Está en la página 1de 109

Background document to EN 1991- Part 2

- Traffic loads for road bridges - and


consequences for the design
G. Sedlacek, G. Merzenich, M. Paschen, A. Bruls, L. Sanpaolesi, P. Croce, J.A. Calgaro, M. Pratt,
Jacob, M. Leendertz, v. de Boer, A. Vrouwenfelder, G. Hanswille

Support to the implementation, harmonization and further development of the Eurocodes

First Edition, XXXXX 2008

List of contents of the background document to EN 1991-2

Part A

Background to load models for road bridges


Traffic and impact loads
Wind loads
Temperature effects
Safety approach

Part B

Consequences for design


Bridge structures
Bridge decks
Bearings
Transition joints

Part C

Aspects related to sustainability of load assumptions


New measurements and conclusions
Trends and requirements for transport policy

Part A Background document to load models for road bridges


List of contents of sections 1-8 of Part A
1.

Objectives

2.

Basic procedure

3.

Evaluation of measured traffic data to construct the traffic load model for ULSassessments
3.1
History of measurements
3.2
Results of measurements and evaluations
3.2.1 Choice of representative traffic
3.2.2 Measurements of axle weights and conclusions for axle loads
3.2.3 Measurements of distance of axles
3.2.4 Measurements of vehicle weights
3.2.5 Definition of flowing and congested traffic
3.2.6 Results of simulations with influence lines
3.2.7 Composition of load models LM1 and LM2

4.

Representative values of traffic loads for SLS assessments


4.1
General
4.2
Results of extrapolation
4.3
Conclusions

5.

Fatigue loading models


5.1
Basic principles of fatigue
5.2
Counting method
5.3
Damage accumulation hypothesis
5.4
The concept of damage equivalence
5.5
Results of evaluation of load-distributions
5.6
Simulation of bridge responses for static fatigue actions
5.6.1 Purpose of the simulation
5.6.2 Results for fatigue loads dependant on span L
5.6.3 Comparison of the fatigue effects of the Auxerre traffic with the
fatigue load models in EN 1991-1
5.7
Conclusions

6.

Procedure for the dynamic analysis of bridges under the loads of single
vehicles
6.1
General
6.2. The mechanical models
6.2.1. General
6.2.2 The mechanical model of the bridge
6.2.3 The vehicle model
6.2.4 Model for the roadway roughness
6.3
Calibration of the models with test results
6.4
Parameter studies to determine physical impact factors
6.4.1 Impact factors for the ultimate limit state verification
6.4.1.1
Local dynamic effects
6.4.1.2
Dynamic effects of individual vehicles
3

7.

8.

6.4.1.3
Dynamic effects with several vehicles
6.4.2 Impact factors for fatigue assessments
6.4.3 Conclusions for loading codes
Dynamic simulations for justifying the load models in EN 1991-2
7.1
Procedure and assumptions
7.2
Results of the simulations for LM1 and LM2
7.3
Determination of representative values
7.4
Dynamic simulations for fatigue assessments
7.4.1 General procedure
7.4.2 Span factor a
7.4.3 Factors 2 ,3 , 4
7.3.4 Results of simulations
Braking and acceleration forces
8.1
Code provisions
8.2
Calculative model for braking forces on stiff bridges
8.3
Dynamic effects from braking
8.4
Determination of braking forces in dependence of the bridge length
8.5
Specification of braking forces in EN 1991-2, 4.4.1 and conclusions

1.

Objectives

(1)

This Part A of the background report deals with actions on road bridges and
gives the results of studies that justify the rules and recommendations for
traffic loads on bridges, see EN 1991-Part 2.

(2)

It also gives the background for the recommended choices of Nationally


Determined Parameters in the National Annexes.

(3)

The background document to the vertical traffic loads is structured such that it
gives in its section 2 to 5 the basic justification for the development of the load
pattern and of the amplitudes of the loads by evaluations of traffic with static
influence lines and in its sections 6 to 7 by using dynamic analysis of bridges.

2.

Basic procedure

(1)

The basis for the preparation of the traffic loads model in EN 1991-Part 2 has
been developed in parallel at various locations in Europe with studies
performed at SETRA, LCPC, University of Pisa, University of Liege, RWTH
Aachen, TU Darmstadt, Flint & Neill, London that comprised the following
steps:
Step 1:

Evaluation of measured data related to the composition of traffic,


traffic density, axle loads, vehicle loads, inter-axle distances and
inter-vehicle distances.
These evaluations gave the basis for the choice of the
geometrical pattern of the traffic load model being composed of a
group of single loads and uniformly distributed loads suitable to
determine the effects of traffic both for local assessments of the
components of the bridge deck and for global verifications of the
main structure of the bridge.

Step 2:

Calculative determination of the magnitudes of the loads in the


traffic load model by simulating traffic effects with a static
simulation dynamic model on one side and a dynamic simulation
model on the other side, that comprised a dynamic model for the
bridge and their parts, models for the various vehicles as rigid
body kinematic systems and a model for the surface of the bridge
including the surface roughness and its effect on excitation of the
axles of the vehicles.
For consistency reasons the static simulation model was run with
dynamic magnification factors obtained from the dynamic
simulation model.

Step 3:

Comparison of the results of static and dynamic simulations to


agree on conclusions for the load model in the Eurocode.

Step 4:

Probabilistic limit state assessments to determine recommended


values for Nationally Determined Parameters as the partial
factors for the traffic load model.

(2)

In the following the works executed in these steps are reported.

3.

Evaluation of measured traffic data to construct the traffic load model for
ULS assessments
History of measurements

3.1
(1)

Until 1970 singular vehicles were taken from the flowing traffic and weighed on
movable weigh-bridges, so that the statistical relevance of data was doubtful.

(2)

Since 1970 the LCPC used weigh-bridges for weigh in motion measurements
so that automatic records of axle loads with time gaps were possible.

(3)

In 1975 to 1978 a measuring campaign was carried out supported by the


ECSC to determine axle loads and inter-axle distances for fatigue
assessments of steel bridges.

(4)

Since 1980 the LCPC has worked with piezo-electric measuring methods to
simplify the measurements.

(5)

From 1988 on the data available from measurements in Europe were collected
by the Eurocode-working group to identify the development and distribution of
the vehicle weights.

3.2
Results of the measurements and evaluations
3.2.1 Choice of representative traffic
(1)

The fig. 3-1 gives a survey on the measurements that were carried out a
different European locations, which comprise different types of roads as

highways
other roads
urban roads

with different types of traffic as

long distance traffic


short distance traffic
special traffic, e.g. in the vicinity of gravel pits, quarries, etc.

Figure 3-1: Measurements of traffic at various locations in Europe


(2)

Significant are differences in the average daily traffic density (number of


vehicles/24 h), e.g. for Doxey (UK) with 34.500 vehicles/24 h. Important for the
extreme traffic situations are however roads with an extreme density of heavy
vehicles, for which fig. 3-2 gives an indication.

(3)

Fig. 3-3 shows the accumulated distribution of the axle loads as measured. In
this figure n10 is the number of axle loads with PA 10 kN. In fig. 3-4 also the
accumulated distribution of total vehicle loads is given, where n30 is the
number of vehicles with G 30 kN.

(4)

It is apparent, that the traffic at Auxerre does not exhibit the largest axle loads
but the largest frequency of large axle loads; this is in particular caused by the
large frequency of articulated vehicles in the traffic, see fig. 3-1.

Figure 3-2:
(5)

Percentage of types of vehicles

In conclusion the data from the Auxerre traffic were selected as the basis for
the development of the Eurocode traffic load model.

Figure 3-3:

Comparison of accumulated densities of axle loads


4

Figure 3-4:
(6)

Justification for choosing the Auxerre traffic data as basis for the
European traffic loading model

In order to produce the pattern of the Eurocode traffic-load model more


detailed measurements were necessary to complete the informations on
-

the frequency distribution of axle loads,


the frequency distribution of the distances between axles,
the frequency distributions of different types of vehicles,
the frequency distribution of distances between vehicles.

3.2.2 Measurements of axle-weights and conclusion for axle loads


(1)

Fig. 3-5 gives as an example the frequency distribution of loads for axle no 2
of an articulated vehicle in Rheden.
The frequency distribution is bimodal caused by the frequencies of unloaded
and loaded vehicles.

(2)

These frequency distributions can be approximated by Rayleigh-distributions


which are close to normal distributions for large values.

Figure 3-5:

Frequency distribution of weights of axle no 2 of articulated


vehicles in Rheden
5

(3)

From the extreme values PA 14 kN an extreme value distribution as in fig. 36 for the Caronte-bridge can be derived that allows to extrapolate
representative values for the code.

Figure 3-6:
(4)

Extreme value distribution and extrapolation

From the half normal distribution y

y=

1
2

Z2
2

where
z=

x xo

x=Q
x o = Qo

=
f

( x x o )2 f

is the variable of distribution


is the variable of weight Q
is the average value of weight Q
is the standard deviation
is the frequency of x

and the following input values from measurements

Qo = 90 kN
= 32 ,64 kN

the following representative values can be derived


daily extreme:
annual extreme:
1000 years extreme:
(5)

Such extrapolations have been carried out for various locations where
measured data were available, see fig. 3-7.

Figure 3-7:
(6)

195 kN (measured 1 x /day)


236 kN
279 kN (characteristic value).

