collapse had in fact

 25 
occurred several times before. Almost exactly 100 years earlier, for example, the Brighton
Chain Pier structure had collapsed in the same way. At that time, there were no film cameras,
but the wreck was documented by artists who described the course of events in simple
pictures.
Other suspension bridges which had been wrecked by the influence of the wind were the first
suspension bridge over the Menai Straits in Wales, which was damaged in 1854, and the
Clifton Bridge at Niagara, which was totally wrecked in 1889. One learned early that
dynamic processes can cause a bridge to collapse. The most serious known bridge accident
took place when a marching troop caused a suspension bridge, the Basse-Chaine in Angers in
France, to collapse and 226 soldiers drowned, as depicted in Figure 1.12. The accident is
considered to have been due to a combination of unfortunate circumstances, where both the
wind and the rhythm of the troop marching in step made the bridge vibrate to such an extent
that the rust-damaged structure broke.

Figure 1.12 The collapse of the Basse-Chaine-bridge in Angers in France in 1850.

1.8.2 Fatigue failure and other phenomena in metallic materials

Less known than the cases referred to above but equally instructive are the large number of
steel bridges that collapsed during the 19th century and some way into the 20th century.
When steel became increasingly less expensive, one started to build e.g. railway bridges in
this material in the middle of the 19th century. Steel was a new material which opened the
way to new construction principles, but the adoption of a new and untried technology led to
serious consequences. Approximately one bridge in four suffered serious damage or
collapsed. It was realized that it was vibrations from the trains which led to the collapse of the
bridge, but the concept of fatigue was unknown at that time.

 26 
A typical example of fatigue fracture is the collapse of the Point Pleasant Bridge in Ohio,
USA, in 1967. The case is interesting, since it teaches us that it is inappropriate to design a
structure so that the load-bearing ability is dependent on a single connection. The collapse,
which was the worst bridge disaster in modern times, led to the loss of 46 human lives.
The accident occurred in the rush-hour traffic on a cold day in 1967, after the bridge had been
in service without any problem for 40 years. The failure, which began in the suspension rod
shown in Figure 1.13, meant that the whole bridge (445 metres) including approach spans,
fell into the river. This was a typical fatigue fracture which can largely be referred to the steel
grade used when the bridge was built. An inappropriate structural design, with no extra
margin of safety, i.e. no redundancies, led to the total collapse.
A similar case where the I-35W Bridge over Mississippi River in US totally collapsed due to
one faulty designed joint between the steel members. 13 people were killed and many injured.

Figure 1.13 The figure, taken from the book Why buildings fall down, shows the structure
and bridge type for the Point Pleasant Bridge in Ohio, USA, which collapsed
in 1967. The figure is redrawn from Levy & Salvadori (1992).

1.8.3 The designer’s nightmare – the unforeseen load

An interesting example of an unforeseen load is the collapse of the first Tjörn Bridge in
Sweden. Its load-carrying structure consisted of two tubular arches which were anchored in
the rock a distance up on the land. In January 1980, the superstructure of a ship collided with
one of the arches, whereby the plate material in the arch buckled so that the arch immediately
lost almost all its load-bearing ability, and the bridge collapsed, as shown in Figure 1.14.
There were no people on the bridge and nobody was injured on the ship, but there was no
time to stop the traffic approaching the bridge, and several cars drove straight out into the air
and six motorists died.

 27 
Figure 1.14 The first Tjörn Bridge after having been hit by a ship. The picture shows the
steel tube arch structure and a part of the roadway resting on the boat deck.

The piers for bridges are normally dimensioned for being driven into by cars or by ships,
depending on whether the bridge goes over land or water. In this case, the piers were on the
land and there was thus no risk that a ship could hit a bridge support. Obviously, no one had
considered the risk that the superstructure of a ship might hit one of the arches. Concern for
this type of unforeseen accidental event has increased considerably in recent years. The new
Tjörn Bridge is a modern cable-stayed bridge with supports so far up on the land that the risk
of being hit by a ship is non-existent. Modern bridges which are built over traffic routes, such
as for example the Öresund Bridge between Sweden and Denmark, are designed so that they
can withstand impacts from even very large ships.

1.9 Safety during construction

In Sections 1.7 and 1.8, examples of failures have been discussed for already finished struc-
tures. It is however much more common with accidents during the construction phase.
Typically about 10 construction workers are every year killed during construction in Sweden.
This is a very high number considering that only about 10 persons has been killed due to
structural failures during the last 100 years. Most of these accidents are due to falling from
large heights or due to objects falling and injuries due objects falling. But a certain amount of
accidents have occurred during the construction process due to the collapse of shuttering and
scaffolding. The largest accident of this kind occurred when the centering support of the
Sandö bridge arches collapsed with the deaths of 17 workers on the day on which the Second
World War broke out in 1939.
It is not the intension to discuss these issues here, since these questions will be described in
manuals and text-books on the design of provisional structures and formwork.

 28 
1.10 Safety Regulations within the building sector
There are many regulations dealing with questions of safety within the building sector.
Table 1.2 presents a list of authorities in Sweden which have the formal responsibility within
a number of different fields. There is probably no sector in the society that is so regulated as
the building sector.

1.10.1 Swedish regulations and ordinances within the building sector

Table 1.2 shows some of the standards and regulations issued by Swedish authorities
regulating the building sector.

1.10.2 Planned joint regulations in Europe

Within EU, work is in progress to create joint regulations, so-called Eurocodes, for safety
within the construction sector. The intention is that they shall apply for all building work in
Europe. The reason for establishing joint regulations is of course the desire to increase the
opportunities for trade between the countries and thereby give lower building costs for the
consumers. The work on the regulations has been rather complex, since the different countries
have very different traditions and experiences. It will probably take a long time before the
same regulations apply over the whole of Europe.

 29 
Table 1.2 Examples of Swedish governmental authorities which publish regulations
regarding safety issues in the building sector. The table also shows examples
of the type of regulations issued.

General regulations regarding building and supervision

The National Board of Housing, Building Provisions regarding building structures
and Planning
(e.g. BBR, BKR)
-”- Elevator regulations
-”- Competence requirements for fire experts
-”- Provisions regarding type approval and
production control
The National Electrical Safety Board Provisions for electric installations
The Swedish Chemicals Inspectorate Provisions for dangerous substances
Activity-dependent regulations
The National Board of Occupational Safety Design of working premises.
and Health
The Swedish Board of Agriculture Requirements on structures in which animals
shall be kept and cared for
The Swedish National Food Administration Water quality
The National Swedish Police Board Safety requirements on wells, basins etc.
The Civil Aviation Administration Requirements regarding airports
Material responsibility which influences building and property administration
Banverket Railway tunnels and bridges
The Swedish Radiation Protection Authority Radiation emissions
The Swedish Road Administration Road tunnels and bridges

1.11 Concluding comments

1.11.1 What have we learned and how we shall avoid mistakes in the
future?

What can we learn from these examples from the history of technology? Obviously, it is
enormously important to check all the details in a design again and again, and constantly to
ask oneself whether anything will break and what the consequences then will be.

 30 
Reliability technology is needed to guarantee that the system has a safe function and a
sufficient lifetime. Here, we can only mention a couple of methods that would have
functioned excellently, for example, in the hotel building in Kansas City. A design exami-
nation with a check list would have directed the necessary attention to the different details in
the design. Failure Mode and Effects analysis (FMEA) is another method that is particularly
suitable for identifying the weak points in a design. The adoption of this latter method would
probably have been able to prevent the disaster with the Sleipner A oilrig. Both methods build
ultimately on the concept of ensuring that there is time for extra reflection and for respecting
common sense, and they also presuppose that staff with experience and competence are
engaged.

1.11.2 Questions of risk and safety are very important in all building
work
In this chapter, we have described different approaches to the question of safety within the
building sector. The factors which distinguish this field from many other fields of technology
are the long lifetimes for which one projects, and also the close interplay with nature and with
the forces of nature which influence everything which we build. In spite of the difficulties in
predicting the influence of nature, the safety level within the building sector is very high
compared with that in many other spheres. Some people consider that it is perhaps unneces-
sarily high, since the costs are also high. On the other hand, it has become customary to have
a very high safety level, since everybody wishes to live and reside in comfort and in safety.
In the development of products, mistakes are always made. This is also true and will be true
in the development of bridges. The greatest development in the art of bridge-building
th
occurred from the beginning of the 19 century when railway bridges began to be built. This
meant that new and much greater loads than had been experienced before now had to be borne
by bridges. In addition, the loads were dynamic, which created new problems and also new
knowledge of e.g. fatigue loads. At first, the problems were considerable and many bridges
collapsed or had to be repaired. There were indeed major disasters where there was
considerable loss of life. One of the well-known disasters was when a bridge over the Firth of
Tay in Scotland collapsed in 1879 when a train was on the span and 79 persons died. There
have also been several accidents during the building work. The most well-known is the
collapse of the first Quebec-bridge when 75 workers died.
The important feature of this is that knowledge derived from the accidents gave new under-
standing to be used in later projects.

This has however unfortunately not always been the case. An interesting example is the
collapse of the first suspension bridge over the Tacoma Narrows. This bridge, which had
stiffening beams with a low dead weight, large slenderness and low stiffness, collapsed due to
flutter in a fairly mild wind. If better attention had been given to history, it would have been
found that this phenomenon had been observed previously in several cases. The Brighton
Chain Pier had collapsed almost exactly 100 years earlier for the same reason.
In general, many steel structures collapsed during the period prior to ca. 1940, due to fatigue,
brittleness and corrosion. Since steel structures are often sensitive to a single fault, extensive

 31 
collapse occurred in many cases. Nowadays, thanks to a better understanding and to much
better steel grades, the collapse of steel structures is very rare.
Recently, considerable damage has been observed on concrete bridges due to corrosion.
Thanks to the very good “fail-safe” properties of concrete structures, this damage has seldom
led to any extensive collapse. “Fail-safe” means that the structure does not totally collapse
due to failure in a single section. The faults have, however, led to the need for extensive
repair and maintenance work.

 32 
 2. Reliability Analysis Methods
2.1 Preliminary remarks
Amongst other requirements structures have to exhibit the following most important features:
x Safety
x Serviceability
Both requirements must be guaranteed for some predefined period of time (durability) and
achieved by minimum cost (economy).
Similar demands are also placed on other technical systems.

Each requirement can as a rule be brought into the form of a so-called limit state condition,
which can generally be written as:

G (a0 , X1 , X 2 , ... X n ) t 0 (2-1)

The Xi represent the random variables2, which describe both the problem and the requirements for a
particular basis of assessment. Random variables are not always physical values like dimensions,
strengths, loads, etc., but could well be abstract values, such as the opinion of a group of
people about the admissible value of a beam deflection.

The so-called limit state equation separates the acceptable region from that which is characterised
as failure:

G (a0 , X1 , X 2 , ... X n ) 0 (2-2)

Failure is defined by the failure condition as:

G (a0 , X1 , X 2 , ... X n )  0 (2-3)

Of interest in the present connection is the probability of failure pf. This can be written as follows:

pf P >G (a0 , X1 , X 2 , ... X n )  0@ (2-4)

The methods for determining these probabilities or the corresponding indices are classified
according to their degree of complexity (so-called “levels of sophistication”):

Level I: The variables Xi are introduced by one single value only, e.g., by the mean value
or some other characteristic value. This is the level of the older codes. Statements about

2 In this chapter statistical methods are discussed. Notations and calculation aids for some of the issues treated in
this chapter can be found in the Appendix, Sections 6.2 to 6.5.

 33 
 the probability of failure are not possible. The committees responsible for writing the
codes carry to a certain extent the responsibility that the failure probability is acceptably
small.

Level II: The variables Xi are introduced by two moments or parameters, e.g., by the
mean value and the standard deviation. Codes could be constructed in this way.
Statements about the probability of failure obtained on this basis have a nominal cha-
racter and can only be used for comparison purposes. Such statements should not be used
outside the considered context.

Level III: If the variables Xi are introduced using suitable distribution functions, then the
results so obtained can, given the adequacy of the input values, lead to results which, for
all but rather small probabilities, can be used in an extended context.

The peculiar difficulty of certain problems in civil engineering lies in the fact that often
one has to deal with values that are far away from the mean value. In these areas the
probability densities are very small and the obtained results are very dependent on the shape
of the so-called “tail” of the distributions.

Besides, one must remember that pf is a subjective probability. Basically, it is a question of

the degree of confidence in the statement that what has been assessed could fail. As discussed
in Section 1.4, this subjective probability is not an inherent property — e.g., of a bridge —
but depends very much on the amount of information available to the person who makes the
assessment. Written formally pf is a conditional probability, dependent on the state of know-
ledge of the person doing the assessment.

pf P ¬ªG (a0 , X1 , X 2 , ... X n )  0 info ¼º (2-5)

There are two further limitations to be mentioned here:

x It is assumed in the following that the variables in a limit state function are inde-
pendent of each other. Correlations between variables are difficult to determine and
considerably complicate the algorithms. This limitation is acceptable, since, if there is
uncertainty, both extreme cases — complete correlation and no correlation at all —
can be analysed separately, compared, and the differences in the results assessed.
Some of the more sophisticated computer programs, however, allow for correlations.
x Human error does not enter this kind of analysis. Failure probabilities pf discussed
here are conditional to the assumption that there are no errors in what is analysed. For
reducing errors, special strategies such as the Quality Assurance measures should be
applied.
This chapter introduces the basic methods that are available to calculate, with sufficient accu-
racy, probabilities of failure pf under the given assumptions.

 34 
2.2 The Monte-Carlo method
No method is as easy to understand and as readily compatible to engineering thinking as the
Monte-Carlo method. And, if a powerful computer and a suitable program are at hand, none is
as adaptable and accurate as this method, although only in the last decade with the emergence
of powerful computers has it found increasing application.
With the Monte-Carlo method (or Monte-Carlo simulation) the exact or approximate calcu-
lation of the probability density and of the parameters of an arbitrary limit state function of
variables

G G (a0 , X1, X 2 , ... X i ... X n ) (2-6)

is replaced by statistically analysing a large number of individual evaluations of the function

using random realisations xik of the underlying distributions Xi. The index “k” stands for the
“k”-th simulation (k = 1, 2 ... z) of a set of xi . Each set of the k realisations gives a value

gk G (a0 , x1k , x2 k , ...xik ... xnk ) (2-7)

The resulting z numbers g k are evaluated statistically according to the rules given in
textbooks on statistics.
The heart of the method is a random number generator that produces random numbers aik
between 0 and 1. Such a number is interpreted as a value of the cumulative distribution func-
tion FX i ( xi ) and delivers the associated realisation xik of the variable Xi, see Figure 2.1.

1,0
FXi (x i )

0,9
0,8
0,7
0,6
0,5
0,4
0,3
0,2
0,1
0,0
xi

Figure 2.1 Transformation of rectangular distributed random numbers [0,1] into a cumu-
lative distribution function FX i ( xi ) .

 35 
Now, the number z0 of failures, i.e., the number of all realisations for which gk < 0, is coun-
ted. Thereby pf can be calculated according to the frequency definition of probability as:

z0
pf | (2-8)
z

In this expression z is the total number of all realisations of G. The greater the number of z0,
the more reliable is the value of p f. This becomes clear when looking at the coefficient of
variation of the probability of failure pf. This coefficient, for small pf, can be written as

1
v pf | (2-9)
z ˜ pf

If a small coefficient of variation is required – e.g., 10 % – then for probabilities of failure of,
e.g., pf 104 as many as z = 10 simulations have to be produced. For such large numbers
6

and for more complicated limit state functions even fast computers give the impression of
being slow.

In addition to counting z0 and z, gk could be analysed statistically according to basic statistical

methods, by determining the mean value mG and the standard deviation sG (and, if of interest,
higher moments too). From these two values the safety index E (described in detail later) can
be determined, and from it an estimated value for the probability of failure pf:

mG
E| ; pf | ) u E (2-10)
sG

As can be seen, in this estimate it is assumed that the density of G is normally distributed.
This estimate is quite good even for relatively small values of z.
In some computer programs the resulting values gk are continuously presented in a histo-
gram, thus giving immediately an idea of the probability density of the variable G. Other
methods to be discussed later do not offer this possibility.
In order to reduce the computational effort, a number of so-called “Importance Sampling”
methods has been developed, in which the realisations gk can be focused on the failure region,
that is the area of the limit state function where failure is most probable. It is beyond the
scope of this work to go into greater detail on these methods.

2.3 The problem G = R  S

2.3.1 Introduction and example
From the beginning of attempts to solve probability problems, the Monte-Carlo method has
been considered, at least intuitively. In practice, however, only in the last 20 years or so has

 36 
its application become feasible. Before the last decade alternative methods were developed,
which, though in some cases less accurate, still today prove to be very powerful and state-of-
the-art.
Certain names are attached to the development of such methods. Among the first was Max
Mayer who formulated his ideas on this topic in 1926 in a book which even today is inte-
resting to read. A. M. Freudenthal took up this question in 1947. Julio Ferry-Borges has
played an important role since 1966 in the further development of both the theory and
solution methods, especially during his long and fruitful time as President of the JCSS, the
Joint Committee on Structural Safety founded by the international associations in the field of
structural engineering (CIB, ECCS, fib, IABSE, IASS, and RILEM). Other key contributions
are to be found in the reference list.
The development of such alternate methods is illustrated and discussed in detail here using, in
terms of examples, the limit state function G = R  S.
In almost all examples in the following, reference will be made to the following variables
(Example 2-1), which can be regarded as bending moments in the critical section of a beam.
These variables are either considered as normally distributed values Table 2.1 a) or as rect-
angular distributions Table 2.1 b) also Figure 2.3.