Characteristic values for single axle loads and tandem, tridem


and vehicle loads

On the basis of fig. 3-7 the characteristic value of the axle load in Load-Model
1 LM1 of EN 1991-2 was taken as
Q = 300 kN

It is also the basis for the axle load in Load-Model LM2, Q = 400 kN .
(7)

The model for the axle load in LM1 includes a certain dynamic factor resulting
from the roughness of the road surface where the measurements were made.
The magnitude of the dynamic factor has been determined according to fig. 38 from dynamic simulations of the flowing traffic at the points of
measurements.

Figure 3-8:
(8)

Determination
measurements

of

dynamic

factor

implicite

in

weight

For Load-Model LM2 the amplification factor = 1.14 for axle loads was
considered to be not sufficient. From dynamic simulations with a local
irregulatory as given in fig. 3-9 on additional amplification factor of 1.3 was
obtained, that leads to the value

Q = 1.3 300 400 kN

Figure 3-9:

Impact factor from irregulatories on the road


8

(9)

The following conclusions can be drawn:


-

the characteristic value of the axle load in LM1 is about 2,5 times
higher than the legally permitted load. This is a result of systematic
overloading,
the characteristic value of the axle load in LM1 differs from location to
location,
the variation of the characteristic value of the axle load in LM1 with
time is small.

3.2.3 Measurements of distance of axles


(1)

To choose an appropriate distance of axle loads in the traffic load model


frequency distributions of distances were determined for articulated vehicles,
see fig. 3-10.

Figure 3-10: Frequency distributions of distances


articulated vehicles for Rheden

between

axles

for

(2)

According to fig. 3-10 the distances between axles 1 and 2 are in the range of
x 1,20 m .

(3)

Therefore for the double axle in Load-Model LM1 a distance of 1,20 m was
chosen.

3.2.4 Measurements of vehicle weights


(1)

For determining the effects of traffic for lengths of influence areas greater than
10 m the statistics of vehicle loads and of the inter-vehicle distance are
necessary.
9

(2)

As reference location for the measured data Auxerre was chosen, because of
the following reasons, see fig. 3-1:
-

(3)

the composition of the traffic corresponds to the estimation of future


trends,
the portion of lorries in the traffic composition is 32% in lane 1 and
10% in lane 2 and in relation to other locations rather high,
the portion of loaded lorries from all lorries is 66% and hence mirrors
the trend for an improved transport management,
data were fully documented for a large time period for lane 1 and lane
2 in a 4-lane highway.

In fig. 3-11 the most important types of vehicles from the full traffic are isolated
as given in fig. 3-2 for which this figure also shows the distribution of these
vehicle to lane 1 and lane 2 if the full lorry traffic is reduced to these 4 types of
vehicles.

Figure 3-11: Reduction of full lorry traffic to 4 important types of vehicles and
distributions of these types to lane 1 and lane 2 of the 4 lane
highway at Auxerre
(4)

Fig. 3-12 gives the distributions of weights for those 4 vehicle types.

10

Figure 3-12: Distribution of vehicle weights


(5)

The most important vehicle type is type 3 (articulated vehicle), which is


dominant both for ultimate limit state and fatigue verifications.

(6)

The distributions in fig. 3-12 can be approximated by 2 normal distributions as


illustrated for type 3 vehicle in fig. 3-13, one for light, one for heavy traffic.

Figure 3-13: Approximation


distributions
(7)

of

distributions

as

measured

by

normal

In fig. 3-14 the statistical data of the distributions are given from the data as
measured. In fig. 3-15 the statistical data are corrected after filtering out the
dynamic effects. This filtering influences mainly the standard deviations.

11

Figure 3-14: Statistical data for the lorry-traffic in lane 1 and lane 2 at Auxerre
from measured data

Figure 3-15: Statistical data of lorry-traffic in lane 1 and lane 2 at Auxerre after
filtering out dynamic effects
(8)

Furthermore the distributions of the vehicle loads on the various axles and the
distribution of the distances of axles can be described in statistical terms, see
fig. 3-16.

12

Figure 3-16: Distribution of vehicle loads on axles and of distances between


axles in statistical terms
(9)

Such descriptions of data are necessary to use them


-

for generating artificial traffic situations for dynamic simulations of


bridge responses
to extrapolate representative values of maximum vehicle weights from
the data available.

(10)

A list of extrapolations of maximum vehicle loads is given in fig. 3-7 with


characteristic values in the range up to 960 kN for a mean return period of
1000 years.

(11)

From the Auxerre traffic a characteristic weight of the type 3-vehicle can be
determined in this way with a total weight of 880 kN, see fig. 3-17.

Figure 3-17: Characteristic axle loads for type 3-vehicles


(12)

For lengths of influence areas exceeding the lengths of vehicles particular


studies must be carried out to determine the relevant traffic load models from
action effects in bridges.

(13)

To this end traffic simulations have been carried out using a two lane box
girder type bridge and shapes of influence lines along the bridge axis with
varying values of lengths as indicated in fig. 3-18.

13

Figure 3-18: Simulation of traffic effects with influence lines


(14)

The simulation was carried out with the following conditions:


- flowing traffic as recorded for lane 1 in
(25% lorries, 75% cars)
- congested traffic: only 100 % lorries as
recorded
in lane 1 in Auxerre
Traffic on two lanes:
- flowing traffic as recorded for lane 1 in
Auxerre (for each lane)
(25% lorries, 75% cars)
- congested traffic as recorded in lane 1 in
Auxerre
(25% lorries, 75% cars)
Traffic on three or four lanes: - flowing traffic artificially generated from
Auxerre (for each lane)
data (10% lorries, 90% cars)
- congested traffic artificially generated from
Auxerre data (10% lorries, 90% cars).
Traffic on one lane:
Auxerre

(15)

These conditions are by 1 step more severe than the conditions measured in
Auxerre. One could imagine that for bridges with 2, 3 or 4 lanes in each lane
congestions of only lorries could occur. This case has however not been
considered in the loading model because of the low probability of occurrence.
14

3.2.5 Definition of flowing and congested traffic


(1)

In congested traffic situations the inter-vehicle distance was assumed to be 1


m (3 m between axles for jam).

(2)

For defining flowing traffic the distances between vehicles depending on the
speed must be known.

(3)

A possibility to define the density function for the inter-vehicle distance is by


the density function y of Davenport
y=

k
x k 1 e x
(k )

where
x=d
xo = d o
xn = d n

is the distance between the vehicles


is the mean value of distance
is the modal values corresponding to the maximum frequency

1
xo xn
xo
k=
xo xn

(k )

is the Gamma-function of k

Hence the function is controlled by two parameters:


the mean value d o and
the modal value d n .
(4)

Fig. 3-19 gives distributions of measured distances and the calculative


distribution using d o = 120 m and d n = 30 m.

15

Figure 3-19: Distribution of distances between vehicles


(5)

Another possibility for density distributions is given in fig. 3-20, where the
constant part between 20 m and 100 m covers the probability of development
of convoy and a linear increase up to 20 m is due to the minimum distance.
The exponentially decreasing part for distances greater than 100 m covers
free flowing traffic.

Figure 3-20: Comparison of measured and theoretical values for density


function of inter-vehicle distances
(6)

The -value in fig. 3-20 gives the probability of occurrence for lorry distances
less than 100 m, and the -value has been obtained from traffic records of 24
representative traffics in Germany.

(7)

A simplified solution that could also be used for the inter-vehicle distance is
the minimum distance that results from the reaction time of a driver to avoid a
collision with the front vehicle in case of braking. Assuming a minimum braking
reaction time Ts = 1 s of the driver, the minimum distances is give by
16

a=

v
Ts

where v is the mean speed of the vehicles. The distance is limited to a


minimum value a = 1 m in case of jam situations.
3.2.6 Results of simulations with influence lines
(1)

From the simulations that have been carried out to obtain-load effects, e.g. the
bending moment M or the shear force V , the equivalent load Q effecting
these load effects can be determined by
Q' =

M
k
L

or Q' = V k

where
k

(2)

is a factor resulting from the influence line considered.

Fig. 3-21 and fig. 3-22 give these characteristic equivalent loads for traffic on
one lane and the associated values from Load-Model LM1.

Figure 3-21: Characteristic values of equivalent loads Q' determined from


traffic simulations for the mid-span bending moment of a single
span bridge for 1 lane traffic

17

Figure 3-22: Characteristic values of equivalent loads Q' determined from


traffic simulations for the hogging moment of a two span
continuous bridge for 1 lane traffic
(3)

Fig. 3-23 gives the single loads Q [kN] and the uniformly distributed loads q
[kN/m] that result from various influence lines.

Figure 3-23: Characteristic value of traffic loads for 1 lane traffic


(4)

Fig. 3-24 demonstrates the effects of the Auxerre traffic as measured (25 %
lorries, 75 % cars), that can be represented for flowing traffic by
QK = 800 + 12 L

and for congested traffic


18

QK = 800 + 31L

Figure 3-24: Global characteristic loads Q'


(5)

For traffic on more than 1 lane the studies have shown, that for L > 30 m
always congested traffic is relevant.

(6)

Fig. 3-25 and fig. 3-26 give the equivalent loadings for 2 lane traffic.

Figure 3-25: Characteristic values of equivalent loads Q' determined from


traffic simulations for the sagging moment of a single span bridge
for 2 lane traffic
19

Figure 3-26: Characteristic values of equivalent loads Q' determined from


traffic simulations for the hogging moment of two span bridges
for 2 lane traffic
(7)

Fig. 3-27 and fig. 3-28 give the relevant equivalent loads for 4 lane traffic.

20

Figure 3-27: Characteristic values of equivalent loads Q' determined from


traffic simulations for the sagging moment of single span bridges
for 4-lane traffic

Figure 3-28: Characteristic values of equivalent loads Q' determined from


traffic simulations for the hogging moment of two span
continuous bridges for 4-lane traffic
(8)

For other influence curves the results are in between these extreme values.

(9)

Fig. 3-29 gives the single loads Q [kN] and the uniformly distributed loads q
[kN/m] that result from various influence lines and numbers of lane.

Figure 3-29: Characteristic values of traffic loads for 2, 3 and 4 lane traffic

21

(10)

The fig. 3-30 and fig. 3-31 demonstrate the effect of the characteristic vehicle
in fig. 3-16 on the equivalent value Q' . Such a vehicle could be used for the
assessment of existing bridges.

Figure 3-30: Effects of the characteristic value in fig. 3-16 on equivalent


values Q' for sagging moments

Figure 3-31: Effects of characteristic vehicle in fig. 3-16 on equivalent values


Q' for hogging moments
22

3.2.7 Composition of load models LM1 and LM2


(1)

The composition of Load-Model LM1 as concluded from the evaluations of


measurements is given fig. 3-32.

Figure 3-32: Load-Model LM1 in EN 1991-2


(2)

Using a reference length of 11 m it can be interpreted as a simultaneous


assembly of

a 900 kN-vehicle in the first lane in a row of 450 kN vehicles in a jam


situation with 5 m inter-vehicle distance
a 500 kN-vehicle in the second lane with 120 kN-vehicles in a row
a 300 kN vehicle in the third lane with 120 kN vehicles in a row

where the 120 kN vehicles may also be understood as a mixture of 450 kN


lorries with personal cars.
(3)

The Load-Model LM2 is simply a single axle representing the impact effect of a
characteristic axle load from irregularities on the road surface.

4.
4.1

Representative values of traffic loads for SLS assessments


General

(1)

Other representative values of traffic loads than characteristic values are non
frequent values, frequent values and quasi permanent values.

(2)

The non frequent or rare values are those having a mean return period of 1
year from either flowing or congested traffic, e.g. considered for the limit state
of decompression in concrete.
23

(3)

The frequent values have a mean return period of 1 week. They are based
on frequent traffic situations as e.g. flowing traffic and good surface quality of
road. The extreme situations applied for the determination of characteristic
loads are not considered for frequent values.

(4)

For quasi-permanent values of traffic (e.g. for creeping effects) usually the
values are zero.

(5)

To determine the magnitudes of representative values the half-normal density


distribution y for extreme action effects is used as for characteristic values:
1

y=

z2
2

where
z=

x x0

is the reduced variable

is the effect considered


is the particular average value of x used in fitting to the real
distribution curve
is the standard deviation.

x
xo

The values x o and are adjusted in such a way that the correlation coefficient
is maximized.
(6)

The value corresponding to a return period R is

x R = x0 + z R = x0 1
z R
x0

where z R results from the following equation


Y (z R ) =

1
2

zR

z2
2

dz =

1 Ts
2 N S TR

where:
NS =

TS
TR

is the number of vehicles corresponding to xi > x o and the


simulated time period
is the simulated time period
is the required return period.

4.2

Results of extrapolation

(1)

A typical example for such a simulation is given in fig. 4-1.

24

Figure 4-1:

(2)

Fig. 4-2 gives in the scale of half-normal distributions straight lines with
intersection points for the return periods required, where xo and hence z o are
adjustment values for the effect considered (best fit parameters, see fig. 4-1).

Figure 4-2:

(3)

Example for the approximation of the distribution of bridge


response by a half-normal distribution and determination of
representative values

Identification of Y (z ) depending on the return period required for


different curve fitting parameters

Fig. 4-3 gives the results of many simulations of the Auxerre traffic. In this
figure x K corresponds to a return period of 1000 years. The factor

2N s
nL

25

where n L is the number of lorries used in the simulation varies between 1 and
4, but z R various only by 10 % for short return periods (1 day) and by 5 % for
long return periods (1 year).

Figure 4-3:

Factors

xR
xk

to

determine

representative

values

from

characteristic values of traffic effects


4.3

Conclusions

(1)

Taking account of the results in fig. 4-3 the following choices have been made
in EN 1991-2:
-

1 (non frequent) = 0,80

In this choice a reduction of about 10 % from 1 = 0 ,92 has been made


to take account of advantages of good surface quality (better than for
characteristic values) for the mayor part of service life and the fact that
characteristic values for large spans in EN 1991-2 are conservative.
-

1 (frequent) = 0,75 for axle loads Qk


= 0,40 for uniformly distributed loads q k

In this choice also a reduction of about 10 % from 1 = 0,82 for flowing


traffic and 1 = 0,85 for congested traffic has been made due to good
surface quality of for the mayor part of service life.
Also the frequent values from flowing traffic are not greater than 50 % of
the frequent values for congested traffic, because the characteristic
values differ by this amount, see fig. 3-23.
The values 1 for axle loads and uniformly distributed loads are also
quite close to the values for fatigue-frequent loading model FLM1 for
fatigue assessments.
26

5.
5.1

Fatigue loading models


Basic principles of fatigue

(1)

Fatigue is the progressive, localised and permanent structural change


occurring in materials subjected to fluctuating stresses initiating and
propagating cracks through a structural part after a sufficient number of load
cycles.

(2)

Fatigue is induced in bridges mainly by heavy vehicles. The development of


appropriate load models and verification concepts is a main topic of EN 1991-2
and the bridge parts of the other Eurocodes.

(3)

The same reasons that apply for the choice of the Auxerre traffic for the
development of the load models LM1 and LM2 also apply for the preparation
of the Load-Models for fatigue.

(4)

In principle the way of determining the fatigue load models is as follows:


1.

Choose typical bridges for simulating bridge responses to traffic flow


either by static simulations using influence lines with dynamic factors
or by dynamic simulations with appropriate vibrating systems
comprising the mass and damping systems of the bridges, the vehicles
and the surface roughness of the road surface.

2.

Evaluate the bridge-responses documented as stress-histories by an


appropriate counting method to achieve a histogramme of the stress
ranges applicable as fatigue actions to the material.

3.

Apply an appropriate damage accumulation rule (in general the Miner


rule) to transfer the fatgiue actions expressed by a histogramme with
varying stress ranges into a histogramme (or spectrum) with constant
stress ranges (damage equivalent stress ranges), to make the fatigue
actions comparable with the fatigue resistances.

4.

Select an appropriate fatigue resistance curve (Whler curve) for the


material and structural detail chosen that originates from large scale
test specimen fabricated in the way as the real structure and that
constitutes a characteristic value of fatigue resistance.
In general such fatigue resistance curves have been determined for
constant amplitude stress-ranges.

(5)

5-

Apply a reliability concept to achieve sufficient reliability for the fatigue


assessment.

6.

Simplify the procedure by producing damage-equivalent fatigue loads,


the effects of which can be more easily compared with the fatigue
resistances.

Fig. 5-1 gives the main steps of this procedure .


27

Figure 5-1:
(6)

For steel structures and for reinforcing and prestressing steel the fatigue
strength curves according to fig. 5-2 may be used.

Figure 5-2:
(7)

Fatigue assessment: from influence lines to damage assessment

Fatigue strength curves for structural steel and reinforcement


and prestressing steel

The fatigue strength curves for welded steel structures are defined by the
fatigue strength category c (fatigue strength at 2 106 cycles) and the
constant amplitude fatigue limit D at 5 106 cycles.
For stress ranges above D the slope m of the curve in a double logarithmic
scale is m = 3 . Stress ranges below D dont produce fatigue if the maximum
stress range max D .

28

For stress spectra with stress range above D and below D the damage
accumulation may be performed with a modified fatigue resistance curve with
a slope m = 5 for stress ranges below D . This modification takes into
account that the constant amplitude fatigue limit D is reduced by damage
effects from stress ranges > D
For stress ranges of the stress range spectrum which are below the cut-offlimit L at 108 cycles it may be assumed that these stress ranges do not
contribute to the calculated cumulated damage.
(8)

Typical examples for fatigue strength categories in steel and composite


bridges are shown in fig. 5-3. The stress ranges in this figure relate to nominal
stresses.

Figure 5-4:

Typical examples for fatigue strength categories in steel and


composite bridges

(9)

For reinforcement the fatigue strength curve is given in EN 1992-1-1 and


described by a two linear functions in the double logarithmic scale without any
constant amplitude fatigue limit, see fig. 5-2.

(10)

Whereas for steel structures normally a linear relation can be assumed


between the fatigue loading and the stresses, for concrete structures the nonlinear behaviour due to cracking of concrete has to be taken into account for
the determination of the time history of the stresses. In this case in addition to
the fatigue load also the permanent loads and effects from climate
temperature actions have to be considered.

(11)

In the following the effects of the counting method and of the damage
accumulation hypothesis (Miner Rule) are explained in more detail.

5.2

Counting method

(1)

The following counting methods give the same results and can also identify the
mean stress level for each cycle
-

the reservoir method,


29

(2)

the rainflow method

The reservoir method, see fig. 5-5 and also fig. 5-1 defines a cycles by the
difference of water level in emptying a reservoir. Secondary stress ranges
originate by water pockets left when starting with emptying the reservoir at the
lowest point. The method is easily understood but difficult to programme.