Example 2-13
Input values for this and more examples in the following are presented in Table 2.1.

Xi Type P Xi V Xi Xi Type aXi bX i

S Normal 90 30 S Rectangular 38,1 141,9

R Normal 150 20 R Rectangular 115,4 184,6

Table 2.1 a) Input values for a normal b) Input values for a rectangular distribu-
distribution of the two functions S tion of the two functions S and R. The values
and R. are chosen to give the same average and
variation values as the normal distribution.

With the aid of these numbers, the fractile values and the conventional deterministic safety
factors can be determined.
From the mean values of both variables the “central” safety factor J C can then be derived
as:

PR 150
JC 1, 67
PS 90

The so-called safety margin amounts to:

3 This example is taken from Schneider (1997).

 37 
 m PR  PS 150  90 60

Usually, without looking at the type of distribution, the 5 % and 95 % fractiles are calcu-
lated. These are given for a normal distribution by the following expression:

x95% or x5% μ r 1, 65V

Numerically, this results in

r5% 150  1, 65 ˜ 20 117, 0

s95% 90  1, 65 ˜ 30 139,5

and thus to the nominal safety factor of

r5% 117, 0
JN 0,84
s95% 139,5

Obviously, this is unacceptable, as safety factors are deemed to be greater than one. Looking
at the rectangular distributions, the 5 % and the 95 % fractile values are 118,9 and 136,7, re-
spectively. The nominal safety factor, therefore, remains practically unchanged.
These are numbers that attempt to describe safety. But is difficult to draw conclusions about
whether the situation is acceptable. Much more meaningful are failure probabilities, the deri-
vation of which will be discussed below.

2.3.2 The classical solution

The variables R and S in the limit state function G R  S are shown, with their respective
probability density functions, in Figure 2.2.

Figure 2.2 Calculating the probability pf P( R  S  0) .

 38 
In calculating the probability pf P( R  S  0) , firstly, the probability that R is smaller than
a given value x is of interest. This can be written according to basic statistical principles as the
value of the cumulative distribution function at the position x:

P ( R  x) FR ( x) (2-11)

The probability that S = x is obtained from the probability density function of S at x:

P(S x) f S ( x) (2-12)

The probability that both expressions (2-11) and (2-12) are valid, is since
p( A ˆ B) p( A) ˜ p (b) given as a product of these values.

Since x may take on any value between f and  f , integration results in:

f
pf ³f f S ( x) ˜ FR ( x)dx (2-13)

As can be readily seen, this is the same as

f
pf 1 ³ FS ( x) ˜ f R ( x)dx (2-14)
f

This integral, known as the “convolution integral”, looks simple (and formally it is), but can
only be solved in a closed form for certain simple cases. Computer programs provide different
methods of numerical integration.

Example 2-2

The formal integration is possible for the rectangular distributions of the numerical
Example 2-1 introduced above, see Figure 2.3. The following is valid4:

1
f S ( x) 9, 63 ˜103 ; ^115, 4 d x d 184, 6`
141,9  38,1

1
FR ( x) x  1, 67 14, 45 ˜103 x  1, 67; ^115, 4 d x d 184, 6`
184, 6  115, 4

4 Calculation aids are to be found in the Appendix, Section 6.4.

 39 
 fX(x)
R

S
fR(x)
fS(x)

r, s, x

100

150
50

200
FX(x)
Integration range
FS(x) FR(x)

r, s, x

Figure 2.3 Integration of probability

Using the formula (2-13) the integration is as follows:

2 º141,9
3 ª x
141,9
3 3 3
pf ³ 9, 63 ˜10 14, 45 ˜10 x  1, 67 dx 9, 63 ˜10 ˜14, 45 ˜10 « »
«¬ 2 »¼115,4

115,4

9, 63 ˜103 ˜1, 67 > x @115,4

141,9
0, 0483 4,8%

This is the probability with which G = R  S < 0. This value would definitely be too large for
structural safety problems. Here, values in the range 104 to 106 are usually observed. For
serviceability problems, however, this might be an acceptable order of magnitude.

2.3.3 Analysis of the function G = R – S using normal probability

density functions
To explain the methodology an example for analysis of the problem G = R  S will be
performed, first more general and then for the normal density distribution.
Figure 2.4 shows the problem with its variables R, S and M. In fact G is the so-called safety
margin M = R  S. As the sum of two variables this margin is also, of course, a variable and is
normally distributed if the variables R and S are normally distributed. In this case all the
variables may be introduced by their mean and their standard deviation only.
Using the rules for normally distributed functions the two first moments of M can be quickly
determined. They are

PM PR  PS (2-15)

 40 
 VM V R2  V S2 (2-16)

Figure 2.4 The three probability density functions R, S, and M= R – S.

From Figure 2.4 the so called safety index E can also be seen. It may be determined from the
following quotient

PM
E (2-17)
VM

Expressed in word: E shows how often the standard deviation of the random variable M may
be placed between zero and the mean value of M. The probability of failure is obviously the
same as the probability that M is smaller than zero, or pf P( M R  S  0) . Assuming
normally distributed variables R and S, the failure probability pf can be read from standard
normal distribution tables, see Appendix, Section 6.2.

pf )( E ) (2-18)

Of special interest are the so-called weighting factors D , which show with which weight the
corresponding variable participates in the value of the probability of failure. These can be
calculated from

VR ½
DR °
V R2  V S2 °
¾ (2-19)
VS °
DS
2 °
VR VS ¿
2

Leading to

 41 
 D R2  D S2 1 (2-20)
As a numerical example, the variables introduced in Example 2-1 are used. The numerical
analysis delivers the following:

PM 150  90 60
VM 202  302
60
E 1, 66
36,1
pf ) (1, 66) 0, 049 4,9 %
20
DR 0,555
202  302
30
DS 0,832
20  30
2 2

The results are exact for normally distributed variables. For rectangular distributions the
result is only approximate. From the D-values it can be seen that the variable S exerts a
greater influence on the result than the variable R, due to its larger coefficient of variation.
The procedure of Basler/Cornell can be easily developed further into a design condition. The
requirement is E t E 0 , whereby E0 is representing the ”safety level” prescribed in a code.
This normally lies, depending on the assumptions, in the range E 0 | 3 to 6 . Simply using
algebra the following is obtained:
P R  P S t E0V M
VR VS
t E0 V R  E0 VS
V R2  V S2 V R2  V S2
t E 0D RV R  E0D S V S

Ordering the terms by R and S leads to:

P R  D R E 0V R t P S  D S E 0V S
P R 1  D R E 0 v R t P S 1  D S E 0 vS

This condition may be abbreviated as follows:

r ts (2-21)

and states the well known and, thus, rather trivial condition that the design value r of the
resistance must be greater than the design value s of the action. The design values are the
co-ordinates of the so-called design point.

 42 
In principle, the expressions 1 r D X E 0v X are safety factors. These differ, however, from
traditional factors by their transparent structure. The expressions area function of the desired
reliability level E0; they respond to the size of the standard deviation V X or the coefficient
of variation v X of the respective variable X and are moderated by means of the weighting
factor DX.

2.3.4 Representation as a joint probability density

A different representation of the problem G = R – S is shown in Figure 2.5. The respective
two-dimensional joint probability density, is represented as a hump. Its volume is 1 and the
contours are concentric curves.

f R,S( r,s)

(r*,s*)

f R(r)

S
0
f S(s)

Figure 2.5 Representation of the problem G = R – S as a joint probability density func-

tion.

In Figure 2.5, R and S are plotted as marginal probability density functions on the r and s
axes. The limit state equation G = R – S = 0 separates the safe from the unsafe region,
dividing the hump volumetrically into two parts. The volume of the part cut away and
defined by s > r corresponds to the probability of failure. The design point ( r , s ) lies on this
straight line where the joint probability density is greatest: if failure occurs it is likely to be
there.

Example 2-3

Turning again to Example 2-2, it follows that the joint probability density function is box-
shaped, see Figure 2.6.

 43 
 r

fR(r) 200

150 R

S
100
S
R
=
M

50

50 100 150 s
fS(s)

Figure 2.6 Representation of the problem G = R – S for a rectangular probability density

function.

fX,Y is constant within the defined region. Its value, i.e., the height of the ”box”, can be
readily determined, since the volume of the joint pdf is equal to 1. Thus:

1 1 1
f X ,Y ˜ 139 ˜106
141,9  38,1 184, 6  115, 4 7183

The volume cut away and corresponding to the probability of failure pf can then be easily
calculated. It corresponds to the cut-away triangular area multiplied by fX,Y:

˜139 ˜106 ˜ 0,5 0, 0488

2
pf V 'S ˜ 'R ˜ 0,5 ˜ f X ,Y 141,9  115, 4

This result has already been obtained in Example 2-2.

2.3.5 The method of Hasofer and Lind

Following the procedures of Basler/Cornell for some more complex limit state functions,
Ditlevsen (1973) discovered that the results depend on how the function is formulated. This
was called the invariance problem. An important step in solving this problem was made in
1974 by Hasofer and Lind (Hasofer, 1974). They transformed the limit state function into the

 44 
so-called standard space. This transformation is shown here for the two variables R and S
only.

The random variables R and S are transformed and standardised into Ul and U2:

R  PR
U1 o R U1V R  P R (2-22a)
VR

S  PS
U2 o S U 2V S  P S (2-22b)
VS

Thus, the new variables have a mean value of 0 and a standard deviation of 1. The greatest
rise of the joint probability density hump discussed in the previous section then coincides
with the origin of the coordinates. Now that both marginal pdf’s are of the same (0, 1)
character, the figure is axially symmetric, and all contours (points of equal probability) are
concentric circles, see Figure 2.7.

u1

0
=
fU1 (u1 )
G u2
E
pf

u2

fU 2 (u2 )

Figure 2.7 Transforming the limit state function into standard space.

In the new coordinate system the straight line G = R  S no longer passes through the origin.
The line transforms into the following:

G RS
U1V R  P R  U 2V S  P S (2-23)
P R  P S  U1V R  U 2V S

 45 
The design point ª¬u1 , u2 º¼ still lies at the highest elevation of the joint pdf above the straight
line G = 0. Due to the axial symmetry of the hump, the distance from the design point to
the origin is equal to the distance marked with E, the so-called HL (Hasofer/Lind) safety
index. The further the straight line passes from the origin the greater is E and the smaller
is the cut-away volume and thus pf.

Example 2-4

A short example should illustrate the procedure. Using the same limit state function G and the
parameters of the variables R and S, the transformation results in the following:

R  150
U1 oR 20U1  150
20

S  90
U2 oS 30U 2  90
30

Substituting the limit state function G gives

G RS (20U1  150)  (30 U 2  90)

which then gives the straight line

20U1  30U 2  60 0

The distance of this straight line from the origin of co-ordinates can be quickly found by com-
paring this equation with the Hessian normal form for representing straight lines:

A˜ x  B ˜ y  C 0

Comparing the coefficients

A 20, B 30, C 60

are obtained. From this we get:

A 20
cos [1 0,555 o D1 corresponding to u1
A B
2 2
20  30
2 2

B 30
cos [ 2 0,832 o D 2 corresponding to u2
A B
2 2
20  30
2 2

 46 
 C 60
h E 1, 66
A2  B 2 202  302

Using the table in Appendix, the probability of failure pf is obtained:

pf = ) 1, 66 0, 049 4,9 %

This value is correct in the case of normally distributed variables R and S or Ul and U2,
respectively, and, for non-normally distributed variables, the value is a good approximation.
The co-ordinates of the design points are obtained as

u1 E ˜ D1 1, 66 ˜ 0,555 0,924

u2 E ˜D2 1, 66 ˜ (0,832) 1,384

or in the r-s coordinate system:

r 20u1  150 20 ˜ (0,924)  150 131,5

s 30u2  90 30 ˜1,384  90 131,5

All these numerical values have already been observed in the previous sections. The method
applied here seems to be more complicated and to bring no advantages. This impression,
however, is wrong. The Hasofer/Lind method actually permits extension to arbitrary limit
state functions and any kind of distribution types as will be shown in the following.

2.4 Extensions of the Hasofer/Lind method

What was presented in the previous section is strictly valid only for linear limit state functions
and for independent, normally distributed variables Xi. In all other cases the results are –
though often quite good – approximations.
The extensions now to be discussed concern the transitions from
x two variables to many variables
x linear limit state functions to non-linear functions
x normally distributed variables to any kind of distribution types.
Because it can account for all these extensions, the Hasofer/Lind method is state-of-the-art in
reliability analysis.

 47 
2.4.1 Linear limit state functions with many variables
A linear limit state function with several variables is generally written in the form:
n
G a0  ¦ ai ˜ X i (2-24)
i 1

A transformation to the u-space is not necessary but is often advantageous when

programming for the computer. The algorithm is just a simple extension of the two-
dimensional case (see Section 2.3.3) and proceeds as follows (no proof is given):
n
PG a0  ¦ ai ˜ Pi (2-25)
i 1

n
¦
2
VG ai ˜ σ i (2-26)
i 1

PG
E o pf )( E ) (2-27
VG

Vi
Di a (2-28)
VG i

xi Pi  D i ˜ E ˜ V i (2-29)

Example 2-5, Beam bending safety

Figure 2.8 Example 2-5, a beam carrying load from a slab in a residential building.

 48 
A beam in a residential house is typically loaded by three different types of loading e.g. fixed
dead load, movable and varying loads from furniture and movable variable load from people.
2
The dead load could be assumed to be 5 kN/m with a very small standard deviation
2 2
0,20 kN/m . The load from furniture etc. is assumed to be 0,30 kN/m with a deviation
2
0,14 kN/m . The load from people defined as the yearly maximum is assumed to have an
2 2
average value of 0,70 kN/m with a deviation of 0,35 kN/m . The beam with a span 6 m is
carrying the load from a 6 m strip. The bending resistance of the beam has an average value
of 270 kNm with an assumed deviation of 30 kNm. The measures are assumed to be determi-
nistic. The task is to calculate the central safety factor, the safety index, the failure proba-
bility, the weighting factors Di and the partial factors for the four variables involved in this
study!
The variables can be compiled and defined as follows, see Table 2.2:

X Xnom PX VX vX

Moment capacity RM/kNm 220 270 30 0,11

2
Dead weight qd/(kN/m ) 5,0 5,0 0,20 0,04
2
Load from furniture etc. qf/(kN/m ) 0,35 0,30 0,14 0,47
2
Load from people qp/(kN/m ) 0,80 0,70 0,35 0,50

Table 2.2 Example 2-5, input values.

The first column contains nominal values, as, e.g., could be taken from a code and could thus
also be defined as characteristic values. They could be defined as the 5 %-fractile for the
moment capacity and fixed values for the loads. Using these numbers, the traditional safety
factors – i.e., the nominal safety factor JN and the central safety factor JC – can be calculated.
The mid-span moment due to dead weight and imposed loads is:

L2 L2 L2
MS qd b  qf b  qpb
8 8 8

where L is the beam span and b is the width of the floor strip carried by the beam. To simplify
the common deterministic factor, A, describing the measures are defined as:

L2 62
A b 6 27 m3
8 8

And the external moment can thus be written

MS A qd  qf  qp

and

 49 
 220
JN 1,32
A 5, 0  0,35  0,8

270
JC 1, 67
A 5, 0  0,3  0, 7

The limit state equation can now be defined as:

L2
G ( R, qd , qf , qp ) R b qd  qf  qp R  27 qd  qf  qp
8

With the previous algorithm the calculation of the safety index and the failure probability is
very simple:

PG 1˜ 270  27 5  0,3  0, 7 270  162 108 kNm

2 2 2 2
V G2 1˜ 30  27 ˜ 0, 2  27 ˜ 0,14  27 ˜ 0,35
900  29,16  14, 29  89,3 1032,8 o V G 32,1

PG 108
E 3,36 o pf )(3,36) 0, 0004 0, 04 %
VG 32,1

The weighing factors are:

VR 30
DR 1 0,93
VG 32,1
V qd 0, 2
D qd 27 27 0,17
VG 32,1

V qf 0,14
D qf 27 27 0,12
VG 32,1

V qp 0,35
D qp 27 27 0, 29
VG 32,1

As may be seen R contribute to a much higher extent to the probability of failure, whereas
here the contribution from the loads is much smaller. It should also be noticed that the sign of
the D-values gives further information on the problem at hand: positive are those D -values
whose variables provide safety – i.e., act positively – while the negative sign indicates “hazar-
dous” variables.
The design values of the four variables are:

xR 270  0,93 ˜ 3,36 ˜ 30 176

 50 
 xqd 5, 0  (0,17) ˜ 3,36 ˜ 0, 20 5,11

xqf 0,30  (0,12) ˜ 3,36 ˜ 0,14 0,36

xqp 0, 70  (0, 29) ˜ 3,36 ˜ 0,35 1, 04

The results are exact for normally distributed random values. For other distributions of va-
riables they are quite good approximations.
Here is a place to briefly introduce the notion of partial factors to be discussed in Section
2.5.2. These factors relate design values to nominal (or characteristic) values fixed, for in-
stance in a code. The partial factors for the problem at hand are:

M nom 220
JR 1, 25
Pnom  D M EV M 176

Pqd  D qd EV qd 5,11
J qd 1, 02
qd,nom 5, 0

Pqf  D qf EV qf 0,36
J qf 1, 03
qf,nom 0,35

Pqp  D qp EV qp 1, 04
J qp 1,30
qp,nom 0,80

2.5 Deriving partial safety factors

2.5.1 Design formats
There are, traditionally, quite a number of design formats. One still prevailing in many areas
is the allowable stress format requiring that maximal stress in a structure be less than allow-
able values described and numerically fixed in codes and standards for different materials.
Written in the form of a safety condition, this format reads:

V allow t V max (2-30)

This format is being replaced now in many new codes by the so-called load/resistance-factor
(LRF), or split factor format. Here, two safety factors, applied to the resistance and to the load
side, respectively, serve to keep load effects well below what a structure can resist. The
format reads as follows:

R
I˜R t J SS (2-31)
JR

 51 
wherein J S and J R are load and resistance factors, respectively, and larger than 1, while,
alternatively, I 1 is a resistance reduction factor as included in e.g., North American
JR
codes.
The so-called partial factor format applies factors to all relevant design parameters, e.g., diffe-
rent factors for dead and live loads, different strengths, for instance, in the resistance of a
reinforced concrete beam. The format is best explained using the so-called design values xdi,
defined by multiplying characteristic values xki, by partial factors J i . Both values and factors
are defined for loads and strengths. The format reads

R(rdi ) t S ( sdi ) (2-32)

where R(...) and S(...) represent functions of design values, which are defined, as stated above,
by

sdi J Si ˜ ski ½
¾ (2-33)
rdi rki / J Ri ¿

Often, in addition, a model factor is introduced.