Figure 5-5:
(3)

Reservoir method

A method more suitable for programming is the rainflow method, see fig. 5-6.
This method follows the flow of rain-drops from the top of a fictive roof as
explained by the example in fig. 5-6.

Figure 5-6:

Rainflow counting method

30

5.3

Damage accumulation hypothesis

(1)

For a given stress-range histogramme (spectrum) the application of the MinerRule for a stress spectrum that has stress ranges above and below the
constant amplitude endurance limit D , is demonstrated in fig. 5-7.

Figure 5-7:

(2)

Linear damage accumulation and equivalent amplification factor

fat .

The damage accumulation related to steel structures with m1 = 3 and m 2 = 5


therefore reads:
D=

ni 23


i>

n D D3


i>

ni i5
L

n D D5

(3)

A damage accumulation can be neglected if the frequent values max do


not exceed D .

5.4

The concept of damage equivalence

(1)

The use of the Miner-rule together with the fatigue resistance curve allows to
transfer any stress range spectrum with variable stress ranges into a damage
equivalent spectrum with constant stress-ranges e .

(2)

For a fatigue resistance curve with an unlimited slope m and a spectrum with

(
D=

m
i

ni

2 10
m
c

a damage equivalent constant amplitude spectrum with e and ne results


from:

31

(
D=

m
i

ni

cm 2 10 6

em ne
cm 2 10 6

and reads
e = m

(3)

m
i

ni

ne

In taking into account a linear relationship between a histogramme of loads


S i , ni and the associated histogramme of stress-ranges i , ni , it is possible
to adopt the notion of damage-equivalent stress-ranges to damage equivalent
loads.
Fig. 5-8 shows for the example of axle-loads measured for the traffic in
Rheden the distribution of axle loads (S i , ni ) and the distribution of damages
Di calculated by
Di = S im ni

that for the reference number ne gives


S e = m

ne

(4)

Hence it is possible to calculate fatigue damage equivalent loads from load


histogrammes on the basis of the shape of the fatigue resistance curve without
knowing the relevant fatigue detail class.

(5)

Fig. 5-8 shows that very high loads do not contribute to damage because of
their small number, and small loads with large numbers do not contribute
because of their small amplitudes.

32

Figure 5-8:

(6)

Distribution of fatigue damage from a load histogramme and


determination of fatigue resistance load

There are various possibilities to define the damage equivalent load ranges
S e depending on the choice of ne .
Examples are:
-

reference to the total number of cycles ne = ni which gives


S en = m

D
n

reference to the maximum value of load ranges S e = max S which


gives
ne =

max S m

reference to the most damaging load range S eg which gives


S eg =

(D S )
D
i

33

neg

D
=
S egm

D
=
( D S )
m +1

eg

S eg is equal to the mean values (gravity centre)of the damage distribution Di .

(6)

Fig. 5-8 gives the results of a damage calculation for an axle load
histogramme and shows the values S eg and S e .
The definition S eg gives larger values than S en resulting in less conservatism
in particular where the values of S e are close to L , see fig. 5-9 and fig. 510.

(7)

All definitions of S e are equivalent and based on the same fatigue damage
S em ne = Di = D

Figure 5-9:

Errors in using an unlimited slope m = 3 instead of trilinear fatigue


resistance curve

34

Figure 5-10: Errors in damage calculation when using an unlimited slope


m = 3 instead of trilinear fatigue resistance curve
5.5

Results of evaluation of load-distributions

(1)

The evaluations were performed on the basis of load distributions as given in


fig. 5-8 to determine the following values:
nL

number of lorries per day

n1 , n 2 , n 3

number of single axles, tandem axles or tridem axles

n30

number of single axles exceeding PA = 30 kN

Qd

loads with a return period of 1 day

Qf

loads, the exceedance of which would produce a damage

Qg

of less
than 1 % of the total damage. This load can be
interpreted as the frequent value for fatigue
=
load level that produces the maximum damage Di in the
Di -

ng

signifies
the damage equivalent load
=
number of cycles that together with Q g produce the total
the

(2)

damage distribution. Together with the number n g it

fatigue damage D on the basis of an unlimited slope of


resistance curve m = 3

The results of the evaluations are given in fig. 5-11 for single loads, fig. 5-12
for tandem loads, fig. 5-13 for tridem loads and in fig. 5-14 for the full vehicles.

35

Figure 5-11: Evaluation of distribution of single loads

Figure 5-12: Evaluation of distribution of tandem loads

36

Figure 5-13: Evaluation of distribution of tridem loads

Figure 5-14: Evaluation of distribution of vehicles loads


(3)

Results of the evaluations are the following


1.

for frequent values for fatigue Q f :


single axle: 200 kN
tandem axle: 300 kN = 2 x 150 kN
37

tridem axle: 385 kN = 3 x 130 kN


vehicle:
600 kN
2.

for damage equivalent loads defined by the maximum damage of


damage distribution Q g , n g :
single axle: 131 kN
ng
nL

= 1.37 ;

ng

0.43 0.5

n30

tandem axle: 252 kN = 2 x 125 kN


ng
n2

= 0.42 0.5

tridem axle: 266 kN = 3 x 90 kN


ng
n3

vehicle:

469 kN
ng
nL

3.

= 0.75

= 0.43 0.5

for damage equivalent loads defined by the number of vehicles n L ,


calculated from Qe = Q g 3

ng
n ref

n ref
nL

single axle: 151,10 kN


tandem axle: 120,00 kN
tridem axle: 183,00 kN
vehicle:
378,00 kN
(4)

EN 1991-2 specifies as fatigue Load-Model LM3 a four axles vehicle, see fig.
5-15-

Figure 5-15: Fatigue LM3 in EN 1991-2

38

(5)

This model has a total weight of 480 kN and therefore is slightly heavier than
the most damaging equivalent fatigue load of 469 kN in Auxerre, see fig. 5-14.
This model has 4 axles, as 4 axles represent a lorry in the mean, see

n1
in
nL

fig. 5-11.
The tandem load of 240 kN corresponds approximately to the most damaging
equivalent load for tandem axles 252 kN, and the single axle load of 120 kN is
approximately the value of the most damaging single load of 131 kN
When compared with the damage equivalent loads defined by the number of
vehicles Load-Model 3 is conservative except for single axles.
(6)

For the fatigue assessment of details the stress-situation of which is influenced


by local loads as from single axles the fatigue loading model LM3 is normally
not precise enough, e.g. for bridge decks made of concrete or orthotropic steel
decks.
For that case EN 1991-2 has specified the fatigue load models LM2 and LM4,
where LM2 gives the frequent values and LM3 gives the damage equivalent
loads for 5 important vehicle silhouettes, see fig. 5-16.
The silhouettes give the mean value geometry, and the axle loads reflect more
precise load-distributions for single axles, covering also the maximum value of
the damage equivalent load Qe = 151 kN, determined for ne = n L .

(7)

These load models LM2 and LM4 which are assembled in fig. 5-16 give also
more precise contact surfaces of the wheels needed for local assessment, see
fig. 5-17.

39

Figure 5-16: Fatigue load models LM2 and LM4

Figure 5-17: Dimensions of pressure areas of wheels


(8)

The damage equivalent load Q g representing the most damaging weight has
the advantage, that it is less sensitive to a variation of the slope of the fatigue
resistance curves than the value Qe related to ne = ni , see fig. 5-18.

40

Figure 5-18: Histogramme of axle loads and associated distribution of


damage Di for m = 3 and m = 4

5.6
Simulation of bridge responses for static fatigue actions
5.6.1 Purpose of the simulation
(1)

The simulation of bridge responses to traffic, that has been performed for
determining characteristic values of load effect, as described in 3.26 has also
been used to determine the frequent Fatigue Load-Model (with 1 % fatigue
damage only) and to check the damage equivalent fatigue loading models as
described in 5-5 in view of the dependency on the shape and size of influence
lines.

(2)

To this end the following loading scenarios have been used:

AMT
AJ 1T

as recorded in Auxerre on lane 1 (25 % lorries and 75 % cars)


the same traffic as AMT , but with an inter-vehicle distance of
1m
41

(3)

As the recorded axle loads from Auxerre contain already a certain amount of
dynamic impact from good surface quality, no additional dynamic impact
factor was applied, see fig. 5-19.

Figure 5-19: Damage equivalent amplification factors for good pavement


quality
(4)

The results of the simulation of damage equivalent loads are valid for lengths
of influence lines L > 10 m , so that local effects of single axles are not
considered but only effects from full vehicles.

(5)

As fatigue is only relevant, where a sufficient ratio of variable loads to


permanent loads, applies, accuracy of the damage equivalent loads is only
required for influence lines with small and medium spans. Fig. 5-20 gives an
indication on the fatigue critical field of span length and deck width of bridges.

42

Figure 5-20: Example for calculative service life of composite box girder
bridges for different span length and deck widths for (1)
continuous girders with reinforcements at supports, (2)
continuous girders, (3) single span girders
5.6.2 Results for fatigue loads dependant on span L
(1)

Fig. 5-21 gives the total frequent fatigue load Q f and the damage equivalent
load Q g related so the length L of single span bridges for the flowing traffic and
the congested traffic chosen.

Figure 5-21: Total fatigue loads on span L


43

(2)

The frequent fatigue load is composed of a single load of 600 kN which is the
frequent load of a lorry and a uniformly distributed load that depends on the
type of traffic being either 9 kN/m for flowing traffic or 22 kN/m for congested
traffic.
If frequent loads for fatigue are related to flowing traffic only, then the
frequent values can be defined by coefficients applied to the characteristic
traffic loads:

1Q

to axle loads Qk :

1q

to uniformly distributed loads q k

1 = 0 ,8 (~ 0 ,75 )
1 = 0 ,30 (~ 0 ,40 )

The reduction of 1q in relation to 1Q is caused by the increase of distances


between vehicles.
(3)

The values 1Q and 1q for frequent loading for fatigue are very close to the
frequent values defined by a return period of 1 week in 4.3 (see values in ( )).

(4)

Fig. 5-22 gives a comparison of frequent and damage equivalent fatigue load
models from fig. 5-19.

Figure 5-22: Comparison of frequent and damage equivalent fatigue load


models
(5)

Figure 5-21 also gives the damage equivalent loads Q g , with the maximum
damaging effect together with

ng
nL

and the damage equivalent loads Qe related

to n L for the Auxerre traffic.


44

It is evident, that Q g for a span length of 10 m corresponds to the damage


equivalent load determined from the Auxerre vehicle loads, see fig. 5-14, that
is constant for a length up to L = 50 m , if flowing traffic is used as the basis.
For congested traffic the maximum span length is about 20 m.
(6)

Fig. 5-23 shows for traffic other than in Auxerre that for fatigue the critical intervehicle distance is not as for congested traffic, e.g. for a inter-vehicle distance
of 4,8 m, but in the range of 24 m.

Figure 5-23: Damage equivalent loads dependant on inter-vehicle distance


(7)

ne
represent more or less the ratio of the number ne of loaded
nL
lorries to the total number n L of lorries.

The values

For larger span lengths L > 50 m resp. L > 20 m the fatigue load grows slowly,
see figure 5-23, so that an adjustment is necessary.
5.6.3 Comparison of the fatigue effects of the Auxerre traffic with the fatigue
load models in EN 1991-2
(1)

To check the effects of the various Load-Models in EN 1991-2 a comparison is


made between the effects of these load models and the fatigue effects of the
Auxerre traffic.

(2)

Fig. 5-24 gives the ratio of moment ranges due to Fatigue Load Model
(LM1) M f ,LM 2 which is the frequent fatigue model related to the
characteristic Load-Model (LM1), and the frequent fatigue value M fA from
the Auxerre traffic, for different slopes m
45

Figure 5-24: Comparison FLM1 and Auxerre-traffic


(3)

For the influence lines M 0 , M 1 , M 3 the Eurocode-model is safe-sided, and for


L 20 m also sufficient, whereas for influence line M 2 the model is not safe.

(4)

Fig. 5-25 gives the ratio

M fLM 2
M f , A

for the fatigue load model LM2 for frequent

fatigue load. The greater precision of this model is reflected by a smaller


scatter for the various influence lines in the range L 20 m . It needs however
an adjustment by an impact factor of about 1,10 to be safe sided.

46

Figure 5-25: Comparison FLM2 and Auxerre traffic


(5)

Fig. 5-26 gives the ratio

M e ,LM 3
for ne = n L
M e , A

Figure 5-26: Comparison FLM3 and Auxerre traffic for ne = n L

47

(6)

It is evident that FLM3 is safe sided for the influence lines M 0 , M 1 , M 3


however unsafe for larger spans for influence line M 2 .

(7)

Fig. 5-27 shows

M g ,LM 3
M g ,A

if ne is modified by

ne = k 1 k 2 k 3 n L
where
k1 =

percentage of loaded lorries, for Auxerre k 1 = 2 / 3 (highest value so far


observed)

k2 =

correction for span length L < 10 m to consider the influence of axles on


amplitudes and number of cycles e.g.
k 2 = 0 ,6 +

k3 =

4
for 1,18 L 10 m
L

correction for influence line M 2


k 3 = 1 + 0 ,01 (L 10 ) for 10 m < L < 100 m

Figure 5-27: Comparison FLM3 and Auxerre traffic for ne = k 1 k 2 k 3 n L

48

(8)

It is evident that for LM3 and ne = k 1 k 2 k 3 n L the scatter and the influence of m is
smaller; however the use for influence line M 2 is still unsafe for m = 5 and
m = 9.

(9)

In order to correct the situation for influence line M 2 a second FLM3-vehicle is


necessary as specified in fig. 5-28 with 40 m distance to the first vehicle

Figure 5-28: Supplement to FLM3


(10)

This additional vehicle does not modify the values as given for influence lines
M 0 , M 1 , M 3 but improves substantially the results for influence line M 2 , see
fig. 5-27, in which the second vehicle has already been taken into account.

(11)

The fatigue load model FLM4 gives 5 types of equivalent vehicles, see fig. 5M e ,LM 4
16, for which the ratios
, are given in fig. 5-29.
M e , A

Figure 5-29: Comparison FLM4 and Auxerre traffic


(12)

This model is mainly used for local assessments as in orthotropic steel decks.

49

5.7

Conclusions

(1)

EN 1991-2 gives fatigue loads, from which FLM1 and FLM2 are mainly
intended to perform a check for infinite durability:
frequent D

This check is always relevant for a large number of cycles as e.g. experienced
in short spans and structural components of the deck.
Note: If each lorry produces a stress range cycle this would lead to 1000 to
8000 cycles per working day, e.g. 25 to 200 millions of cycles for 100 years.
For small components loaded by axles or wheels this number is even higher.
(2)

Fig- 5-30 shows the cut-off by the fatigue frequent Load-Models FLM1 and
FLM2.

Figure 5-30: Cut-offs for fatigue load range spectra


(3)

The damage-equivalent Load-Models FLM3 and FLM4 are intended to assess


a specified limit of service life of the bridge.
To this end FLM3 is for main components of the bridge with influence lengths
L 20 m whereas FLM4 is for deck-components with L 20 m .

(4)

The fatigue Load-Model FLM3 needs a supplementary vehicle to cope with


larger spans and with influence line M 2 applicable for hogging moments
above supports.

(5)

The supplementary vehicle to FLM3 and any preparation of time histories and
cycle counting can be avoided, if FLM3 is used together with damage
equivalent factors , directly obtained from the Auxerre traffic for different
types of influence lines.

(6)

Any fatigue assessment both for infinite durability and limited service life needs
partial factors for reliability sake. These partial factors are specified in the
material related fatigue codes or bridge design codes.

50

(7)

A feature that controls the value of the partial factor is robustness for fatigue
damages which is expressed in terms of damage-tolerance. The relationship
between several levels of damage tolerance and the partial factors is specified
in the material related fatigue codes.

6.

Procedure for the dynamic analysis of bridges under the loads of single
vehicles
General

6.1
(1)

This section deals with a procedure used to determine the dynamic effects of
traffic loads in the development of a European load model for road bridges. On
the basis of the results of experimental tests carried out on bridges loaded with
defined moving vehicles a dynamic simulation program has been developed
which allows to predict the dynamic behaviour of road bridges loaded with
moving traffic loads.

(2)

In the following first the mechanical models which have been used in the
program are described. After that the results of parameter studies to determine
the magnitudes and dependences of the dynamic increments for road bridges
under the actions of single vehicles are presented.

6.2
The mechanical models
6.2.1 General
(1)

Road bridges and their structural elements react to moving vehicles. The
dynamic behaviour of the bridges is mainly determined by firstly the vehicles,
which by themselves represent dynamic systems, secondly the unevenness of
the road surface and thirdly the dynamic characteristics of the bridge. Hence
the total mechanical model is subdivided into three submodels which
represent the bridge, the vehicles and the roadway roughness.

6.2.2 The mechanical model of the bridge


(1)

The dynamic calculation of the bridge structure aims at two goals:


1.
2.

(2)

the determination of the eigenfrequencies and eigenmodes of the free


undamped vibrations and
the determination of the deflections and action effects which result from
the excitement of the structure by traffic.

For the solution of both problems the differential equations for the deflection of
the bridge are needed which are obtained by considering the kinematic
equilibrium. For general continuous Systems these differential equations in
general cannot be defined and resolved analytically. Therefore numerical
procedures are used where the structure is subdivided in discrete elements.
This leads to the use of the Finite Element Method. To reduce the expenditure
efforts in discretization for the dynamic calculation of the bridge under moving
traffic the following simplications have been made:
-

In the serviceability range a linear behaviour of the bridge has been


assumed.
51

The formation of plastic deformations is excluded.


The behaviour of the three-dimensional bridge structure may be
modelled by a planar system with sufficient accuracy.

The damping of the bridge which in general is rather small is assumed


to be viscous.
First priority is given to the vertical deflections of the bridge structure.

(3)

From this follows that a bridge structure can be discretized by planar


continuous beams and skeleton elements with distributed masses and with
linear elastic material behaviour and viscous damping.

(4)

The translatory displacements of the beam elements are described by the


edge values u i for the nodes of the elements, see fig. 6-1.

Figure 6-1:
(5)

Components of the translatory displacements of the beamelement.

In using approximative deflection functions between the nodes of the beam


elements and by considering the elastic stiffness and the equilibrium with the
damping and inertia forces the following differential equation for a total system
can be obtained:

[M (t )] + [D]& U& + [K ] U = F (t )
(6)

Here [M] is the mass matrix, [D ] is the damping matrix and [K ] is the stiffness
matrix of the total system. , U& and U are the vectors of acceleration, speed
and translatory displacements of the discrete nodes. F (t ) represents the
vector of the time dependent external forces on the nodes which are given by
the actions of the wheels of the vehicles. Due to the moving masses of the
vehicles the mass matrix is not constant but time dependant.