The probabilistic format advocated in this book reads

pf P (G G ( X i ) d 0) d Pf,adm )( E0 ) (2-34)

where E 0 is the target reliability index. This condition is equivalent to

G G ( xi ) ! 0 (2-35)

where,

xi μi  D i ˜ E0 ˜ V i (2-36)

are the design values derived for instance using the Hasofer/Lind method in its various exten-
sions as explained in Section 2.4. It is obvious from Eq. (2-34) and (2-35) that the differen-
tiation between load and resistance variables disappears, as well as does the need to allocate
variables to the left or to the right side of the safety condition.

2.5.2 Partial factors

The similarity between expressions (2-32) and (2-35) is evident. This similarity may be used
to derive partial safety factors. The following is valid:

xdi xi

J i ˜ xki Pi  D i ˜ E 0 ˜ V i

 52 
From the comparison, the partial factor relevant for the variable Xi can be derived as

Ji Pi  D i ˜ E 0 ˜ V i / xki (2-37)

Introducing the coefficient of variation vi V i / Pi into the above the expression (2-37) reads

Ji Pi 1  D i ˜ E 0 ˜ vi / xki (2-37´)

All important characteristics of the respective variable come into play in Eq. (2-38), e.g., the
mean μi of the variable, its coefficient of variation vi, a characteristic value xki of the variable
(freely chosen, or prescribed in a code), the relative importance of the variable within the
safety condition (expressed by Di), and, finally, the target reliability level expressed by E0.

The above is valid for normally distributed variables. As the D-values for “dangerous” variab-
les become negative, the term in brackets becomes larger than 1. The inverse is true for
“favourable” variables.
Since the D-values are dependent on the specific design situation under consideration, the
partial factors to be applied for a variable depend on that same situation. This fact is un-
acceptable as for practical reasons constant partial factors are a necessity.

2.5.3 Linearization
In order to arrive at simple design rules D-values are arbitrarily fixed to some values that, for
the more frequent design situations, are checked to be on the safe side. So, e.g., for the most
important resistance variable in a design situation, DR = 0,8 is often chosen, while for all other
resistance variables D Ri = 0,0 is chosen resulting in the mean (or nominal) values are consi-
dered. For the most important load variables, e.g. the leading variable in a hazard scenario,
likewise DS = 0,7 may be used. For all other load variables, rather small D-values are chosen.
Of course, the sum of all D i2 must be larger than 1 in order to be on the safe side.

2.5.4 Calculation of partial factors according to codes

As was discussed in Section 2.3.3 the safety format used in e.g. the Eurocodes where the E-
factor is given as E t E0 depending on the design situation as

P R  P S t E0V M (2-38)

or
VR VS
P R  PS t E0 V R  E0 VS E0D RV R  E0D SV S (2-38´)
V R2  V S2 V R2  V S2

Ordering the terms by R and S leads to:

 53 
 P R  D R E 0V R t P S  D S E 0V S
(2-39)
P R 1  D R E 0 v R t P S 1  D S E 0 vS

In practise there is a lack on information on many of the input values that must be used for
calculating the partial factors. As stated above the tradition is to use partial coefficients larger
than or equal to 1,0. This means that if the variable no i in Eq. (2-37) represents the resistance
the equation for calculating the partial coefficient for resistance reads:

P R 1  D R ˜ E 0 ˜ vR
J Rc which result in a partial factor  1 if D R ! 0 (2-40)
xkR

To see that the partial factor is larger than one, Eq. (2-40) is re-written

xkR
JR !1 (2-40´)
PR 1  D R ˜ E 0 ˜ vR

The equation for the action reads

1  D S ˜ E0 ˜ vS
JS PS ! 1 since D S  0 (2-41)
xkS

Eq. (2-32) can then be written

xkR
t J S xkS (2-42)
JR

xkR
Or using the notations xdR and J S xkS xdS , where the index “d” stands for “design”
JR
xdR t xdS (2-42´)

In the standards the characteristic load values for the main load should be the 98 % fractile of
the yearly maximum of the load or “the fifty year load” i.e. the maximum load occurring in
average every 50 year. The characteristic design resistance value shall be the lower 5 %
fractile of the resistance distribution.
In practise the characteristic values for loads and resistances are based on experience and
comparisons with older standards, but if enough input data are at hand, these basic input
values for the safety can be used.

2.5.5 Examples of calculation of input data

In the analysis of the safety of structures using the partial factor method input data on action,
resistance, geometry and design models are needed. Of these four parts building up the design
parameters, the actions is the most complicated part at least for advanced structures like
bridges and thus this part is discussed in a separate chapter, Chapter 3. Issues regarding the

 54 
resistance, geometry and uncertainties in methods for analysis are discussed on textbooks on
different materials, see e.g. Design of Reinforced Concrete, Design of Prestressed Concrete
Structures. Some examples will though be discussed here.

Variation in geometrical parameters

A structure and its different components can not be constructed exactly according to the
drawings. Tolerances always occurs. These tolerances are usually relatively larger for smaller
structures. Usually these tolerances are implemented in the standards for the different
materials giving different characteristic values for different size of the structure or its parts.
For steel structures the effect of the size, is in practice counteracted by the fact smaller size,
due to cold working during the rolling process increase the strength.
In cases where the variation in e.g. the area or moment of inertia of a structural component is
of importance the formulas in Appendix can be used. The Example 2-6 will show an example.

Example 2-6

The area A of a cross-section is the product of the lengths of two sides. If the sides are
designated b and h,

A b˜h (a)

If the side length have statistical means Pb and Ph with deviations V b and V h formulas (A-
24) and (A-25) in the Appendix can be applied since A, b and h are subject to random
variations and are thus usually uncorrelated stochastic variables

PA Pb ˜ Ph (b)

2 2 2
V A2 PbV h + P hV b + V bV h (c)

or, in terms of coefficients of variation

v A = V A /P A , vb = V b /Pb and vh = V h /Ph

v 2A = vb2 + vh2 + vb2 vh2 (d)

The final term in Eqs. (c) and (d) can usually be neglected.
v 2A = vb2 + vh2 (d’)

The case discussed in Example 2-6 can in most cases also be used for a more general relation-
ship such as:

f = f ( x , y , z...)

 55 
then, to a good approximation,

2 2 2
§ wf · § wf · § wf ·
V 2f | ¨ V x ¸  ¨ V y ¸  ¨ V z ¸  ...
© wx ¹ © wy ¹ © wz ¹

Other variations that especially for structures acted on by axial forces, are of importance are
variations in end eccentricity, initial curvature and initial slope. These factors are typically in-
troduced in technology standards for different materials.
An important factor influencing the safety of compressed structural members is the initial
slope for a single column or a system of columns stabilising a framework.

Example 2-7

An example how to take the initial slope of columns into consideration is presented in the
Swedish standard for concrete structures, BBK 04. With definitions according to Figure 2.9
and using notations according to this standard, the initial slope Ti is dependent on the number
of columns contribution in the stabilisation according to the following equations.

Figure 2.9 Definition of parameters for describing

geometrical initial slope of columns or
column systems.

Ti TiD hD m

Dh 2 / L ; but 2 / 3 d D h d 1 ; L in meter

Dm 0,5 1  1/ m ; m is the number of columns contribution in the stabilisation.

Examples of variation in strength values

In Figure 2.10 results from measured yield point stress values are presented for two different
sizes of reinforcement steel bars, diameter 8 mm and diameter 16 mm. The nominal yield
point stress for the samples was 600 MPa.

 56 
 30

%
) 16, n = 524, P = 657 MPa, V = 33 )
25
) 8, n = 40, P = 676 MPa, V = 26 )

20

15

10

0
560

580

600

620

640

660

680

700

720

740
Yield point stress (MPa)

Figure 2.10 Measured yield point stress for two series of tests of reinforcement bars.

The characteristic value to be used in the design is according to standards typically the lower
5 % fractile of the measured distribution. This gives for the two series the following values
fsy,k,)16 657  1, 65 ˜ 33 603MPa and

fsy,k,)8 676  1, 65 ˜ 26 633MPa

It must be observed that the characteristic values defined according to the equations above is
based on the assumption that the number of samples is large. According to the laws of
statistics fewer samples must be compensated by a larger standard variation to really get the
5 % fractile.

2.6 Idealizations of structural systems

Structural systems or their sub-systems may be idealized into two simple categories: series
and parallel. Some structural systems will consist of combinations of these and others still
will be more complex, containing conditional aspects as well.

2.6.1 Series systems

In a series system, typified by a chain, and also called a “weakest link” system, attainment of
any one element limit state constitutes failure of the structure, see Figure 2.11.

 57 
 Figure 2.11 A series system consisting of a chain with
3 links, where the failure of one chain con-
stitutes failure of the whole structure.

It is evident that a statically determinate structure is a series system since the failure of any
one of its members implies failure of the system. Each member is therefore a possible failure
mode. It follows that the system failure probability for a weakest link structure composed of
m members is

pf P Fs P F1 ‰ F2 ‰ F3 ‰ ... F j ... ‰ Fm (2-43)

In Eq. (2-43), the event “structural failure” Fs is the union of the failure of all m members,
where Fj is the event “failure of the j:th member”.

The reliability R of a series system is given by the probability that neither F1 nor Fj nor Fm,
nor any of its elements will fail. This probability is, if there is no statistical dependence of the
participating elements, given by

m
R 1  pf,1 ˜ 1  pf,2 ˜ ... ˜ 1  pf,m – 1  pf,j (2-44)
j 1

The probability of failure pf of a series system is given by the complement of the reliability of
the system

m m
pf 1  – 1  pf,j | ¦ pf,j (2-45)
j 1 j 1

The summation approximation is valid for small probabilities of failure pfj. Obviously, the
probability of failure pf of a series system increases with the number of elements and largely
depends on the probability of failure of its most unreliable element.
Figure 2.12 and Figure 2.13 show two examples of statically determined structures of the
series type. Figure 2.12 shows a truss structure with diagonals with smaller area than the

 58 
upper and lower cords. The forces in the diagonals are the same for all three members. The
strengths of the diagonals are denoted R1, R2 and R3. The structure will collapse if just one of
the diagonals fails.
Figure 2.13 is a so called Gerber beam system which is statically determinate and a hinge
formed in any of the three points R1, R2 and R3 due to yielding will lead to collapse of the
whole structural system.
Example 2-8 demonstrates an example of the analysis of the safety of a series structural
system.

R1 R2 R3

Figure 2.12 The figure shows a truss structure with diagonals with smaller area than the
upper and lower cords. The strengths of the diagonals are denoted R1, R2 and
R3.

F
R1 R3
A R2 C
L
D
L B L
L2

Figure 2.13 A so called Gerber beam system is statically determinate and a hinge formed
in any of the three points R1, R2 and R3 due to yielding will lead to collapse of
the whole structural system.

Example 2-8

Consider a chain shown in Figure 2.11 consisting of three links of strength R1, R2 and R3
described by their mean and standard deviation as (110;20), (115;25) and (105;15). The
applied load is assumed to be deterministic Q = 50.
For a given value of the applied load Q q , the probability that the capacity of the system
will be less than this value is given by

 59 
 Probability Q  q Probability ª¬ R1  q ‰ R2  q ‰ R3  q º¼

if the properties of the links in the chain are independent.

For a given value Q q the probability of failure then becomes

§ 50  110 · § 50  115 · § 50  105 ·

pf )¨ ¸ )¨ ¸)¨ ¸ ) 3  ) 2, 6  ) 3, 67
© 20 ¹ © 25 ¹ © 15 ¹
0, 00135  0, 00466  0, 000121 0, 0061

2.6.2 Parallel systems

When the elements in a structural system (or subsystem) behave in such a way or are so inter-
connected that reaching of the limit state in any one or more elements does not necessary
mean failure of the whole system, the system is said to be a “parallel” or a “redundant”
system.
A simple parallel system is depicted in Figure 2.14.

Figure 2.14 A simple parallel system with three

parallel members.

The failure of a parallel system requires that R1 and Rj and also Rm fail. Only when all ele-
ments fail does the system fail. This condition can if the elements are statistically indepen-
dent, using probability theory, be written:

m m
pf pf1 ˜ pf2 ˜ ... ˜ pfm – pf,j and the reliability R 1  – pf,j (2-46)
j 1 j 1

Analysing general parallel structural systems can be rather complicated as the force-defor-
mation relations often are different for the different structural members building up the struc-
ture. In parallel systems of low redundancy and with brittle elements, failure of one element
often is sufficient to cause failure of the system. Usually the failure of one of the elements
leads to overloading of other elements in sequence (causing so-called progressive collapse).
The situation is quite different if the members are ideal plastic. Such systems are often occur-
ring in practical cases.

 60 
Ideal plastic parallel systems

Ideal plastic parallel systems often occur in practical structural situations, since brittle
systems should be avoided. Brittle materials are seldom used for applications, but buckling í
even if the material involved is ductile í can lead to brittle behaviour of the system.
Examples of ideal plastic systems are continuous beams and frames built up by members that
have a theoretical moment rotation relationship according to Figure 2.15.

Figure 2.15 Moment-rotation curve for a structural member that, in the ultimate limit state,
can be treated as ideal plastic.

For typical cases shown in Figures 2.16 the collapse or failure mode (i.e. the ultimate limit
state) can be represented by an equation of the following type

¦ F jG j  ¦ M jT j 0 (2-47)
j j

where
Fj (j = 1, 2,…) are the external loads,
Gj are the deflections corresponding to Fj,
Mj (j = 1, 2,…) is the plastic moment resistance at section j = 1,2,… and
Tj is the plastic rotation at section j. (Tj is a function of the deflection Gj and the dimen-
sions of the structure).

Expression (2-47) is clearly of the parallel since each resistance Mj must be mobilized to
develop the total resistance against the loading Fj.

It may here be appropriate to discuss some definitions

1) Ultimate limit state is often defined as a case where fulfilment of conditions of type (2-
42) is necessary for all points of a structure. This does not always mean collapse of the
structure, because undetermined structures may need insufficient strength capacity in
more than one point as is shown in the following text.
2) There is no strict coupling between the number of elastically undetermined forces or
moments and the number of hinges needed for collapse.

 61 
 a)
b)

c)

Figure 2.16 Three examples of structures, which if the structural members involved are
ideal plastic can be treated as parallel ideal plastic systems:
a) Horizontal stabilizing system with three columns.
b) A portal frame 5 times structurally undetermined, for which four or under
certain circumstances only three hinges have to be developed for the collapse
of the system.
c) A continuous beam for which three hinges have to be developed at the
ultimate limit state, compare Figure 2.13.

The idealizations are discussed below in two simple examples.

Example 2-9

Consider a parallel member system in principle shown Figure 2.14. Let there be n members
with each member assumed to take an equal proportion of the load and each member assumed
to have rigid-plastic behaviour as shown in Figure 2.15. Let the plastic resistance Rj, j = 1,
2,… n, of each member be identical and be a normal distributed random variable with mean
value P j P and standard deviation V j V . The coefficient of variation for each member is
Vj V j /Pj V / P . For independent members, the total capacity of the system RS is then
given by
n
RS ¦ Rj (a)
j 1

and using Eqs. (A-17) and (A-18) it follows that the mean and variance are respectively
n
PS ¦Pj nP (b)
j 1

 62 
 n
V S2 ¦ V 2j nV 2 (c)
j 1

It follows that the standard deviation of the total capacity is given by

VS nV (d)

This result shows that the variance of the total system capacity increases with the number of
members if their strengths are independent. The total capacity also increases with increased
number of members. Hence another way of expressing this, is to rewrite Eq. (d), since V is the
same for all members, as

V
VS (e)
n

If the strengths of the members are not independent, such as if all members are taken from
one piece of homogeneous material Uij 1 in Eq. (A-19). In this case it is easily shown that
variance in Eq. (d) and the coefficient of variation in Eq. (e) becomes, respectively

VS nV (f)

VS V (g)

This result implies that there is no advantage in increasing the number of members.
In a case when the task is to hold the system capacity to a given mean value by building up
the capacity by n members with independent properties, the system standard deviation is
given by

V
VS (h)
n

This result and Eq. (e) show that the random variability of the overall capacity decreases as
the number of independent redundant members in parallel system increases.