(7)

The solution of the differential equations is achieved numerically for time steps
t by using a direct integration method (Newmark) [2]. From this solution the
values of the translatory displacements, speeds and accelerations of each
node for the time t are obtained by which the action effects in the bridge can
be fully described.
52

(8)

The afore described modelling of the structure allows the simulation for
different bridge structures as shown in fig.6-2. Simple and continuous beams,
frames and arch bridges as well as truss bridges may he analysed including
also secondary structural elements as e.g. cross-beams and stringers of
orthotropic decks.

Figure 6-2:

Bridge structures considered in the calculation-model

53

Figure 6-3:

Vehicle vibration model for a) a single-vehicle and b) an


articulated vehicle

6.2.3 The vehicle model


(1)

The vehicles represent the exciters of the vibrating system bridge', and have
to be described as vibrating systems for themselves. Due to the roughness of
the roadway the vehicles are excited and produce dynamic wheel loads. The
frequency spectrum of the wheel loads is controlled by:
-

the speed of the vehicles;


the spectrum of the roadway roughness;
the characteristics of the vehicles.

(2)

Therefore it is necessary that the modelling of the vehicle allows a good


approximation of the actual behaviour of the vehicle for a wide range of
varying conditions. For this modelling discrete rigid multibody systems with
nonlinear behaviour are used [3].

(3)

Two types of vehicles are discerned according to fig.6-3 : On one hand single
vehicles with a set of individual axles which are connected to a rigid body
54

mass, on the other hand articulated vehicles where the trailor is connected to
the motor tractor by an elastic coupling.
(4)

For each individual axle the idealized tire suspension system according to fig.
6-4 is assumed. The mass of the wheels and the axle structure of each axle
are represented by a concentrated mass between the tire spring and the
suspension structure. The suspension structure is idealised as a parallel
system consisting of a linear elastic spring for the suspension, a friction
element and a viscous damper. The body mass is considered as a rigid
element with translatory and rotatory mass inertia.

Figure 6-4:
(5)

Vibration model for a single-axle

The degrees of freedom of the vibrating system of the vehicle are determined
by the vertical translatory displacements of the axle and the translatory and
rotatory movements of the rigid body mass. Using the equilibrium conditions
the following coupled differential equation can be obtained

[M F ] &z& + [D F ] z& + [C F ] z = FF (t )
where

[M F ]
[DF ]
[C F ]

= mass matrix of the vehicle


= damping matrix of the vehicle
= stiffness matrix of the vehicle
z , z& , &z& = vector of the translatory displacements, speeds and accelerations
FF (t ) = vector of the exciting forces

(6)

These differential equations formally are identical with those of the bridge
system and therefore are resolved with the same numerical time step method.
55

(7)

Particular attention has to be paid to the suspension structure with inter-leaf


friction which is frequently used in commercial vehicles.

(8)

The inter-leaf friction can result in vibrations where the leaf spring between the
axles and the rigid body may be blocked and only the tire springs may act.
Due to the small damping effect of the tires then great dynamic effects may be
caused also for even road surfaces. This particular behaviour of the leaf spring
is considered by a nonlinear load deflection characteristic according to fig. 6-5,
which represents the "friction hysteresis".

(9)

In this vehicle model all significant parameters of the vehicle which influence
the dynamic behaviour are considered. These parameters are:
-

the vehicle structure (number of axles, inter-axle distance, distribution


of masses);
the suspension system of the axle (leaf spring suspension, pneumatic
suspension, hydraulic suspension);
the damping system of the suspensions (pneumatic, hydraulic or by
friction);
the properties of the tires.

Figure 6-5:

Force-Displacement diagram of a suspension with inter-leaf


friction

6.2.4 Model for the roadway roughness


(1)

Three different kinds of roadway roughness may be found on road bridges:

56

local extreme unevennesses, e.g. transition joints, road holes or


artificial obstacles (timber);
regular unevennesses, e.g. surface waves due to fabrication (e.g.
according to distance of cross-beams);
irregular unevennesses.

(2)

The two unevennesses mentioned first are deterministic and may be


described geometrically. The irregular unevenness however is stochastic and
statistic characteristics and functions have to be used to describe them.
For this the unevenness u (x ) , see fig. 6-6, is presented by its complex fourier
spectrum
u (x ) =

U ( ) e ix d

Stochastic unevenness profile u (x )

Figure 6-6:
(3)

The fourier spectrum displays the amplitudes and phase relations of the
different harmonic components of the unevenness. In assuming a Gaussian
distribution of the irregular unevenness and the applicability of a stationary and
ergodic random process a relation between the square of the mean value of
the unevenness profile and its power spectral density may be derived.
1
x 2 X

u 2 = lim

+X

u 2 ( x ) dx =

( ) d

(4)

Hence it follows that the power spectral density indicates how the squares of
the mean values of small frequency ranges are distributed in dependence of
the frequencies. It represents a scale for the intensity of the unevenness in
different ranges of wave lengths.

(5)

From measurements on roadways the power spectral densities for the


roughness of different roadway pavements are known [5]. In plotting these
values versus the cyclic frequency of path in a double logarithmic scale
according to fig. 6-7 it becomes obvious that all kinds of roadway pavements
can be characterized by similar functions. The densities of roughness
decrease by an exponential function with the frequency of path and may be
approximated sufficiently by straight lines.

57

Figure 6-7:

Power spectral density of roadway roughness of different road


pavements (from [3])

Table 6-1: Classification of roadway roughness (ISO-TC 108)

Quality of pavement ( 0 ) cm 3 for 0 = 1 m 1 , w = 2


lower
mean
limit
value
very good
0.5
1
good
2
4
average
8
16
poor
32
64
very poor
128
256
(6)

Therefore for the approximation of the power spectral density an exponential


function is used as mathematical model.

( ) = ( 0 )
0

(7)

upper
limit
<
2
<
8
< 32
< 128
< 512

The spectral roughness coefficient ( 0 ) and the spectral roughness


exponent w are used as characteristics for the description of the roadway
roughness. In using mean and extreme values for the parameter ( 0 ) a
classification of roadway roughnesses as given in table 6-1 may be used.

58

(8)

In order to use this stochastic description of the unevenness for a deterministic


calculation the unevenness has to be described deterministically as a discrete
profile. This is realized by approximating the profile by a fourier series. The
amplitudes of the harmonic elements of the fourier series can be determined
from a given power spectral density distribution by
i = 2 i ( i )

(9)

As the power density distribution does not contain any information on the
phase difference of the different components of the fourier series, the phase
angles for the different harmonic waves are determined by the Monte-CarloMethod on the base of a Gaussian normal distribution. Roughness profiles
which have been generated that way showed a good approximation to the
input power spectral density distributions and allow to consider the stochastic
roadway roughness in a realistic way.

6.3

Calibration of the models with test results

(1)

The mechanical models of the bridge, the vehicles and the roadway surface as
described above and their interaction were used for the development of a
computer program. This computer program can simulate the crossing of one
or more vehicles on a bridge. The results of the simulation are time histories of
any displacements, forces or action effects of the vehicle or the bridge. The
main interest is in the bridge response, in particular in the ratio of the extreme
dynamic response in the bridge to the static response of the bridge which by
definition is the dynamic increment or the impact factor.

(2)

The program was used to simulate the behaviour of bridges where


measurements for crossing of test vehicles were carried out in order to prove
that the calculations yielded to reliable results.

(3)

In fig.6-8 a summary of the input parameters for dynamic loading tests on a


road bridge in Switzerland is given [5]. The bridge system is a continuous
three span prestressed concrete sway frame bridge with a hollow section. This
bridge was crossed in several tests series by a 160 kN-vehicle with two axles
and leaf suspension with different crossing speeds in the range of 10 to 70
km/h. The test series were carried out with two different pavements with
different roughnesses. Both roughness profiles were measured to be used
directly and the power spectral density distribution was determined to generate
roughness profiles indirectly. To get the bridge responses a set of time
histories for deflections and for strains which were measured by strain gauges
were recorded. In the following only the results from the measurement of
deflections in the third bridge span are considered.

59

Figure 6-8:

Input parameters of the calibration experiment

(4)

For the numerical simulations of the tests the bridge was modelled as a frame
[7]. The discretisation of the system allowed for 32 beam-elements with 72
degrees of freedom. The vehicle was modelled as a two-axle single vehicle
with nonlinear behaviour the dynamic characteristics of which were
determined in vibration tests. The roughness profiles used in the simulations
were those directly recorded.

(5)

Fig. 6-9 shows a plot of the dynamic increments of the mid-span deflection in
the third span versus the vehicle speed. The figure presents the measured
and the calculated values. The strong variations of the dynamic increments
reveal that there is no significant functional dependence between the dynamic
increment of a bridge and the crossing speed of the vehicles. There are great
values of dynamic increments for small speeds between 10 and 20 km/h. The
reason for this effect is the inter-leaf friction of the leaf suspension which
causes frequent blockings of the leaves for small speeds by which the total
vehicle vibrates on the tires which exhibit only a small damping. The
60

eigenfrequency of the vehicle with blocked suspension, which is in the range


of 2.8 Hz, coincides with the fundamental frequency of the bridge which is in
the range of 3.0 Hz. This results in resonance effects. The comparison of
measured and calculated values proves that these dynamic particularities are
very well reflected by the calculation model.