Example 2-10

A gusset plate carrying the deterministic given force P is attached to the main structure by
four screws F1 … F4, see Figure 2.17. If the deformation capacity of the screws is large, the
system is of the parallel ideal plastic type.

 63 
 Figure 2.17 Example 2-10.

The mean values and standard deviations for the screws and the external force are compiled in
Table 2.3.

Variable Mean value Standard deviation

P/N PP 0
F1/N P1 V1
F2/N P2 V2
F3/N P3 V3
F4/N P4 V4

Table 2.3 Example 2-10, mean values and standard deviations for variables.

Typically the mean values and standard deviations for the individual members for this type of
structure are assumed to be the same, which implies that the formulas in Example 2-9 can be
used. It is also usually assumed that the strengths of the screws are dependant implying that
Eqs. (f) – (g) in Example 2-9 should be used.
However if the screws are assumed to come from 4 different manufacturers Eq. (d), Eq. (e)
and Eq. (h) can in theory be used. This will formally increase the safety of the system. This
methodology is however not allowed to be used for this kind of structure, but can maybe be
applied for other cases.

Example 2-11

The rigid-plastic portal frame shown in Figure 2.18 is three times statically indeterminate. If
the rotation deformation capacity of the members is large, four different ultimate limit state
collapse modes can occur with hinges formed at different locations according to Figure 2.19.

There are seven variables involved in this problem. The possible plastic moments M1, M2, M3
and M4 formed at points 1, 2, 3, and 4 respectively, the external loads H and V and the mea-

 64 
sure L. The measure L is supposed to be deterministic, while the other variables
X j ( M 1, M 2 , M 3 , M 4 , H ,V ) are assumed to be random. Let the properties of these variables
be as shown in Table 2.4.

Figure 2.18 Example 2-11, geometrical properties and external loads.

a)
b)

c)
d)

Figure 2.19 The four possible failure modes for the rigid-plastic frame shown in Figure
2.18.

 65 
 Variable P V
M1/Nm 1,0 0,15
M2/Nm 1,0 0,15
M3/Nm 1,0 0,15
M4/Nm 1,0 0,15
H/N 1,0 0,25
V/N 1,0 0,50

Table 2.4 Example 2-11, mean values and standard deviations for variables.

For each mode an equation of equilibrium can be written, since the size of the rigid-plastic
moments are given.

Mode a): +M2 +2M3 +M4 V·L =0 (a)

Mode b): +M1 +M2 +M4 H·L =0 (b)

Mode c): +M1 +2M3 +2M4 H·L V·L =0 (c)

Mode d): +M1 +2M2 +2M3 H·L V·L =0 (d)

Using the methodology presented in Section 2.4 a limit state function can be set up for each
of the four failure modes. Let us start with Mode a) and to simplify the value of L is assumed
to be 1 m.

G( X ) M 2  2M 3  M 4  V ˜1 (e)

so that

PG 1  2  1  1 3 (f)

and

2
V G2 0,152  2 ˜ 0,15  0,152  0,52 0,385 (g)

and thus

PG 3
Ea 4,83 (h)
VG 0,385

The failure probability can then be found in a table for the complementary standard normal
function i.e. Table A.1. For large E-values the approximate formula (A-3) is exact enough
considering the approximate input values.

 66 
 pf,a 6,9 ˜ 107 (i)

The other failure modes give values according to Table 2.5

PG V G2 E pf

7
Mode a): 3,0 0,385 4,83 6,9·10
8
Mode b): 2,0 0,130 5,55 1,5·10
5
Mode c): 3,0 0,515 4,18 1,5·10
12
Mode d): 5,0 0,515 6,97 1,6·10

Table 2.5 E-values and probabilities for the four failure modes.

The first order failure probability is, see Eq (2-46), bounded by

m m m
max( pf , j ) d pf d 1  – 1  pf,j | ¦ pf,j (j)
j 1 j 1 j 1

Inserting the values from Table 2.4 gives

1,5 ˜ 105 d pf d 1,6 ˜ 105 (k)

It is evident that modes b) and d) and even mode a) have negligible effect on the failure pro-
bability of the structure.

 67 
 The loads on bridges can be very large. The picture shows an arch bridge
loaded by three locomotives each with a mass of approximately 85000 kg.

 68 
 3. Loads
3.1 General about loads on bridges
3.1.1 Different load types
What is special with bridges compared to the situation for structures for houses and industries
is that bridges shall be able to carry high loads from heavy lorries. Another thing compared to
loads in houses is that the loads vary significantly in time and room. Another special feature is
that the loads on bridges have increased much over time due to the demands from the trans-
portation sector. This development is still going on due to increased demands for higher
transport capacity and greater speeds. This development is true both for road and especially
railway transport.
A key and special point in the design of bridges is thus, in the right way, to treat the traffic
loads and their special features.
An attempt at grouping the different types of loads actual for bridges is as follows.
A first division is due to the basic source of the load
x loads caused by the gravity field of the earth and
x other loads.
To the first group belong
x permanent loads,
x variable loads during the usage of the bridge and
x temporary loads.
Permanent loads include
x dead weight of the bridge structure,
x dead weight of complementary items belonging to the bridge, i.e. surfacing, railings
etc.
x earth pressure and
x water pressure.
The reason why it is needed to separate the dead weight of the structure itself from the dead
weight of other parts of the bridge is that the dead weight of the structure itself will probably
remain constant during the lifetime of the bridge, while complements like surfacing can be
changed.
The variable loads during the usage of the bridge include:
x loads from vehicles,
x train loads and
x loads emerging from people, cyclists and the like.

 69 
The temporary loads caused by the gravitation include natural loads such as
x snow load and
x flooding
and also various loads caused by construction or other temporary activities on the bridge.
Other loads not directly caused by the gravity field can be of different types. One classi-
fication can be
x secondary loads caused by the traffic,
x different loads caused by nature,
x forces caused by the structural behaviour of the bridge and
x temporary loads.
Secondary loads caused by the traffic include
x horizontal centrifugal forces due a horizontal curvature of the road or railway.
x braking or acceleration forces and also
x additional forces depending on the dynamic movement of the railway vehicles on the
rails.
In addition to the abovementioned ”normal” loads there are many other loads that could be
characterized as accidental loads. Depending on the requested level of safety, the list of
possible accidental loads can be very long. Some of this kind of accidental loads are discussed
in Sections 1.7 to 1.9.
Accidental loads usually prescribed in standards are e.g.
x impact loads caused by vehicles colliding with structures belonging to the bridge e.g.
forces due to ship collision on piers in water or
x forces due to derailment of trains.
For major projects it could be described that the bridge should be able to withstand
x crashing airplanes and
x sabotage.
Loads caused by nature that are not directly coupled to gravity are e.g.
x wind loads,
x ice pressure and
x forces due to earthquakes.
In addition there are various types of loads and actions caused by the structural behaviour of
the bridge, such as
x long-term effects due to
- creep and shrinkage of concrete,

 70 
 - secondary forces due to pre-stressing,
- uneven settlement of the foundation structures
x and short-term effects due to
- temperature variation over the whole structure or
- temperature gradients over the sections.
As can be understood, there is a long list of actions that must be met when designing a bridge
or other important structure.
In Figure 3.1 some of the most important loads have been compiled in a scheme, so that it is
possible to see the connection between the different actions on a bridge structure.

Figure 3.1 Schematic compilation of different actions on bridges.

In addition to these external loads and actions, consideration must be paid to environmental
actions, which in course of time will break down all construction materials. To be able to
analyse these issues, consideration must be paid both to time and the environment in which
the structures serve.

3.1.2 Load positions

The non-dynamic loads can be of two main types. Loads whose position in space are almost
constant and those which vary in space, e.g.:
x fixed loadings
x variable loadings
Examples of fixed loads are dead weight of structures, soil and also water pressure. In certain
cases loads that in practise can move freely in the design are considered to be fixed. An
example of this kind of load is snow load. When designing the bridge, the fixed loads are

 71 
assumed to be stationary in relation to the structure while the free loads can be distributed in
the most unfavourable position.

3.1.3 Load duration

Dynamic effects due to moving loads etc. is discussed in Section 3.2.2. In this section the
non-dynamic loads will be discussed.
With respect to their variation in time, the loads and actions are classified in the following
main categories
x Permanent actions (G), which vary so little or so slowly that this action can be
assumed to be constant in time.
x Variable actions (Q) consist of all other normally occurring loads and other actions.
Examples are snow load, loads due to furnishing and people and also traffic load. The
first two actions are natural loads while the other types are the loads for which the
house or the bridge is constructed (functional load).
x Accidental actions (Qa) that occurs extremely rarely and then in combination with
some type of accident or serious mistake.
Variable actions are classified into three levels:
x Characteristic value Qk, usually defined as the 98 % fractile for the statistical distri-
bution for the yearly maximum. This can also be expressed so that Qk is the value that
is on average exceeded more than once every 50’s year.
x Usual value for the load \Qk, where \ is a reduction factor and \ d 1. This value for
an action is used when different variable actions are combined.
x Long-term action \1Qk that describes the average value of the action over a long time.
This level of action is used when designing structures due to deformations in turn due
to creep, shrinkage and other long-term effects.
The principle for the different \-values is depicted in Figure 3.2.

Last

Qk
Qk

1 Qk
Tid
a
50

Figure 3.2 To compensate for different durations of the characteristic variable actions, the
loads are multiplied by reduction factors \.

 72 
3.2 Traffic loads
3.2.1 General
That which is special with bridge structures compared to building structures is that the bridges
should be able to carry movable, dynamic and large loads from heavy lorries, and in the case
of railway bridges the loads can be really large.
The design loads for bridges are of course the same as for the road or railway for which the
bridge is an essential part. Often the bridges constitutes the limit for the maximum capacity of
the road or railway. For a society it is thus of great importance that the load capacity of the
bridges are as high as possible because a large capacity gives a more effective and also a more
environment-friendly transport system.
One must observe that the loads allowed for heavy traffic are not the same as the design
loads. The road administrations must take into account that there may be overloads, dynamic
effects and of course also some effects taking into account the statistical variation of the
actions. The real loads on the bridges can be measured by e.g. Bridge-In-Motion technique.
Measurements show that overloads are rather common, which implies that the design loads
must be larger than the allowed loads. These factors will shortly be described in the following
sections.

3.2.2 Dynamic load effects

The traffic on the bridge causes different types of dynamic effects, since the vehicles usually
moves fast along the bridges. The most important dynamical effects are fatigue and what is
called the “dynamic load factor” or defined in an other way “dynamic load allowance”. (The
difference between these definitions are shown in the following.)
Fatigue is discussed in the steel courses and is not discussed in this textbook, since the effect
of fatigue is most important for metallic materials, even if fatigue is of some importance even
for other materials.
The dynamic load factors are due to the dynamic interaction between the vehicle and the
structure. To be able to analyse this interaction, knowledge of at least the following factors
are needed
x the vehicle speed, the spring and damping properties of the vehicle and also
x the elastic and damping properties, the vertical and horizontal curvature and also the
smoothness of the bridge.
Figure 3.3 shows the forces acting on a vehicle during its passage over a structure. Due to
different forces of inertia it generally holds that

¦ Pi (t ) z mg (3-1)

 73 
 mg

P1 P2 P3 P4

Figure 3.3 Forces acting on a vehicle during its movement on the bridge.

The effects that influence the difference between the forces ¦ Pi (t ) acting on the bridge and
the weight mg of the vehicle are
x the vertical road curvature that can give rise to centrifugal forces. These forces can in
turn depend on
- initial curvature due to the vertical road alignment or
- curvature arising during the action of loads according to the formula
R 1 / y cc  EI / M . Strictly this is only valid for the part of the vehicle mass that
is not supported on the vehicle spring system, since the movement of the centre of
gravity of the vehicle is depended on this spring system.
x Inertia forces depending on the vertical movement of the vehicle and the bridge. These
forces are dependent on both the vehicle spring system and the deformation properties
of the bridge. These inertia forces are also dependant on the unevenness of the road
surface, or in the railway case, on the unevenness of the rails.
In practise the abovementioned very complicated factors can be summarized by the simplified
principle

(¦ Pi )design J ˜ mg (3-2)

(¦ Pi )design 1  D ˜ mg (3-3)

J = dynamic load factor

D = dynamic load allowance
Normally 1  J  2 , i.e. 0 d D  1 . There is theoretically no upper limit for the dynamic load
factor, but standards usually present values J = 1,05 à 1,6.
In the textbook “Svängningar, utmattning och vågutbredning för infrastrukturkonstruktioner”
a theoretical formula is derived for the dynamic amplification factor for a simply supported
beam.

 74 
 1
J (3-4)
v
1
2 f1L

In Eq. (3-4) v is the speed of the vehicle, f1 is the lowest natural frequency of the beam and L
is the bridge span.
This formula is valid under certain simplified conditions and e.g. it does not take many im-
portant factors like the bridge unevenness into consideration.
For road bridges the dynamic amplification factor is included in the standard for designing
new bridges, Bro 2004, but is given in the special rules for classifying existing bridges.
In the international standard for railway bridges UIC 71, the dynamic load coefficient is given
as:

1,44
J 0,82  but always 1, 00 d J d 1, 67 (3-5)
Lbest  0,2

In the Swedish standard for railway bridges the dynamic load amplification factor is

4
J 1, 0  (3-6)
8  Lbest

In Eqs. (3-5) and (3-6) Lbest is a notional length defined in the standard. For a simply sup-
ported beam Lbest is the span length of the studied bridge. Both Eqs. (3-5) and (3-6) gives
smaller amplification factors for longer spans which is theoretically right, but otherwise this
kind of formulas must be considered to be very formal and approximate.
For railway bridges, that are to be used for high speed trains, the dynamic effects should,
according to the standards be analyzed using the real dynamic properties of the bridge and the
trains.

3.3 Traffic loads on roads and road bridges

3.3.1 Allowed loads on roads
In each country the authorities issues rules for allowed total weights, axle weights, lengths,
widths etc for vehicles and vehicle combinations (lorry + trailer).
Figure 3.4 show examples of allowed gross weight as a function of the total length of the
vehicle or vehicles. As can be seen the largest allowed total weights are higher in the Nordic
countries compared to the ones allowed in the rest of Europe. The reason for this is the need
for using the roads for heavy timber transport in the Nordic countries.

 75 
In US the allowed gross weights and also axle weights are lower than what is allowed in
Europe. In all countries special transports with high total and axle loads are allowed but with
special restrictions e.g. regarding speed and the lateral position of the vehicles on the bridge.

70
Austria
60
Belgium, Germany, France..
Max total weight /ton

50 Switzerland
40 Sweden

30 Denmark

20 Finland

Great britain
10
Norway
0
0 5 10 15 20 25 30
Total length of vehicle + ev. trailer /m

Figure 3.4 Examples of maximum allowed gross weight as a function on the total length of
the vehicle or the vehicle including trailer.

The allowed load in the Nordic countries are rather similar. The allowed loads as a function of
the vehicle length in Sweden are presented in Figure 3.4 and Figure 3.5. The vehicle lengths
are defined as the distance between the first and last axle in a vehicle or a vehicle combina-
tion.

 76 
 65,0

60,0
55,0
BK 1
Total weight /ton

50,0
45,0

40,0
35,0 BK 2
30,0
25,0

20,0
5,0 7,5 10,0 12,5 15,0 17,5 20,0

Distance between first and laste axle /m

Figure 3.5 Total allowed weight for vehicles on the Swedish road network as a function of
the distance between the first and last axle of the vehicle or the vehicle combi-
nation including trailer. Two different levels are presented. BK1 is for all new
roads and BK2 can be used for older roads with bridges that do not have
enough load carrying capacity.

65,0 4,00

Weight per meter /(ton/m)

60,0 3,50
55,0
Total weight /ton

3,00
50,0
2,50
45,0
2,00
40,0
1,50
35,0
30,0 BK 1 1,00

25,0 Load/m 0,50

20,0 0,00
5,0 10,0 15,0 20,0 25,0

Distance between first and last axle /m

Figure 3.6 Allowed weight per meter for vehicles on the Swedish road network as a
function of the distance between the first and last axle of the vehicle or the
vehicle combination including trailer.

 77 
3.3.2 Design characteristic loads for road bridges
The most important characteristic designing load in the Swedish bridge standard Bro 2004 is
the equivalent load depicted in Figure 3.7. The idea is not that this notional load combination
explicitly shall describe a real load, but using this notional load is intended to cover most of
the loads and load combinations on the road network. For short bridges or for the design of
the bridge slab and other bridge elements other loads shall be used, see Figure 3.8.
Both in the Swedish Standard and in Eurocode the loads are distributed over a notional lane
with width 3 m. The standards also reduce the design load for bridges with many lanes, see
Figure 3.9. The bridges should be designed for the number of lanes that can be accommo-
dated within the bridge width. A bridge with a free width of 8 m shall thus be designed for
two lane loads and a bridge with a width of 9 m for three lane loads.