Figure 6-9:
(6)

Comparison of measured and calculated time-histories of


deflection for a vehicle speed of 3rd span

Fig. 6-10 and fig. 6-11 allow to compare the measured and calculated time
histories of the deflections for two different crossing speeds. The time histories
demonstrate similar functions from measurements and calculations in
particular in view of the change of phases with predominant excitement and
phases with reduced dynamic responses. These changes of phase are mainly
caused by the different local distributions of exciting roughnesses and the
position of the vehicle. The frequency analysis of the time histories from
measurements and calculations demonstrates that the vibrations of the bridge
are composed of the natural vibrations of the bridge and the excited vibrations
due to the vehicle excitement.

61

Figure 6-10: Comparison of measured and calculated time-histories of


deflection for a vehicle speed of = 15 km/h

Figure 6-11: Comparison of measured and calculated time-histories of


deflection for a vehicle speed of = 65 km/h

6.4
Parameter studies to determine physical impact factors
6.4.1 Impact factors for the ultimate limit state verification
(1)

In order to clarify the mechanical causes for the observed dynamic responses
of road bridges and the main controlling parameters, parameter studies were
carried out.

(2)

The dynamic effects of crossing vehicles may be best differentiated according


to the influence length of the structural elements in question. The following
distinctions can be made:
-

(3)

In each of these cases different mechanical processes, influencing parameters


and dynamic magnification factors are relevant.

6.4.1.1
(1)

local dynamic effects caused by the load of a wheel or of an axle;


dynamic effects caused by the loading of a complete vehicle;
dynamic effects caused by the simultaneous action of several vehicles.

Local dynamic effects

Local dynamic effects are investigated for the Stresses of stringers in


orthotropic decks. For this investigation the stringer which is directly situated
under the wheel track of a vehicle is modelled as a continuous beam on elastic
supports with masses according to fig. 6-12. The results of the study are the
following:

62

Figure 6-12: Calculation model for the stringers of an orthotropic deck


(2)

In general the stringers have short influence lengths which result in very high
eigenfrequencies. Therefore these structural parts are not excited to significant
eigenvibrations by the dynamic wheel loads of the crossing vehicles. Hence
there is no dynamic interaction between the vibrations of the stringers and the
vibration of the vehicles. The dynamic stresses in the stringers are mainly
determined by the dynamic wheel loads and therefore controlled by the
dynamic properties of the vehicle, the speed and the surface roughness. Fig.
6-13 demonstrates that the dynamic wheel loads increase with the square of
the speed. The same functional dependence can be observed between the
speed and the dynamic increment of the load effects in the stringers.

Figure 6-13: Relation between dynamic wheel-loads and the dynamic


increment of direct loaded components

63

(3)

As the dynamic properties of the vehicles vary in a large range, the maximum
values of the dynamic increments have a large scatter. Fig. 6-14 gives the
scatter and the extreme values of the calculated dynamic increments versus
the type of vehicle and the roughness of the surface for a maximum speed of
100 km/h. It seems not to be justifiable to differentiate according to structural
parameters, as e.g. cross beam distance or stringer stiffness.

Figure 6-14: Dynamic increments of the stringers depending on the vehicle


type and quality of pavement
(4)

A particular attention has to be paid to structural elements adjacent to the


transition joints, because the dynamic wheel loads in this region are
significantly greater than in other areas of the bridge due to the local
unevenness of the joints. From these local unevenness the dynamic
increments may attain values which may be up to 50 % greater than those
given in fig. 6-14. In concluding a constant dynamic magnification factor for
local wheel effects depending only on the roadway roughness seems to be
appropriate.

6.4.1.2

Dynamic effects of individual vehicles

(1)

For the dynamic action effects of the main girders of a bridge with mean or
long spans the frequency contents of the wheel load variation which influences
the resonance behaviour has greater influence than the maximum values of
the wheel loads.

(2)

Therefore the dynamic behaviour of the total vehicle plays an important role
and the dynamic increment of the bridge is mainly controlled by the interaction
of vehicle and bridge.

(3)

These mechanical processes shall be first considered for the case of a single
vehicle crossing the bridge. The calculations have been carried out for three
span bridges with equal span lengths where the geometry and stiffness were
chosen such that a wide spectrum of usual span lengths and eigenfrequencies

64

were represented. Each dot in fig. 6-15 is referred to one of 35 bridges that
were investigated in the study.

Figure 6-15: Relation between the fundamental frequency and span length of
investigate bridges
(4)

As the regression curve, which from empirical studies gives a relation between
the fundamental frequency of a bridge and its span length [6], is independent
on the type of material and way of construction, the following results are
applicable to steel bridges, reinforced concrete bridges and composite
bridges.
-

The dynamic response of a bridge is mainly determined by the


fundamental frequency of the bridge and the spectrum of the dynamic
wheel loads. The bridges exhibit increased dynamic increments when
their fundamental frequency is in the range of the frequencies of the
vehicles. It is assumed however that the roughness of the surface is
such that the dynamic wheel loads in this frequency range are excited.

The spectrum of the wheel loads is determined by the dynamic


properties of the vehicle and the roadway roughness. Hence in addition
to the fundamental frequency of the bridge these two parameters are
the significant parameters for the dynamic reactions of a bridge under
traffic loads.

Bridges with short spans have an increased sensitivity to resonance


vibrations due to critical frequencies caused by the sequence of axles.

The feedback effects of the vibrations of the bridge to the vibrations of


the vehicles are rather small, i.e. the amplitudes and frequencies of the
wheel loads are almost not influenced by the vibrations of the bridge.

There is a great scatter of the maximum dynamic increments for a given


bridge even in case of constant surface quality and constant vehicles.
The reason for this scatter is that a given power spectral density
function for the roadway roughness can be modelled by an infinite
number of deterministic roughness profiles which yield to different
results.
65

(5)

The dynamic increments for the deflections are greater than those for
the bending moments. The dynamic increments for the hogging
moments are greater than those for the mid-span moments.

The damping of the bridge has less importance than the other
influence-parameters.

The significance of the fundamental frequency of the bridge and of the


characteristics of the vehicles can be taken from the fig. 6-16 6-18 which are
given as examples.

Figure 6-16: Relation between the fundamental frequency and span length of
investigate bridges
(6)

Fig. 6-16 demonstrates the dynamic increments of the bending moments


versus the fundamental frequency of the bridge for a two axle vehicle with
pneumatic suspension and a good quality roadway pavement. The envelope
demonstrates resonance effects in the ranges of 1.5 Hz to 2.5 Hz and beyond
7.0 Hz. In the ranges between these resonance areas the dynamic increments
are significantly reduced.

(7)

The resonance effects are caused by the fundamental vibrations of the bridge
and the eigenfrequencies of the rigid body masses and the axle masses of the
vehicles. The frequency ranges with low dynamic increments are outside the
excitement ranges of the vehicles. The results of the calculations are in good
agreement with the results of experimental tests [6].

(8)

The influence of the variation of the dynamic characteristic of the vehicle on


the dynamic increments of the bridge can be taken from fig. 6-17. The vehicle
considered had a conventional leaf suspension. The difference in the axle load
spectrum yields to a widening of the frequency ranges where the bridges get
higher dynamic increments and to greater values of the dynamic increments.

66

Figure 6-17: Dynamic increments of the bending moment as a function of the


fundamental frequency (vehicle B)
(9)

Fig. 6-18 shows a reduction of the resonance effects in the range of lower
frequencies due to articulated vehicles. Only for bridges with short Spans the
dynamic increments are increased due to critical frequencies caused by the
sequence of double and triple axles.

Figure 6-18: Dynamic increments of the bending moment as a function of the


fundamental frequency (vehicle )
(10)

The influence of the stochastically distributed roadway roughness on the


dynamic increments can only be determined by statistics, because a given
power spectral density distribution gives a set of equivalent deterministic
roughness profiles which cause maximum dynamic increments that scatter
very much. In ordering the impact factors of different bridges versus mean
values it can be shown, that the scatter increases with increasing mean values
and a correlation between the means values and the standard deviation is
possible. The mean values have been obtained on the basis of ten different
roughness profiles generated for a given roughness quality. From this any
extreme values of the dynamic increment can be determined when the
probability of exceedance is specified. The study has yielded to the result that
the maximum values of the impact factors of a bridge may be correlated to the
roughness quality "very good", "good" and "average" by the factors 1 : 2 : 4.

67

6.4.1.3

Dynamic effects with several vehicles

(1)

A third item of the parameter study was devoted to determine the dynamic
effects caused by the simultaneous action of several vehicles of equal
dynamic properties. The study was carried out for a convoy of 32t-vehicles
with four axles and an inter-vehicle distance of 5.0 with by a maximum speed
of 40 km/h.

(2)

The results of the study demonstrate that the magnitudes of the impact factors
due to resonance with the rigid body eigenfrequency of the vehicles are
significantly reduced and the dependences on the fundamental frequency of
the bridge is less significant than for single vehicles, see fig. 6-19. The reason
for this result is the wide band spectrum of excitation where the dynamic
effects of different vehicles are partly compensated. The influence of the span
length of the bridge becomes more important. In fig. 6-20 the results are
therefore plotted versus the span length. The tendency of the dynamic
increments decreasing with increasing span length can be clearly identified.
This is caused by the decreasing influence of the vehicle dynamics in relation
to the static vehicle effects with increasing span length.