A kN A kN A kN A kN A kN
2 2

3,0 m
p kN
m

a) Longitudinal direction b) Transverse direction

Figure 3.7 Equivalent lane load type 21.2221 according to Bro 2004. For
lane 1: A = 250 kN, p = 12 kN/m,
lane 2: A = 170 kN, p = 9 kN/m and for
lane 3: A = 0, p = 6 kN/m. See also Figure 3.9.

For designing e.g. the bridge slab or other smaller part of road bridges a check must be per-
formed for a single axle load according to Figure 3.8.

B kN B kN B kN
2 2

3,0 m

a) Longitudinal direction b) Transverse direction

Figure 3.8 Equivalent lane load type 21.2222 according to Bro 2004. For
lane 1 B = 310 kN,
lane 2 B = 210 kN.

 78 
 a) 3,0 m 3,0 m 3,0 m 3,0 m 3,0 m

A1/2 kN
A2 /2 kN p3 kN /m

p1 kN /m p2 kN /m
Carriageway with 4 lanes

Footway Shoulder Shoulder Footway

b) 3,0 m 3,0 m 3,0 m 3,0 m 3,0 m 3,0 m

A1/2 kN
A2 /2 kN
p3 kN /m

p1 kN /m p2 kN /m
Carriageway Central Carriageway
with 2 lanes reserv with 2 lanes
Shoulder Shoulder Shoulder Shoulder

Figure 3.9 For bridges with many lanes the figure shows the design load combination
according to Bro 2004, cf. Figure 3.7.

A kN A kN A kN A kN A kN A kN

p kN
m

Figure 3.10 Equivalent lane load type 21.2225 according to Bro 2004. For
lane 1 A = 250 kN,
lane 2 B = 170 kN.
This combination shall be used for longer bridges.

 79 
 150 kN 150 kN 180 kN 180 kN 75 75
kN kN
90 90

3,0 m

a) Longitudinal direction b) Transverse direction

Figure 3.11 Equivalent lane load for fatigue calculations type 21.2226 according to
Bro 2004.
The Eurocode presents the main design loading according to Figure 3.11. This nominal load
consists of just one bogie and a much greater distributed load. This loading is as in the
Swedish standard acting on a notional lane with a width of 3 m. The second and third lanes
have much smaller designing loads.

A ( kN )

p ( kN )
m

a) Longitudinal direction b) Transverse direction

Figure 3.12 Design load group type 1 according to Eurocode. For

lane 1 A = 300 kN, p = 27 kN/m,
lane 2 A = 200 kN, p = 7,5 kN/m and
lane 3 A = 100 kN, p = 7,5 kN/m.

In Figure 3.13 the load effect measured as the maximum moment on a simply supported
bridge with just one notional lane and varying span using the Eurocode load combination
(Figure 3.12) compared to the load combination according to Bro 2004 (Figure 3.7). In the
figure the design load effects are compared with the load effect from one vehicle with maxi-
mum allowed weight.

 80 
 16 000

14 000 Max allowed vehicle

Bro 2004
12 000
Eurocode
Moment/kNm

10 000

8 000

6 000

4 000

2 000

0
10 15 20 25 30 35 40 45 50
Span /m

Figure 3.13 Maximum moment on a simply supported bridge with just one notional lane
using Eurocode and the Swedish Standard Bro 2004. The design load effects
are compared with the load effect from one vehicle with maximum allowed
weight.

3.4 Loads on railways and railway bridges

3.4.1 General
A railway track is of course designed to be able to carry the train loads. The loading capacity
can be considered as some type of classification of the railway line. For the transport
company that is using the track, there is one set of rules for what is allowed and for the
structural design there is an other set of rules. Typically the design loads are higher than the
allowed loads.
A big difference between road and railway transport system is the much larger loads trans-
ported. A freight train for transporting iron ore could be built up with 50 wagons and the total
weight could typically be 5000 tons or more. Even a passenger train can be rather heavy. The
Swedish moderate high speed train X2000 consisting of 6 wagons has a gross weight of 318
tons.

3.4.2 Allowed loads on railways

For the companies using the railways, there are rules for the amount of loads that are allowed.
Typically the allowed load is divided into two categories:

 81 
 Stax5 = the maximum allowed axle load (mass) in ton and
Stvm = the maximum allowed load (mass) per meter track measured in ton/m.
Railway tracks are also classified for the maximum speed (sth) (km/h) that is allowed for each
line.

3.4.3 Designing traffic loads for railway bridges

From a design point of view most of the principles for the design of railway structures, don’t
differ so much from the design of road structures. There are however distinctive features that
are special when designing railway structures compared to the design of road structures.
Especially the following points need to be considered in the design of railway structures:
x The loads are higher.
x The dynamic allowance is relatively larger.
x For good riding comfort at high speeds the structures need to be stiffer.
x The load actions do not vary in the transversal direction as is the case for road vehic-
les.
x Especially for high speed train lines, both the vertical and horizontal alignment need to
be much more restrict. Typically railways are designed for a maximum vertical slope
in the range 1 % - 2 %, while roads can be designed for slopes 3 times as high. The
horizontal radius must also be much larger not to produce to high centrifugal forces at
high speeds.
The restrictions mentioned, result in higher construction costs for a railway track system com-
pared to roads, since the amount of bridges and tunnels are higher especially in a typical
Nordic landscape with mountains and valleys.
The cost for a railway bridge or a tunnel is typically 10 to 20 time as high per meter track
compared to the ordinary track.
Regarding the difference between the allowed loads and the structural design loads, the same
conditions as for road bridges is valid because
x dynamic effects produce higher loads than the dead weights of the train, see Section
3.2.2.
x and because overloads and other factors causing higher axle loads are not uncommon.
Figure 3.14 shows some of the design load system used for the design of railway bridges. The
loads depicted are the total load per track.

5 The abbreviations are the ones used in Sweden. Similar notations are used in other countries

 82 
a) 4·250 kN

80 kN/m 80 kN/m

0, 8 m 3·1,6 m 0, 8 m
6, 4 m

b) 4·330 kN

110 kN/m 110 kN/m

0, 8 m 3·1,6 m 0, 8 m
6, 4 m

c)

Figure 3.14 Examples of design loads for railway bridges. Load pattern a) is the common
European (UIC) load used for allowed axle load 22,5 tons. Load pattern b) is
the system used for new railways in Sweden. Type c) is a special load pattern
for heavy transport.

3.5 Comparison between prescribed load effect on a road

bridge compared to load effect on a railway bridge
Figure 3.14 shows a comparison between the load effect measured as the maximum moment
on a simply supported bridge with varying span acted on by one notional road lane or one
railway track.

 83 
 18 000 2,50

16 000
BRO 94 Typ1
14 000 BV UIC 71
BV/VV
Maxmoment/kNm

12 000

10 000
2,00
8 000

6 000

4 000

2 000

0 1,50
10

15

20

25

30

35
Balklängd/m

Figure 3.15 Comparison between the load effect measured as the maximum moment on a
simply supported bridge with varying span acted on by one notional road lane
with load pattern according to Figure 3.7, or one railway track with load
pattern according to Figure 3.13 a) (UIC 71). The quota between the two
cases is shown on the right hand axle.
From studying Figure 3.15 it is obvious that the railway design loads are much higher than
the design loads on road bridges, and the difference is still larger using the load pattern for
new bridges according to Figure 3.14 b).

3.6 Load combination

When different actions are combined the probability that all large loads are acting so that the
maximum effect occurs at the same time is very small. This is true for the probability that
they act at the same time as well as the probability that different loads can produce the same
maximum action in the same part of the structure. How to handle probabilities for combina-
tion of actions has been discussed in Chapter 2 and how the loads are acting in time is dis-
cussed in Section 3.1.3. In the standards these questions are summarized in Eq. (3-7).

n m
R
t ¦ J Gi Gi  J Q1 Qk,1  ¦\ Q j J Q j Qk,j (3-7)
JR i 1 j 1

 84 
In Eq. (3-7) Gi and Qj are permanent and variable loads respectively and R is the resistance.
J G , J Q and J R are partial coefficients and index k denotes characteristic values. The \ Q j fac-
tors are load reduction parameters because the probability that many variable loads are acting
at the same time is small. Often in standards the combined values \J are given for the diffe-
rent actions.
In practice many of the partial factors in codes are obtained calibrated using comparison be-
tween older Level 1 methods by this Level 2 method, see Section 2.1. However designers try
to calibrate the reliability level using the type of statistical tools discussed in Chapter 2. The
target for this calibration is using the target safety index E for different consequences of fai-
lure according to Table 3.1. The failure probabilities and the target safety indices in Table
3.1 are based on a reference period of one year.

Safety class Consequence of failure Target safety index E Formal probability

of failure
6
3. Very serious Major risk for serious 4,8 10
personal damage
5
2. Serious Some risk for serious 4,4 10
personal damage
4
1. Less serious Little risk for serious per- 3,7 10
sonal damage

Table 3.1 Classification of the safety classes according to the Swedish Standard BKR
(2003).

For road and railway bridges in Sweden the target safety index E is set to 4,8.
Both the Swedish Road Administration and the Swedish Railway Administration are using a
set of load combinations in principle according to Table 3.2.

 85 
Load Combination Control issue
I Serviceability limit check at the construction stage.
II Ultimate limit check at the construction stage.
III For calculating the super elevation of the bridge.
IV:A Main loading case in the ultimate limit state. In this loading case only
one load shall have a large partial coefficient. Only four variable
loads shall be considered simultaneously. The four loads giving the
most unfavourable effect shall be considered.
IV:B Loading case for calculation of the influence of dominating perma-
nent loads in the ultimate limit case.
V:A Main loading case in the serviceability limit case.
V:B Loading case for analysing crack widths. It can also be described as
the design loading case for permanent and long-term duration loads.
V:C Loading case for analysing deflections and horizontal movements due
to traffic loads.
VI Loading case for analysing fatigue.
VII Loading case for analysing vibrations.
VIII Loading case for analyzing accidental loads, one calculation for each
type of accidental loads.
IX Loading case for analyzing mechanical equipment for moveable
bridges.

Table 3.2 Load combinations used for bridges in Sweden according to Bro 2004.

The most important design load combination factors used for road bridges according to
Bro 2004 are presented in Table 3.3. The most serious load combination must always be
checked for all main structural members.

 86 
Load combination Load comb III: Load comb. Load comb. VB: Load comb VC: Load comb VI:
Pre-cambering IV:A: Ultimate Serviceability Serviceability Fatigue limit
limit state limit state (crack limit state state
width) (deformations)
Permanent loads
Dead weight from 1,0 1,0/0,9 1,0 1,0
concrete and steel
Surfacing 1,0 1,0/0,8 1,0 1,0
Soil pressure 1,0 1,1/0,9 1,0 1,0

Differential settle- 1,0/0 1,0/0

ments
Shrinkage 1,0 1,0/0 1,0/0
Variable loads
Traffic (axis + 0,7/1,5 0,8 1,0
lanes)
Pedestrians etc 0,7/1,5 1,0
Wind 0,6/1,3 1,0
Breaking forces 0,7/1,5

Table 3.3 Roadway load factors J\ for some of the most important load combinations.
When two factors are given, the combination that gives the greatest action
should be used.

 87 
Heavy vehicles can move in different positions in the transverse
direction of a bridge, which results in different load actions on
the main beams.

 88 
 4. Load distribution
4.1 General
One of the most important tasks for a bridge engineer is to design the bridge superstructures
so that they can distribute the large concentrated wheel and axle loads to the different main
load bearing systems of the bridge. Different issues in this field are discussed in other text-
books in the series of compendia issued from the div. of Structural Design and Bridges. It is
recommended to study the following textbooks parallel to the study of this section:
x Torsion of concrete beams
x Theory of elasticity for analyzing bridge superstructures.

4.2 Concentrated loads

Since bridges usually are designed based on the theory of elasticity, which for the simplified
thin plate theory produces infinitely large moments close to the concentrated loads it is
important to take the load distribution, that occurs in practice, into account, see Figure 4.1.

Figure 4.1 Under the train and heavy

truck wheels, large con-
centrated pressure arises.
These pressures are by
different structural systems
distributed out to the main
load-bearing structure and
down to the foundation.

The in practice usable load distribution for heavy truck tyres be attributed to three factors
x The distribution of the loads due to the deformation of the tyres (the contact area
between the tyres and the road surface), see Figure 4.2. The contact area is dependant
on the pressure and the stiffness in the tyre and also on the friction between the tyre
and the road surface in the case of a moving vehicle. In standards the contact area
under the design wheel load is often given as conservative values.
x The load distribution due to the distribution in the pavement structure (asphalt, con-
crete protection layers, insulation etc.)
x The distribution of loads that occurs in the deck structure itself. This is mainly due to
the different structural behaviour of plates analyzed by the thin plate theory (Kirchoff
theory) and the thick plate theory (Mindlin theory). The thin plate theory produces

 89 
 infinitely high moments under concentrated loads which of course is not realistic. The
Mindlin theory solves this problem, but is complicated to use in practise. Standards re-
commend thus simplified methods for adequate load distribution under concentrated
loads. Two such methods taking the thickness of the concrete slab and the pavement
into account are described in Figure 4.3.

Rolling direction

Figure 4.2 The distribution of the loads under a heavy truck tyre is dependant on the
pressure, stiffness and movement of the tyre.

2 1
b
p

o o o
45 45 45
2, eff 1, eff

Principle according to Eurocode: d 2,eff d 2  2hb  hp ; d1,eff d1  2hb  hp

Principle according to Bro 2004: d 2,eff d 2  2hb ;d1,eff d1  2hb

Figure 4.3 Principles for the load distribution in a deck structure according to the current
Swedish standard and the Eurocode. According to Eurocode and many other
standards one can utilize the load distribution in the deck slab, while this effect
is disregarded in the Swedish tradition.

 90 
4.3 Concentrated load on a cantilever plate
When calculating the clamping moment caused by a concentrated load on a cantilever plate
consideration must be paid to the load distribution ability in the plate and the distribution
effect an eventual co-operating edge beam can produce.
If the edge beam has a large stiffness the load distribution caused by this structural member
can approximately be analyzed by a “classical method” used in Sweden for a long time. The
method is presented in Wästlund (1964).
In this section this model will be presented because it has some pedagogical value since it
shows how the distribution of forces and moments between different interacting elastic struc-
tural members depend on their relative stiffness. A theoretically more stringent theory is
derived in the textbook Elastic plate theory for bridge superstructures, section 3.5.
The structural model that will be analyzed is shown in Figure 4.4. The following assumptions
and approximations are applied:
x The plate is assumed to be built up by a series of beam strips with a bending stiffness
Ei, where the moment of inertia when bent in the y-direction per unit width is i with
unit, m4/m. The torsional stiffness and the bending stiffness of the plate is thus neglec-
ted in this model.
x The edge beam is assumed to have a bending stiffness EI, where I is the moment of
inertia of the edge beam.
x The torsional stiffness of the edge beam is neglected.
The analytical model received in this way, is in practice the classical ”beam on elastic foun-
dation” problem where the elastic foundation in this case is formed by the plate strips.

Figure 4.4 For the approximate analysis of the load distribution effect in a cantilever plate
due to an edge beam, a model according to the figure is proposed.

 91 
The model presupposes that the concentrated load P is acting on the edge beam and that a di-
stributed load px [N/m] is acing between the edge beam and the plate. Using the notations in
Figure 4.4 the following equation for the cantilever beam is obtained

3k t Ei
px y( x) Fy (4-1)
a3

The factor k t in eq. (4-1) is intended to adjust the stiffness for the case when the cantilever
have varying height and thus varying stiffness. This will be explained further below.

The factor 3k t Ei / a3 , with unit [N/m2] is given the notation F . The equation for the elastic
line for the edge beam is

d4 y
EI  px
dx 4

With the value for the distributed load p x inserted, this equation can be written:

d4 y
4  4O y
4
0 (4-2)
dx

where

3 k ti
O4 (4-3)
4 a3I

The solution for this forth order differential equation can be written:

y eOx ( A cos Ox  B sin Ox )  e Ox (C cos Ox  D sin Ox ) (4-4)

It is obvious that the case presented in Figure 5.4 is symmetrical around the y-axle as well as
that y and yc must tend to zero for large numerical values for x. This implies that the con-
stants A and B must be zero and that yc 0 for x 0 . Inserting these relations and deriving
Eq. (4-4) gives

yc Oe  Ox (  (C  D) sin Ox  ( D  C ) cos Ox ) (4-5)

which in turn implies that C = D and thus

y Ce O x (cos O x  sin O x) (4-6)

An equation for vertical equilibrium gives

 92 
 f
P ³ p x dx (4-7)
f

If Eq. (4-1) is combined with eq. (4-6) and inserted in Eq. (4-7), an equation for P is obtained

f
O x
P FC ³e (cos O x  sin O x)dx
f

The integration of this equation results in the following relations

2 FC PO
P or C (4-8)
O 2F

Insertion of the C-value in Eq. (4-6) and knowing that the moment in the edge beam follows
the equation M x ( x )  EIy cc gives

O P O x
y ( x) e (cos O x  sin O x) (4-6’)
2F

P O x
M x ( x) e (cos O x  sin O x) (4-9)
4O

The clamping moment in the intersection between the cantilever plate and the main beam, see
Figure 4.4 can now using Eqs. (4-1) and (4-6’) be written

PaO  Ox
mi ( x ) apx aF y ( x ) e (cos Ox  sin Ox ) (4-10)
2

Both the moment in the edge beam and the clamping moment in the cantilever plate as their
absolute maximum values for x = 0.

M x,max 1 0,25
= for x 0 (4-11)
Pa 4O a Oa

mi,max Oa
for x = 0, and (4-12)
P 2
S
M x,min 1  0,052 S
 e 2= for x / a (4-13)
Pa 4O a Oa 2O a

In Figure 4.5 and Figure 4.6 respectively, the moment distribution in the edge beam and the
distribution of moment along the clamped end of the cantilever plate for varying values of Oa
are shown.

 93 
Eq. (4-12) can be interpreted so as the moment Pa [kNm] from the concentrated load P [kN]
is evenly distributed over the distance 2 / O .
In the case of two concentrated loads, e.g. a bogie load, close to each other at the distance bb ,
these two loads can approximately be assumed to be distributed over the distance bb  2 / O .
If the average moment is denoted mi,2 , this moment can be calculated using the following for-
mula

mi,2 2/O
2 >1 if bb  2 / O (5-14)
mi,1 bb  2 / O

and where mi,1 is the clamping moment for one concentrated load.

0,50

0,40
Oa
0,30
0,5
0,20 0,75
M x /Pa

1
0,10 2

0,00

-0,10

-0,20
0 0,5 1 1,5 2 x/a 2,5

Figure 4.5 The moment distribution along the edge beam belonging to a bridge cantilever
plate according to Figure 4.4 and the approximate theory presented in this
section, for some values of the stiffness parameter Oa.

 94 
 1,00

0,80
Oa
0,60
0,5
m i/P

1
0,40
2
0,20

0,00

-0,20
0 0,5 1 1,5 2 x/a 2,5

Figure 4.6 The moment distribution along the clamped edge of the cantilever plate accor-
ding to Figure 4.4 and the approximate theory presented in this section, for
some values of the stiffness parameter Oa.

The presented method is approximate, but demonstrates some typical principles on load distri-
bution even if the load distribution capacity of the plate in torsion and bending in the x-axle
direction is disregarded. One way to take this into account is to include a part of the plate
thickness in the stiffness of the edge beam, especially in the case when the concentrated load
acts a distance from the edge beam, as shown in Figure 4.7. When compared to a more
theoretically exact theory, see section 3.5 in the textbook Elastic Plate Theory for Bridge
Superstructures, this correction is not enough to take account of the total load distribution
from the plate, hence an effective stiffness ieff 3 ˜ i could be used when calculating the para-
meter Oa.

Figure 4.7 When the concentrated

load is acting at a dis-
tance from the edge beam,
I a part of the bending stiff-
min

ness of the plate in the x-

a axle direction can be in-
c cluded in the stiffness of
the edge beam.
If the height of the cantilever plate is varying, see Figure 4.7, this can be taken into conside-
ration by the coefficient k t . This factor can be calculated using ordinary beam theory (tip de-
flection for a cantilever beam with varying height) and the result is presented in Figure 4.8.
Eq. (4-12) can be reformulated according to Eq. (4-15)

 95 
 mi,max Oa a 4 3k ti a 4 3k t t 3 t 4 kta
(4-15)
P 2 2 4a 3 I 2 48a 3 I 4 tI

1,00
0,90
Figure 4.8 The figure shows the
0,80 factor k t to be used
kt

0,70 when calculating the

0,60
parameter F and O in
Eqs. (4-1) and (4-3) re-
0,50 spectively, as a func-
0,40 tion of the parameter
t/tmin, see Figure 4.7.
0,00

0,25

0,50

0,75

1,00

1,25

1,50
t /t min 1

0,50 0,50
I b/I s I b/I s
0,40 0,40 1
1
2
m i/P

m i/P

0,30 2 0,30
7,5
7,5
15
0,20 15 0,20

0,10 0,10
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1,0

0,3
0,4
0,5
0,6
0,7
0,8
0,9
1,0

a /c a /c

Figure 4.9 Clamping moment mi in a cantilever plate stiffened by an edge beam according
to Bakht & Jaeger (1985) (left) compared with the simplified theory presented
in this section (right). In these diagrams the notation I s at 3 / 12 has been
used.

Eq. (4-15) can be found in old Swedish standards and are often still used when designing
bridge superstructures even if the total load distribution capacity of the plate is not taken into
account, which is clearly demonstrated if I o 0 . As a comparison between the method shown
in this section and a more accurate theory, results presented in Bakht & Jaeger (1985) are
shown in Figure 4.9. In cases where the relative stiffness of the edge beam is small the differ-
rence between the two theories is large.
Of the results in Figure 4.9 by Bakht & Jaeger (1985) it is evident that an approximate value
for the clamping moment mi 0,3P can be a good approximation for many cases.

 96 
4.4 Load distribution from a concentrated load on a plate
structure
4.4.1 General
For analysis of the distribution of concentrated loads acting at a point on a plate structure to
other parts of the structure or to the supports, reference is made to textbooks on plate
structures or to the textbook Elastic Plate Theory for Bridge Superstructures.

4.4.2 The use of influence area

When concentrated loads are acting on a plate structure carried by beams in a typical girder
bridge according to Figure 4.10, an important design task is to calculate the distribution of
the load and thus finding the designing moment at the interface between the plate and the
longitudinal beams.
A conventional method is to use influence areas presented in e.g. Pucher (1951), see
Figure 4.12, and to distribute the moments according to Figure 4.11, Menn (1990).

Figure 4.10 For analyzing the moments in a plate structure acted on by concentrated forces
from e.g. a bogie to a lorry, a certain portion of the plate structure will assist
in distributing and carrying this load.

 97 
 0,35 a 0,3 a 0,35 a

Figure 4.11 Distribution of the moments in

mx two directions in the top slab
of a concrete girder bridge.
The distance between the
0,25 a 0,5 a 0,25a beams is a and the moments
calculated using the influence
my areas in Figure 4.12 are mar-
ked by small circles. mx is the
moment in the cross direction
0,2a 0,6a 0,2a
and my is the moment in the
longitudinal direction.

 98 
Figure 4.12 Influence areas for concentrated loads acting on a plate between two longi-
tudinal clamped support lines. Diagram a) shows the central moment in the y-
direction, diagram b) shows the field moment in the x-axis direction and
diagram c) the clamping moment in the x-axis direction. All values shall be
multiplied by the factor 0,04. The diagram is re-drawn from Pucher (1951).

 99 
4.4.3 Approximate method using fictitious distribution widths
An approximate method for calculating the moments in plates due to concentrated loads is to
assume that a part of the slab can be considered to behave like a beam carrying the concen-
trated loads to the supports. The part of the slab carrying this beam can thus be regarded as a
certain contributory width. This width can for certain cases be calculated using elastic plate
theory e.g. as presented in Figure 4.12. Figure 4.13 shows a scheme for deciding such a
contribution width according to methods used in Sweden for a long time.

a) b) beff
beff bb
a
d 2, eff

d1,eff d1,eff d1,eff

Support line

Figure 4.13 When calculating sagging and hogging moment emerging from a concentrated
load acting on a slab supported along two support lines a certain amount of the
slab width (beff) can be assumed to contribute to carrying these loads.

The contributory width (beff) that can be used can very approximately be calculated using
Eqs. (4-15) or (4-16)
For one wheel load (Figure 4.13 a):

beff [ m] d1,eff  min(0,75a ; 2,5) (4-15)

For two adjacent wheel loads (bogie), (Figure 4.13 b):

beff [m] 2d1,eff  min(1,5a ; 2,5  bb  d1,eff ; 0,75a  bb  d1,eff ) (4-16)

 100 
Example 4-1

In this example a comparison is made between load distribution according to Pucher (1951)
and the simplified method presented in Figure 4.13 for a bogie acting on a slab carried by
two beams clamped to the slab. The distance between the beam is 5 m and the measures of the
bogie is shown in Figure 4.14. Figure 4.10 gives a 3D-view of this load case.

1,5 m
1,5 m
5,0 m
2,0 m
1,5 m

Figure 4.14 Example4-1. A part of the roadway slab of a beam girder bridge is acted on by
loads from a bogie with measures according to the figure.

It is assumed that the thickness of the surfacing is 0,05 m. It is also assumed that the slab
2
thickness is 0,20 m and that the tyre contact area is 0,20·0,60 m . With this input data and
using the equations given in Figure 4.3, the effective distributed loaded areas that can be used
in the calculation of moments for all four wheels are

0,20  2 ˜ 0,05  2 ˜ 0,10 ˜ 0,60  2 ˜ 0,05  2 ˜ 0,10 0,50 ˜ 0,90 m 2

These effective areas is depicted in Figure 4.14. Figure 4.14 is of course a scaled version of
Figure 4.12. The point with the expected largest hogging moment can be derived integrating
the values of the influence areas in Figure 4.14. Four values are read for each load area which
is enough to get the required accuracy. The result is:

P ª 2 ˜ 6,6  2 ˜ 5,8  3,6  2  3,2  2,6  º

mxp,i  ˜ 0,04 ˜ « » 0,49 P
4 ¬ 2 ˜ 2,5  2 ˜ 1,6  1,6  1,4  1,0  0,8 ¼

For a beam clamped at both ends with a width beff and with a span 5 m, the hogging moment,
using Table A.6 in the textbook Beam and Frame Structures, is

 101 
 1 § 2 P ˜ 1,5 ˜ 3,52 2 P ˜ 1,52 ˜ 3,5 · P
mxa,i  ¨  ¸ 2,1
beff ¨ 52 52 ¸ beff
© ¹

The effective width according to Eq. (5-16) is

beff 0,5 ˜ 2  2,5  1,5  0,5 4,5 m

and thus

P
mxa,i 2,1 0,47 P
4,5

The result using the two methods give almost equal results, which implies that the simplified
method presented in Figure 4.13 can be used in most cases. For more extreme cases, e.g. for
loads close to the supports or for calculating moments close to the loads, greater differences
between the methods are expected. This is principally due to the elastic plate theory behind
Figure 4.12, which gives to large moments close to concentrated loads. For more accurate
values the Mindlin plate theory should be used and in practice solved using FEM, see Elastic
Plate Theory for Bridge Superstructures.

4.5 Load distribution between two beams

4.5.1 General
In the textbook Torsion of Concrete Structures, Section 1.4, is shown that for a two beam
bridge carrying a road way deck slab, the load distribution is dependant on the relative stiff-
ness of the two main load bearing elements, the beams and the slab. If the load on the bridge
is not symmetrical, which is most often the case, the treatment of the load distribution case is
more complicated.
The treatment of this problem is closely related to calculating what is called the “lane factor”.
In the simples case this factor for a two beam bridge tells how much of half of the imposed
traffic load that must be carried by each of the beams. This factor > 1 combined with the load
group of the type shown in e.g. Figure 3.7 is then used for the design of the individual beams.
Usually the bridge section is symmetrical, which implies that both beams should be designed
for the same load. Since the lane factor by definition is larger than one, the total design load
action e.g. for a two lane bridge is larger than two. There are other definitions of “lane
factors”, coupled to a specific standard, but this type of definition is not presented in this con-
text. To start with, the simplest case is studied in the following sections.

 102 
4.5.2 Lane factor not taking the restraint from the beams into con-
sideration
The lane factor is first calculated for the simple case shown in Figure 4.15.

1 2
a

Figure 4.15 The lane factor (f) for the case shown in the figure indicates how large portion
of half of the total load that is acting on each beam. The factor f1 is related to
beam 1 and f2 to beam 2.

The task is to decide how much the load effect on the beam is enlarged due to the load
eccentricity. If the slab is assumed to be simply supported on the beams, equilibrium
equations give:

r1  r2 qc (4-17)

r2 a qc(c / 2  (b  a ) / 2) (4-18)

If q is load per unit area acting on the roadway slab, r1 and r2 are the forces per unit length
that is acting on each beam.

f2 2r2 / qc (c / a  b / a  1) (4-19)

f1 2  f2 (4-20)

Factors f1 are presented in Figure 4.16 as function of a/b and c/b. As expected the lane factor
is large when the relative distance between the beams is small and gets its smallest value
when the beams are located at the slab edges.
A typical case can be c/b = 0,75 i.e. a two lane bridge with width 8 m. When the beams in this
case are situated along the edges of the bridge the lane factor is 1,25 according to Figure
4.16, and for the case a/b = 0,6 the lane factor is 1,42. At the same time the moment in the
slab increases when the distance between the beams increases. This issue will be discussed in
the next section.

 103 
 2,00

1,90

1,80

1,70 c /b
Lane factor f 1

1,60 0,50
1,50 0,60
0,70
1,40
0,80
1,30 0,90
1,00
1,20

1,10

1,00
0,50 0,60 0,70 0,80 0,90 1,00
a /b

Figure 4.16 Lane factor for a girder bridge with two beams as function of the loaded width
(c), the bridge width (b) and the distance between the beams (a) for a case
when the torsion restraint from the beams are neglected.

4.5.3 Optimal distance between beams with respect to moment in the

slab
To illustrate how the moments in the bridge slab carried by two beams changes when the
distance between the beams varies, Figure 4.17 is studied.

M support
qa 2

8
M centre

a
b

Figure 4.17 The moment distribution in a bridge deck slab with evenly distributed load and
which is simply supported on two longitudinal beams receives a moment distri-
bution according to the figure. The question is now which relative beam dis-
tance that gives the smallest moments in the slab.

 104 
Figure 4.17 shows the case with evenly distributed load, but the principal reasoning is the
same even for groups of concentrated loads, which is the situation in most practical cases. In
practise the design of the roadway slab is often dependant on large axle and wheel loads.
The following elementary formulas are valid

M centre qa 2 / 8  M support (4-21)

ba 2
M support q( ) /2 (4-22)
2

An optimal use of material and saving reinforcement, is to chose a moment distribution where
the hogging moment over the beams and the maximum moment in the span are numerically
equal. This gives the condition:

a /b 2  2 | 0,59 (4-23)

and
M centre M support ( 2  1) 2 qb 2 / 8 | 0,17 qb 2 / 8 (4-24)

From this result it is obvious that the absolute maximum values of moments in the slab are
much smaller if the distance between the beams are chosen close to the value a/b = 0,6.
Combining the results from Section 4.5.2 and this section implies that it is most economical to
chose a distance between the beams that is an optimisation between the result giving the
smallest lane factors and smallest moment is the roadway slab. Usually it is value giving the
smallest moments in the slab which is of greatest importance.
For small bridges and for railway bridges it is in many cases practical to chose a solution with
the main beams along the bridge edges. In such a case the main beams can protrude over the
slab and can thus be used for reduction of the noise from the traffic on the bridge.
In this section we have assumed that the slab is simply supported on the beams. In the next
section the case including the interaction with the torsion in the beams will be discussed.

4.5.4 Calculating the lane factor for a two beam bridge including the
interaction between the moment in the slab and the torsion in the
main beams.
The result in the textbook Torsion of Concrete Beams Section 1.4 is combined with the
problem discussed in Section 4.5.2 but now including the torsion in the longitudinal beams.
The deflections in the main beams are disregarded, and to solve the problem the loads are
divided into a symmetrical and an anti-symmetrical case, according to Figure 4.18.

 105 
 e a e
b

b-c 2 c- b b-c

Figure 4.18 To calculate the lane factor for a two beam bridge in the case where the torsion
in the longitudinal are taken into consideration, the load is divided into a sym-
metrical and an anti-symmetrical case.
The symmetrical case c is treated in the same way as the case treated in Section 4.5.2
combined with the equations in Torsion if Concrete Beams, Section 1.4 and the result is

3
§ b c · qa t a
Is ss ¨ , ¸  xs (4-25)
© a b ¹ 24 Ei 2 Ei

t xs
Iscc  (4-26)
GKv

With the solutions

qa 2 cosh( N s x)
t xs ss
12 cosh (ksA/ 2 )

qa 2
ts,max ss (4-27)
12

 106 
 qa 2 1
ts,min ss
12 cosh (ksA/ 2 )

ts,max is received for x =A /2 and ts,min is received for x = 0 and

qa 2 A tanh(N sA / 2)
Ts,max ss (4-28)
12 2 N sA / 2

EiA2
N sA / 2 (4-29)
2GKva

The solutions are as can be seen the same as in Torsion of Concrete Beams, Section 1.4. but
now multiplied with the factor ss. The factor ss can be calculated using ordinary slope deflec-
tion theory, giving

ss ^0,5  3 e / a  0,25(2c / a  b / a)>3  2c / a  b / a @` when (2c/a  b/a) < 1 (4-30)

2 2

ss ^1  3> e / a  e / a  b / a  c / a @`
2 2
when (2c/a  b/a) t 1 (4-31)

In the same way the slope deflection for the anti-symmetrical case d is:

2
§ b c · qa t ( a / 2)
Ia sa ¨ , ¸  xa (4-32)
© a b ¹ 24 Ei 3Ei

t xa
Iacc  (4-33)
GKv

The definition of the slope I a and other conditions are evident from Figure 4.19.

 107 
Figure 4.19 Definition of slope deflections, forces and moments for the anti-symmetrical
load case.

In the same way as before the equations have the solutions

qa 2 cosh (ka x)
t xa sa (4-34)
12 cosh (ka A/ 2 )

qa 2
ta,max sa (4-35)
12

qa 2 1
ta,min sa (4-36)
12 cosh (ka A/ 2 )

ta,max is received for x = A/2 and ta,min is received for x = 0 and thus

 108 
 qa 2 A tanh(N a A / 2)
Ta,max sa (4-37)
4 2 N aA / 2

3EiA2
N aA / 2 (4-38)
2GKva

As can be seen the solutions are in principal the same as in Torsion of Concrete Beams,
Section 1.4, but with the symmetrical load case multiplied by the factor sa and with an other
value for N a A . The factor sa can be calculated using the slope deflection method for ordinary
beams and using the notations in Figure 4.19 the result is:

ª§ b c e · 2 § b c e · 2 § e · 2 º
sa «¨   ¸ ¨ 1    ¸  ¨ ¸ » when (2c/a  b/a) < 1 (4-39)
«¬© a a a ¹ © a a a ¹ © a ¹ »¼

ª§ e · 2 § e b c · 2 º
sa  «¨ ¸  ¨   ¸ » when (2c/a  b/a) t 1 (4-40)
«¬© a ¹ © a a a ¹ »¼

There is no self-evident method for the deciding the lane factor for a case of the type dis-
cussed in this section. The lane factor can be based on comparison of the deflections, the
moments or the forces. Considering the pros and cons the best method is to compare the
support reactions forces acting on the main beams. For the symmetrical case the reaction
forces at both beam ends are:

Rs 1c
(4-41)
qaA 4a

and for the anti-symmetrical case

Ra 1 c § b c· Ta,max
¨  ¸ 2 2 (4-42)
qaA 4 a © a a¹ qa A

The lane factor f1 is then

2( Ra  Rs ) a§ R R ·
f1 4 ¨ s  a¸ (4-43)
qcA c © qaA qaA ¹
2

Using the equations presented in this chapter the lane factor can be decided as a function of
the torsion restraint. Results are depicted in Figure 4.20, Figure 4.21 and Figure 4.22 for the
cases c/b = 0,6, 0,7 and 0,8 respectively. The effect of the torsion restraint is given as a
function of the non-dimensional parameter:

 109 
 EiA2
E (4-44)
GKva

E = 0 corresponds to a case where the main beams have very large torsional stiffness, while
the case E f corresponds to the case where the main beams have very low torsional stiff-
ness, which is the same case as that treated in Section 4.5.2.
The results in Figure 4.20 – Figure 4.22, show that the main variation occurs in the interval
0 , 1  E  10 . It is also obvious that the effect of the torsion restraint is not very important,
but in interesting cases reduces the lane factor, especially when the main beams are close to
the edges of the bridge.

1,85

1,80

1,75 c /b = 0,6

1,70

1,65 E
f1

0
1,60
1
1,55 10
100
1,50
Large
1,45

1,40
0,50 0,60 0,70 0,80 0,90 1,00
a /b

Figure 4.20 Lane factor for c/b = 0,6 when the relative distance between the main beams
(a/b) and the factor representing the torsion restraint E are varied. For
notations see Figure 4.18.

 110 
 1,65

1,60

1,55 c /b = 0,7
1,50

1,45
E
f1

1,40
0
1,35 1
1,30 10
100
1,25 Large
1,20
0,50 0,60 0,70 0,80 0,90 1,00
a /b

Figure 4.21 Lane factor for c/b = 0,7 when the relative distance between the main beams
(a/b) and the factor representing the torsion restraint E are varied. For nota-
tions see Figure 4.18.

1,45

1,40 c /b = 0,8

1,35

E
f1

1,30

0
1,25 1
10
1,20 100
Large
1,15
0,50 0,60 0,70 0,80 0,90 1,00
a/b

Figure 4.22 Lane factor for c/b = 0,8 when the relative distance between the main beams
(a/b) and the factor representing the torsion restraint E are varied. For nota-
tions see Figure 4.18.

 111 
To show the effect of the torsion restraint, Figure 4.23 shows the variation in the lane factor
for a typical case c/b = 0,7, when the E-value and the relative distance between the beams,
a/b, are varied. From the figure it is also obvious that the variation of the lane factor is much
smaller when the torsion restraint is large i.e. when E reaches small values.

1,65

1,60
a /b
1,55
0,50
1,50
0,60
1,45
0,70
f1

1,40 0,80
0,90
1,35
1,00
1,30

1,25

1,20
0,001 0,01 0,1 1 10 100 1000
E

Figure 4.23 Lane factor when the traffic lanes cover 70 % of the bridge area i.e. c/b = 0,7,
when the relative distance between the beams a/b and the factor representing
the torsion restraint E are varied.

The formulas derived in this section can also be used to modify the calculation of the
moments in the slab connecting the two beams. The special case E f is treated in Section
4.5.2. For other cases the analysis can be performed using a combination of symmetrical and
anti-symmetrical load cases and then using the derived values tsx and tax for the clamping
moments. This case is also treated in Elastic Plate Theory for Bridge Superstructures, Section
3.4.

Example 4-2

Using the input values used in Example 1-3, in Torsion of Concrete Beams the E-factor is

EiA2 E ˜ 6,7 ˜ 104 ˜ 122

E 11
,
GKva 0,022 E ˜ 4

when the beams are supposed not to be cracked due to torsion moments and

E 10 ˜ 11
, 11

 112 
when the beams are assumed to be cracked due to torsion. It is obvious that these values are in
a range where the torsion restraint is of importance and that the lane factor thus is dependant
on the torsion stiffness of the beams. It should be noted that a reduction of the lane factor,
which saves material in the bending design of the main beams, results in a demand for torsion
reinforcement in these beams. The whole problem is thus an optimisation of the best usage of
material. According to the Swedish standard for bridges Bro 2004, there shall always be some
torsion reinforcement in all beams. At least this amount should be used for the reduction of
the lane factors.

4.6 Load distribution between many beams

In the preceding section the load distribution for un-symmetrical loads on bridges with two
beams have been discussed. In cases with more beams the same principles can be used, but
then the vertical deflections of the beams must be taken into consideration. The conditions
gets however even for simplified cases very complicated and impractical to use. For these
general cases effective computer tools like FEM shall be used.
For conceptual design, however, very simplified methods can show important trends in the
behaviour of bridges with many beams.
A very simple case is of course to totally neglect the structural behaviour of the bridge and
just define the lane factor, f, as the value f c / 3 , where “3” is the width of the notional lane
and c is the distance between the beams.
Two cases, approximately taking the structural behaviour into consideration, will be dis-
cussed: stiff cross beams and very weak cross beams.

4.6.1 Stiff cross beams

A case with many parallel main beams spliced together with many stiff cross beams is
studied. The beams are carrying a bridge acted on by an eccentric lane load in principle accor-
ding to Figure 4.24.

q c)
q
L
Cross
b) beams
c c c c
c
Main
c
beams
c
c y0

Figure 4.24 A bridge with many parallel main beams spliced together with many stiff cross-
beams is studied.

 113 
With the given conditions the cross-beams can be assumed to be straight, thus giving that the
deflection distribution in the transversal direction is a straight line that can be described using
Equation (4-45)

yi y0  T ri (4-45)

where ri is the distance from beam i to the centreline of the bridge and inserted with its sign.

It is further assumed that the beam no i is loaded with an evenly distributed load qi . All beams
have the same support conditions at both ends.
The following equation can then be used for the deflection of the beams

5qi L4
yi ( x ) f ( x) (4-46)
384 EI i

If it is assumed that all functions f are the same for all beams, it is only necessary to study an
arbitrary section x along the bridge. Eq. (4-46) can then be modified according to Eq. (4-47).

qi
yi (4-47)
ki

Observe that using this methodology and notation, the beams need not to be simply supported
or acted on by evenly distributed loads. The important condition is that the type of support
and load distribution are the same for all beams. Combining Eq. (4-47) and Eq. (4-45) gives:

qi ki y0  ki riT (4-48)

The total load on the bridge is Q and the distance from the centre of gravity of this load to the
bridge centre line is e. Equilibrium for force and moment then gives:

½
n
Q ¦ qi
°
1 °
n
¾ (4-49)
°
Qe ¦ ri qi °
1 ¿

or with the q i -values inserted

n n
Q y0 ¦ ki  T ¦ ri ki (4-50)
1 1

 114 
 n n
Qe y0 ¦ ki ri  T ¦ ri2 ki (4-51)
1 1

A bridge is usually designed with the main beams symmetrically distributed in relation to the
centre line of the bridge. If this is the case then

n
¦ ri ki 0 (4-52)
1

Eqs. (4-50) and (4-51) is in this case simplified giving:

y0 Q / ¦ ki (4-53)

T Qe / ¦ ri2 ki (4-54)

If values from Eqs. (4-53) and (4-53) are inserted in Eq. (4-48) and using the notation
q = Q/n, which implies that q can be regarded as the average value of the load, the result is

qi ki ri ki
fi n en (4-55)
q ¦ ki ¦ ri2 ki
The f i -factors can be considered to be load distribution factors, “lane factors”, thus descri-
bing the amount of the load, in relation to the average load, each beam shall be designed to be
able to carry.
Since the position of the load normally can vary in the transversal direction, Eq. (4-55) must
be tested for several values for e, and the most un-favourable value for each beam must be
chosen.
In the above performed derivation of load distribution factors, it has been assumed that the
bridge is symmetrical in relation to its centre line. It is not much more complicated to derive
factors for an un-symmetrical case which gives:

qi ¦ ri2 ki  e¦ ri ki e¦ k i  ¦ ri ki
fi nki  n ki ri (4-56)
q ¦ ki ¦ ri2 ki  ¦ ri ki ¦ ki ¦ ri2 ki  ¦ ri ki
2 2

Example 4-3

In this example the equations derived above will be tested for the bridge depicted in
Figure 4.25. The total bridge width is 11 m and shall thus be designed for 3 lanes each with a
notional width of 3 m. The bridge is carried by 3 beams each equal in size and stiffness.

 115 
 3,0 3,0 3,0

q2 q3
q1

4,0 4,0

11,0

Figure 4.25 Example 4-3, an 11 m wide bridge section with 3 main beams acted on by 3
lane loads with notional widths 3 m.

Since the bridge is symmetrical Eq. (4-55) can be used. The result is:

k0 4 k0 3
f1 f3 3 e ˜ 3 1  e
3k0 k0 (42  0  42 ) 8
k0
f2 3 1
3k0

Different standards apply different methods for defining design lane loads. One method is to
use the same load for all lanes, but reduce their intensity by the number of designing loads.
This method is given in Canadian standards. An other method is to like in Sweden and the
Eurocode apply different loads on each lane, see Section 3.3.2.
The very simple methods presented in this section can be used for both methods, but the
definition of the total load Q in Eq. (4-49) will be different and thus also the eccentricity e.
The result given in Eqs. (4-53) and (4-54) is directly applicable to the first method, but the
method can also after some alteration be used for the second case.
In Table 4.1, the first case, see Figure 4.24, is illustrated for a case where the load on one
lane is 1,0, if there are two loaded lanes the load on both is 0,9 and if there are three loaded
lanes the load on each is the same but with the intensity 0,8.

Number of Design Max

loaded lanes lane load eccenticity f1 Load/beam f2 Load/beam
Unit - Force/m m - Force/m -
1 1,00 4,00 2,50 2,50 1,00 1,00
2 0,90 2,50 1,94 3,49 1,00 1,80
3 0,80 1,00 1,38 3,30 1,00 2,40

Table 4.1 Example 4-3. Load distribution among the main beams for the case shown in
Figure 4.24. The cross beams are assumed to be very stiff.

 116 
In Table 4.2 the result from second method is shown. In this case the intensity on one lane is
1,0, the intensity on the next lane is 0,9 and the intensity on the third lane is 0,8. This case is
presented in Table 4.2.

Number of Design
loaded lanes lane load Eccenticity f1 Load/beam f2 Load/beam
Unit - Force/m m - Force/m -
1 1,00 4,00 2,50 2,50 1,00 1,00
2 0,90 2,58 1,97 3,45 1,00 1,80
3 0,80 1,22 1,46 3,15 1,00 2,40

Table 4.2 Example 4-3. Load distribution among the main beams for the case shown in
Figure 4.25. The cross beams are assumed to be very stiff.

From this example it is obvious that the outer beams receives higher loads than the beam in
the centre. The final result can though be influenced by the designer, since it is advisable to
give the outer beams higher strength and thus stiffness. This leads to different load distribu-
tion factors. To find the best design is consequently an optimization process where the design
of the beams also affect the load distribution, a frequent situation in the design of bridge
superstructures.
In the above presented analysis, the torsion stiffness of the beams has been disregarded. If the
torsion effect had been included the load distribution factor had been reduced a little. At least
for steel beams the torsion stiffness is so low that effect of torsion is negligible in this kind of
analysis.
The load distribution gained by the cross beams results of course in moments and shear forces
in these beams. The cross beams must be designed to be able to withstand these actions. It is
also important that in the case when the steel beams are to carry a composite concrete deck,
the deformations and possible instability problems are studied carefully. For more informa-
tion about these problems, see the textbook Steel Concrete Composite Bridges.

4.6.2 Structural co-operation between main and cross beams

In the preceding section the case with very stiff cross beams has been studied. The real case is
however that the deformation of the cross beams must be taken into consideration. A struc-
tural system with beams spliced together in two or more directions is denoted grillage. Such
structural systems usually with only two beams are discussed in the text book Steel Concrete
Composite Bridges. The Swedish Bridge Standard require that the distance between cross
beams shall not be more than 8 m.
A case with many beams in both directions is depicted in Figure 4.26. The analysis of these
kind of grillages is complicated to perform by hand and thus computer methods are suitable.

 117 
 a)

Cross beam Main beam

b)
Ia 1 2 3
A A
c
Ib
B B
c I1 I2 I3
Ib
B' B'
c
Ia
A' A'
1 2 3
L

Figure 4.26 Grillage with four main and three crossbeams. Ia and Ib are the moment of
inertia of the main beams respectively. I1 , I2 and I3 are the moment of inertia
of the cross beams.

It is usually assumed that the torsional stiffness of the beams is so small that the effect of the
torsion on the load distribution can be neglected. For the dead load it is suitable to design the
structure so that each beam carries its part of the dead weight and thus no distribution of dead
load amongst the beams is necessary. For the impost load, however, distribution usually must
be taken into consideration and then the stiffness of the cross beams must be included in the
analysis.
One way to analyse the effect of the co-operation between the main beams in a grillage is to
compare the deformations in the node points in the grillage for unit loads placed at each node.
For the example shown in Figure 4.26 this analysis will result in 12 unknowns and 12 load
cases. The load distribution and the moments and shear forces in each beam can then be
calculated using the result from the first calculation. For this kind of problems there are many
standard computer programs available.

4.6.3 Load distribution between stiff longitudinal beams and a roadway

slab
In the case when the longitudinal beams can considered to be very stiff in relation to the cross
beams or when there are no cross beams, but just a roadway slab carried by the longitudinal
beams, the load distribution can be calculated using ordinary theory for influence lines. The
roadway deck slab is in this case modelled as beam strips in the transversal direction.
This assumption is acceptable if the distance between the main beams is short in relation to
the span of the main beams. Using the methods presented in Beam and Frame Structures,
Chapter 2, including Appendix C, load distribution can be analysed for different cases.

 118 
For a bridge with e.g. 3 beams, the influence diagram shown in Figure 4.27 can be applied.
The diagram presents the reaction forces R1 – R3 acting on the longitudinal beams when a unit
force travels from one bridge edge to the other. Observe that the influence diagram also cover
loads on the cantilevering part of the bridge slab.
Depending on the dimensions of the bridge and the extension of the lane loads, the part of the
load acting on each beam can be calculated using numerical integration of the influence line
for each beam. The position of the lane loads giving the greatest load action must be used for
each longitudinal beam.
In Figure 4.28 the case with four main beams is presented and can be used in a corresponding
way as is used for the three beam bridge.

2,0

R1/P
1,5 R2/P
R3/P

1,0

0,5

0,0

-0,5

-1,0
-0,5 0 0,5 1 1,5 2 2,5
x /c

Figure 4.27 Influence lines for beam on three supports and where c is the distance between
the beams. This type of diagram can be used to analyse the load distribution
between the beams for different positions of the lane loads. The diagram is cal-
culated for the case with stiff longitudinal beams.

 119 
 1,4
1,2
1,0
0,8
R1/P
0,6 R2/P
0,4 R3/P
0,2 R4/P
0,0
-0,2
-0,4
-0,6
-0,50 0,00 0,50 1,00 1,50 2,00 2,50 3,00 3,50
x /c

Figure 4.28 Influence lines for a beam on four supports and where c is the distance between
the beams. This type of diagram can be used to analyse the load distribution
between the beams for different positions of the lane loads. The diagram is cal-
culated for the case with stiff longitudinal beams.

Example 4-4

A bridge with four main beams and only one cross beam according to Figure 4.29 is studied.
A concentrated load P = 1 is travelling along the centre line of the bridge. The task is to cal-
culate the influence line for the three principal cases presented in Section 4.6.1 – 4.6.3.

1 P =1 Ia
A A
c =5,25 m
Ib
B B
c E I1
Ib
B' B'
c
Ia
A' A'
1

L =42 m

Figure 4.29 Example 4-4.

Using the input data

Ia Ia c 5, 25 1
3/ 4 ; 20 ;
Ib I1 L 42, 0 8

 120 
and calculating the influence line when the load travels along the cross beam in the bridge
centre gives the result according to Figure 4.30. The calculation also includes the influence of
a protruding deck slab with a cantilever length of 0,3 c. In Figure 4.30 also the two extreme
cases very stiff cross beams (Section 4.6.1) and very stiff main beams (Section 4.6.3) are
presented for comparison.

1,60
1,40 Stiff main beams
1,20 Stiff cross beam
1,00 Grillage

0,80
0,60
0,40
0,20
0,00
-0,20
-0,40
-0,50 0,00 0,50 1,00 1,50 2,00 2,50 3,00 3,50

x /c

Figure 4.30 Influence line for the case when a concentrated load P = 1 travels from one
edge to the other edge along the cross beam on the bridge depicted in
Figure 4.29. The three cases shown are stiff main beams, see Section 4.6.3, stiff
cross beam, see Section 4.6.1 and grillage, see Section 4.6.2.

 121 
The literature in the field shortly presented in this textbook is
comprehensive. Especially in the field of dynamic actions on
structures the research effort is nowadays extensive.

 122 
 5. Literature
5.1 Literature on Risk and safety for buildings and structures
Some interesting literature on risk and safety for buildings and structures which is be found in
my office. The list is approximately classified according to the following list:
A.: General on risk, safety and reliability.
B.: Failure, accidents for building structures.
C.: Safety analysis for bridges and buildings.
D.: Structural design of structures acted on by explosions, missiles and violent sabotage.
E.: Standards on Structural Design, Risk and safety.

Author Title Class

Ahlenius, E., (1999) Explosionslaster och infrastrukturkonstruktioner – D
Risker, värderingar och kostnader. TRITA-BKN.
Bulletin 47, Licentiatavhandling.
Ahlenius, E., (1999) Om lönsam och effektiv riskhantering, Väg- och Vatten- A
byggaren nr 1, 1999
Ahlenius, E., (1999) Uncertainties of Explosion Loads and its Influence on D, A
Reliability, IABSE Symposium – Rio de Janeiro, Aug,
1999
Andersson, L., (1988) Möjlighetsteori och diffusa mängder – Nya idéer för A
riskanalys och beslutsfattande, BFR Rapport R23: 1988,
doktorsavhandling i brobyggnad, KTH, 1988
Arbetsmiljöverket, Arbetsmiljölagen, med kommentarer i lydelse från den 1 E
(2000) januari 2001. Arbetsmiljöverket, Solna 2000.
BBK 04 Boverkets handbok om betongkonstruktioner. Handbook E
on design, construction and maintenance of concrete
structures. Boverket, Karlskrona 2004.
BBR (2006) Boverkets byggregler. General rules for construction of E
houses and structures, Boverket, Karlskorona 2006.
BKR (2003) Regelsamling för konstruktion. General rules for E
structural design, Boverket Karlskrona.
BRE Building Research Building Failure, The Construction Press Ltd, Lancaster, B
Series, (1978) UK,
Bro 2004 (2004) The Swedish Standard for Road Bridges, Borlänge 2004 E
BV Bro (2004) The Swedish Standard for Railway bridges, Edition 8, E
Borlänge 2004.
Carlsson F., (2006) Modelling of Traffic Loads on Bridges – Based on mea- C
surements of Real Traffic Loads in Sweden, PhD Thesis,
Lund University, Report TVBK-1032.

 123 
Ditlevsen, O., & Structural Reliability Methods, John Wiley & Sons, A
Madsen, H., O., (1996) Chichester 1996.
Fellenius, W., (1918) Skredet vid Getå å statsbanan Järna-Norrköping, Teknisk B
Tidskrift, Dec. 1918
Ferguson, E. S. (1995) How engineers lose touch. Invention & Technology, A
New York 1993.
Forty S., (2005) Defining Moments – Disasters, Grange Books, UK, 2005 B
Getachew A., (2003) Traffic Load Effects on Bridges – Statistical Analysis of C
Collected and Monte Carlo Simulated Vehicle Data,
TRITA-BKN. Bulletin 68
Gordon, J. E. (1978) Structures or why things don’t fall down. Da Capo Press, B
New York.
Grimvall G., & Lindgren Risker och riskbedömningar, Studentlitteratur Lund, A
O. (1995) 1995.
Grimvall, G., Jacobsson Risker i tekniska system, Studentlitteratur, Lund. A
P., & Thedéen T.,
(2003)
Hasofer, A., M., & Lind, Exact and Invariant Second-Moment Code Format, Proc. A
N. C., (1974) ASCE, Journal of Eng. Mech. Division, Vol 100, No.
EM 1.
Hult J. (1990) Spänning och brott, Forskningens frontlinjer, Stockholm. A
IVA, (1981) Skydd mot fria GASMOLNSEXPLOSIONER i process- D
industrin, Medd. 238, Stockholm 1981.
James, G., 2003 Analysis of Traffic Load Effects on Railway Bridges, C
TRITA-BKN. Bulletin 70, Doctoral Thesis, Structural
Design and Bridges, KTH
Johnson A., (1959) Statistisk-ekonomisk dimensionering, Tekn. Dr Arne C
Johnson Ingenjörsbyrå, Tekniska meddelanden Nr 3,
danderyd 1959.
Kjellman W., Wästlund Säkerhetsproblemet i byggnadskonsten, IVA Handlingar C
G., (1940) Nr 156, Stockholm 1940
Levy M. & Salvadori, Why buildings fall down. W W Norton & Co, Second ed. B
M., (2002) 2002, New York.
Melchers R., E., (1999) Structural Reliability Analysis and Prediction, 2nd ed. C
2002, John Wiley & Sons Ltd, Chichester, UK.
Nasr T., (2000) Druckenentlastung bei Staubexplosionen in Siloanlagen, D
Massivbau, Baustofftechnologie, Karlsruhe, Heft 38.
Petroski H., 1994 Design Paradigms – Case Histories of Error and A
Judgment in Engineering. Cambridge University Press,
1994.
Petrovski H., 1985 To Engineer is Human, The Role of Failure in Successful A

 124 
 Design. ST. Martin´s Press, New York.
Pettersson, O. 1977 Några byggnadskatastrofer genom tiderna. Väg- och B
Vattenbyggaren, nr. X, Stockholm.
Riskkollegiet, 1995 KEMISKA RISKER – beslutsfattandets problem. Infor- A
mation från Riskkollegiet, Riskkollegiets Skriftserie,
Stockholm 1995
Roddis, W. M. K. 1993 Structural Failures and Engineering Ethics. Journal of A
Structural Engineering, ASCE, New York.
Salvadori M., (1979) The Art of Construction. Chicago Review Press, 3rd ed. C
1990. (A state-of-the-art book even for students without
structural background.)
Scheer, Joachim (2000) Versagen von Bauwerken Band 1: Brücken, Ernst und B
Sohn, Berlin, ISBN-10: 3-433-01802-2, ISBN-13: 978-3-
433-01802-6
Scheer, Joachim (2001) Versagen von Bauwerken Band 2: Hochbauten und B
Sonderkonstruktionen, Ernst und Sohn, Berlin 2001
Schlaich J. & Reineck Die Ursache fuer den Totalverlust der Betonplattform B
K.-H. 1993 Sleipner A. Beton- und Stahlbetonbau, Berlin.
Schneider J., (1997) Introduction to Safety and reliability of Structures, C
IABSE Structural Engineering Documents 5, Zürich
1997.
Shepard R., Frost J., D., Failures in Civil Engineering: Structural, Foundation B
1995 and Geoenvironmental Case Studies, American Society
of Civil Engineers, Reston, 1995
SJ (1918) Årsberättelse för 1918. Statens Järnvägar (SJ), B
Stockholm.
Sjöberg, L., & Ogander, ATT RÄDDA LIV - Kostnader och effekter, Rapport till A
T. 1994 expertgruppen för studier i offentlig ekonomi, Finansde-
partementet, Ds 1994:14, Stockholm.
Stamm, C., (1952) Brückeneinstürze und ihre Lehren, Mitteilungen aus dem B
Institut für Baustatik, Nr. 24, Zürich, 1952
Timoshenko, S. P. History of strength of materials. Mc Graw-Hill, London. B
(1953)
Wirén E. (1998) Planering för säkerhets skull. Studentlitteratur, Lund A
1998

5.2 Literature on load distribution

Bakht B., Jaeger L.,G., Bridge Analysis Simplified, Mc Graw-Hill Book Company, New
(1985) York 1985.
Cusens A.R., Pama R.P., Bridge deck analysis, John Wiley & Sons, Bristol 1975.
(1975)

 125 
Hambly E., C., (1991) Bridge Deck Behaviour, 2nd ed., E& FN Spon, London 1991.
Hetenyi M., (1964) Beams on Elastic Foundation, the University of Michigan Press
1964.
Leonardt F., (1982) Bridges – Aestetics and Design, Deutsche Verlags-Anstalt,
Stuttgart, 1982.
Menn C., (1990) Prestressed Concrete Bridges, Birkhäuser Verlag, Basel 1990.
O’Connor C., (1971) Design of Bridge Superstructures, John Wiley & Sons, Inc, New
York 1971.
Pucher A., (1951) Einflussfelder elastischer Platten, Springer Verlag, Wien 1951.
Sattler K., (1955) Betrachtungen zum Berechnungsverfahren von Guyon-Massonnet
für freiaufliegene trägerroste und Erweiterung dieses Verfahrens
auf beliebige Systeme, Der Bauingenieur, Berlin 1955.
Schlaich J., Scheef H., Concrete Box-Girder Bridges, IABSE Structural Engineering
(1982) Documents 1e, Zürich 1982.
Sundquist H., (1972) Verkningssätt för plattor med avlånga ursparingar, Inst. för
Byggnadsstatik, KTH, Stockholm 1972.
Vlasov V., Z., (1961) Thin-Walled Elastic Beams, Translated from Russian, Israel Pro-
gram for Scientific Translations, Jerusalem 1961.
Wästlund G., (1964) Kompendium i Brobyggnad, del II. KTH, Stockholm 1964.
Young W., C., (1989) ROARK`S Formulas for Stress and Strain, Mc Graw-Hill Book
Company, New York 1989.
Åkesson B., Å., (1970) Handbok i analys av balkars vridning, del 1 och 2, Chalmers
Tekniska Högskola, Institutionen för Byggnadsstatik, 1970.

 126 
 6. Appendix
6.1 Designations and notations for safety analysis
For a random sample X of variables xi (i = 1, ... n) we define the arithmetical mean mX as

1 n
mX ¦ xi
ni 1

As a measure of the scatter the second moment, the so called variance of X, Var(X) is defined
as

1 n
¦ xi  mX
2
Var ( X ) s X2
n 1 i 1

The standard deviation sX is the square root of the variance i.e.

sX Var ( X )

The coefficient of variation is defined as

sX
vX
mX

Observe that we use roman or Latin letters (i.e. mX, sX) for parameters obtained from random
samples. For the corresponding parameters of continuous distribution functions Greek letters
(i.e. PX, VX ) are used.

6.2 The Normal distribution N(0,1)

The Normal distribution is one which arises frequently in practise as a limiting case of other
probability distributions. It is also a reasonable model for observation of many physical
processes or properties.
Its probability density N(PV) and is cumulative distribution functions are given by

1 ª 1 § x  P ·2 º
I X ( x) «
exp  ¨ X
¸ » ; f d x d f (A-1)
VX 2ʌ « 2© VX ¹ »
¬ ¼

x  P X /V X
1
P ( X d x) ) X ( x)
2ʌ
³ exp  12 v 2 dv ;  f d x d f (A-2)
v f

 127 
In Figure A.1 the distribution is depicted for the normalized function N(0,1), and in
Table A.1 the same function is tabulated.

1,00

0,90
Normal cumulative distribution
0,80
function N (0,1)
0,70

0,60

0,50
Normal probability
0,40 density function N (0,1)
0,30

0,20

0,10

0,00
-4,00 -3,00 -2,00 -1,00 0,00 1,00 2,00 3,00 4,00

Figure A.1 The normal density and cumulative distributions N(0,1).

A useful approximate function for calculating the cumulative distribution function is

1 § 1 ·
) X ( E ) exp ¨  E 2 ¸ (A-3)
E 2ʌ © 2 ¹

 128 
 x I x ) x
 ( (
 ( (
 ( (
 ( (
 ( (
 ( (
 ( ( x I x ) x
 ( (   
 ( (   
 ( (   
 ( (   
 ( (   
  (   
  (   
     
     
     
     
     
     
     
     
     
     
     
     
     
     
     
     
     
     
     

Table A.1 The normal density and cumulative distributions N(0,1).

6.3 Log normal distribution LN(OH)

In this distribution, the natural logarithm of the random variable X, rather than X itself, has a
normal distribution. The probability density and the cumulative distribution functions are:

1 ª 1 § ln x  O · 2 º
f X ( x) exp «  ¨ ¸ » (A-4)
2ʌ «¬ 2 © H ¹ »¼

x
§ ln x  O ·
FX ( x) ³ f X (u )du )¨
© H
¸ (Observe there is no explicit form)
¹
(A-5)
f

 129 
Parameters

O E (ln X ) mean of ln( X ) Pln x (A-6)

H2 var(ln x) V ln2 x (A-7)

Moments

E( X ) P X = exp O + 12 H 2 (A-8)

var( X ) V X2 =P X2 ª exp H 2  1º (A-9)

¬ ¼

With vX defined as the coefficient of variation we get

ª P º
O ln xm ln « X » (A-10)
« 1  v2 »
¬ X ¼

or

PX
xm (A-11)
1  v 2X

and

§ V2 ·
H 2 V ln2 X ln ¨1  X2 ¸¸ (A-12)
¨ P
© X ¹

For the function G R  S , where R and S are log normal distribution the safety index E can
be calculated using Eqs. (A-10) and (A12):

ln P R / 1  vR2  ln P S / 1  vS2
PG PR  PS
E
VG V R2  V S2 ln(1  vR2 )  ln(1  vs2 )
§P 1  vS2 · (A-13)
ln ¨ R ¸
¨ P S 1  vR2 ¸
© ¹
ln ª (1  vR2 )(1  vR2 ) º
¬ ¼

 130 
 1,80

1,60 Log normal probability density

function LN (0;0,25)
1,40

1,20

1,00
FX (x )

Lognormal probability
0,80
cumultative function
LN (0;0,25)
0,60

0,40

0,20

0,00
0,00 0,50 1,00 1,50 2,00 2,50 3,00

Figure A.2 Lognormal density and cumulative probability functions for the case

6.4 Rectangular distribution

Figure A.3 Rectangular distribution function

ab ba
P ;V ; f  a b  f (A-14)
2 12

1 xa
f ( x) ; F ( x) ; ad xdb (A-15)
ba ba

 131 
6.5 Combination of parameters for loads or strengths
A parameter e.g. the sum Y of loads Xi, (considered as statistical variables) coming from many
floors in a multi-storey building when calculating the resulting load acting on e.g. the
foundation
n
Y ¦ Xi (A-16)
1

Each variable X i has the mean value P X ,i and the standard deviation V X ,i . Between two
different values X i and X j exist a correlation Ui , j . It then true that

n
PY ¦ PX i
; X i  P X i ;V X i (A-17)
1

n n n
V Y2 ¦ V X2 i  ¦ ¦ UijV X2 i ; X i  P X i ;V X i (A-18)
1 i 1 j 1, j zi

Assume that all mean values are equal P X ,i P X , that all deviations V X ,i V X are equal
and that all correlations are equal Ui , j U . In this special case the resulting mean value and
standard deviation for Y is:
n
PY nP X (this actually not needed PY ¦ PX i
in all cases)
1

n n
V Y2 nV X2  V X2 ¦ ¦ Uij nV X2  > n(n  1) U @V X2 (A-19)
1 1, j zi

Assume first un-correlated variables giving Ui , j 0 and V Y nV X

If totally correlated variables i.e. U = 1 we get V Y nV X

If partly correlated variables the result is a value between n and n.

Combination of the sum of two statistical variables

Y aX1  bX 2 (A-20)

PY a P X1  bP X 2 (A-21)

V Y2 a 2V X2 1  b 2V X2 2  2ab U1,2V X1V X 2 (A-22)

 132 
Combination of the product of two uncorrelated variables

Y a ˜ X1 ˜ X 2 (A-23)

PY a ˜ P X1 ˜ P X 2 (A-24)

a 2 ˜ ª« P X1V X 2
2 2 2º
V Y2  P X 2 V X1  V X 2 V X1 »¼ (A-25)
¬

Or expressed in the coefficients of variation

vY2 v X2 1  v X2 2  v 2X1 ˜ v 2X 2 (A-26)

Remark: The last term in Eqs. (A-25) and (A-26) are often small and can be neglected.

 133