Figure 6-19: Dynamic increments of the bending moment as a function of the


fundamental frequency for a convoy of heavy vehicles

68

Figure 6-20: Dynamic increments of the bending moment as a function of the


span for a convoy of heavy vehicles
6.4.2 Impact factors for fatigue assessments
(1)

Road bridges are loaded by varying traffic loads and have to be verified in
view of sufficient fatigue safety. In the fatigue assessment the fatigue loading
effects during the expected service life have to be accumulated and introduced
into a damage calculation. As the fatigue strength of steel parts is mainly
defined in terms of stress ranges, the traffic influences on fatigue have to be
described by loading ranges which include dynamic increments. The dynamic
increments to be used for the fatigue assessment do not represent extreme
values but have to be derived from damage calculations with actual time
histories where the dynamic response of the structure is considered. Due to
lack of knowledge in this field the fatigue impact factors so far had been
estimated only. In the following a damage equivalent impact factor for the
stringers of an orthotropic deck is determined.

(2)

When a vehicle crosses a certain section of the stringer a set of stress ranges
i and cyclic numbers ni is effected. The damage caused by this crossing
can be determined as
di =

(
i

ni

/ i ) N c
m

(3)

Here m is the slope of the S-N-curve and c is the reference fatigue strength
for a certain detail for N c = 2 106 cycles. The consideration of the dynamic
influence means that in relation to the spectrum of stress ranges i ; which is
obtained for static loading only another spectrum of stress ranges i
including dynamic effects is necessary. The fatigue impact factor can be
derived from a comparison of the different damages caused by these spectra.

(4)

This comparison of damage is carried out for single vehicles. The definition of
the damage equivalent impact factor of a single vehicle is such, that the
damage yielding from a static calculation with a vehicle load multiplied with
69

E , is the same as the damage calculated for dynamic stress ranges. In


assuming a linear relation between the stresses i of the stress range
spectrum, the action effects and the actions, the damage equivalent impact
factor may be determined on the level of action effects:

E =

n (M )
n (M )

i ,dyn

i ,dyn

i ,stat

i ,stat

(5)

The moment ranges which are used for the determination of E , have been
derived from the time histories for the bending moment by using the rainflow
counting method.

(6)

The magnitude of the damage equivalent impact factor for local effects is
controlled by the same parameters as the impact factor for ultimate limit state
verifications. These parameters are the type of vehicle, the vehicle speed and
the roadway roughness. The fatigue impact factor however represents
approximately an average dynamic increment and not an extreme value, due
to the effect of the damage calculation. This also applies to the scatter of
dynamic influences arising from vehicle characteristics. Fig. 6-21 shows the
maximum values of the damage equivalent impact factor as well as the density
distribution of the values depending on the type of vehicle. The distribution
assumed is a Gaussian distribution.

Figure 6-21: Damage equivalent impact factor for the bending moment of the
stringers of an orthotropic deck as a function of vehicle type and
quality of pavement
(7)

For a mean pavement quality and a speed of 100 km/h the mean value for the
damage equivalent impact factor for the fatigue assessment of a stringer of an
orthotropic deck is in the range E = 1.28.

6.4.3 Conclusions for loading codes


(1)

The study on dynamic influences explains the physical phenomenon of


interaction between moving vehicles, surface roughness of the road and
vibration behaviour of bridges all focussed on the action effects needed.
70

(2)

The basis of the study are deterministic traffic scenarios that give an insight
into this interaction.

(3)

Realistic traffic scenarios however are random in view of the composition of


the traffic, the vibration properties of vehicles depending on the loading rate,
the intervehicle distances etc.

(4)

Therefore the results of dynamic analysis need statistical evaluations that may
lead to
-

(5)

characteristic values and fatigue values of action effects from a static


analysis,
characteristic values and fatigue values of action effects from a dynamic
analysis.

If dynamic impact factors are then defined as the ratios between e.g. the
characteristic values of action effects from the dynamic analysis on one hand
and from the static analysis on the other hand they are in general no more
related to the behaviour of the same vehicle but to the behaviour of different
vehicles and therefore may substantially differ from the dynamic impact factor
as defined for deterministic loading situations as assumed in this section, see
fig. 6-22.

Figure 6-22: Definitions of impact factors


(6)

Statistically oriented dynamic analysis using the simulation method explained


in this section are given in section 7 of this report.

71

7.
7.1

Dynamic simulations for justifying the load models in EN 1991-2


Procedure and assumptions

(1)

The simulation program explained in section 6 of this report has been used to
determine dynamic response histories of structures under moving traffic
-

to check the results obtained with evaluations of response histories


based on static influence lines
to determine damage equivalent factors applicable together with the
fatigue load models FLM3 and FLM4.

(2)

The basis of such dynamic simulations were the statistical data of traffic as
indicated in fig. 3-15 that were filtered from dynamic amplification factors by a
numerical simulation of the measured data at the measurement point at
Auxerre, see fig. 3-8.

(3)

This filtering was achieved by calculating the dynamic wheel loads and vehicle
loads from the axle weights with dynamic effects as recorded with assuming a
good surface roughness of the road at the point of measurement, so that
dynamic increments could be obtained, that then could be subtracted from the
dynamic wheel loads and vehicle loads in order to get the static wheel loads
and vehicle loads.

(4)

In this simulation the dynamic properties of the vehicles were described by


statistical parameters as given in table 7-1.

Table 7-1:
(5)

Statistical parameters of dynamic properties of vehicles

In order to obtain a large number of data necessary because of the definition


of characteristic values by
-

a mean return period of 1000 years or equivalently


a return period of 50 years for 5% of the bridges or
a 10% fractile for a period of 100 years equal to the nominal service life
of a bridge

a sufficiently large number of simulations had to be performed.

72

(6)

Table 7-2 gives a survey on the bridges that fulfil the frequency conditions as
given in fig. 6-15, that were used for the simulations for
-

single span bridges with span L,


continuous bridges with 3 equal spans L.

Damping was assumed to be of Rayleigh-type with a critical value 1 = 0.6%


for the first mode and 2 = 0.9% for the second mode.

Table 7-2:
(7)

Characteristics of the bridges simulated

Fig. 7-1 gives a survey on the action effects as bending-moments and shear
forces that were determined by the simulation.

73

Figure 7-1:

Survey on the influence lines considered in the simulations

(8)

In the transverse direction a constant value of the influence line was assumed
which corresponds to box girder type bridges.

(9)

The assumptions made for the surface quality ( = 2.0) were


-

for characteristic values (related to ULS-verifications):


-

good
average
poor

( o ) = 4.0 cm
( o ) = 16.0 cm
( o ) = 64,0 cm

In addition a local irregularity as specified in fig. 3-9 was


checked.
-

for representative values (related to SLS-verifications):


-

for fatigue equivalent values:


-

(10)

only a good and average surface quality was assumed

only a good surface quality was assumed,


the effects of an irregularity was checked to model the effects of
transition joints.

The assumption for the type of traffic were:


Traffic 1 (artificial):

slow lane (lane 1) as in Auxerre with lorries only (by


eliminating all cars)

Traffic 2 (measured):

Auxerre traffic as measured in the slow lane


(mixture of lorries and cars)

Traffic 3 (measured)

Auxerre traffic as measured in lane 2


74

Traffic 4 (artificial):
(11)

traffic with cars only.

Depending on the deck width the allocations of the various types of traffic to
the lanes are given in fig. 7-2.
These assumptions correspond to the assumptions made for the evaluation of
static influence lines, see 3.2.4(14).

Figure 7-2:

Deck widths and assumption for composition of traffic

(12)

Inter-vehicle distances have been fixed as specified in 3.2.5(7) with a


minimum value a = 5m.

(13)

In total for each type of bridge, each width of bridge deck, each surface quality
and speed of vehicle 100 simulations were performed with about 25 vehicles
selected according to the Monte-Carlo method.
For each simulation the static and dynamic maximum values were determined
and plotted in a diagramme with an accumulated normal frequency
distribution.
Using the data from Auxerre the characteristic value was determined by
extrapolating the distributions to the 1 1.5 10 8 fractiles which in the normal
distribution corresponds to the mean value plus 5 x standard deviations. For
the extrapolation only the half-normal distribution fitted to the upper part of the
real distribution was used.

(14)

Fig. 7-3 shows the example of a bridge response (static and dynamic) for 10
vehicles with a speed v = 80 km / h and the selection of the maximum static and
dynamic values.

75

Figure 7-3:
(15)

From 100 extreme values determined for the situation in fig. 7-3 the
accumulated frequency distributions as given in fig. 7-4 were determined that
could be further evaluated to get characteristic values, see fig. 7-4.

Figure 7-4:
(16)

History of static and dynamic bridge responses and choice of


extreme values

Example for the accumulated frequency distributions for static


and dynamic bridge responses

Various simulations have shown that flowing traffic (with a speed of vehicle
v = 60 km / h 80 km / h ) is relevant for span lengths L 20 m , whereas
congested traffic (10 km / h 20 km / h ) or jam situations (0 km / h ) are relevant
for span lengths L 30 m , see fig. 7-5.

76

Figure 7-5:
(17)

Characteristic values of bridge responses for flowing and


congested traffic

Fig. 7-6 shows the influence of different surface qualities and flowing and
congested traffic on the bridge responses, and fig. 7-7 gives the associated
dynamic factors.

Figure 7-6:

Characteristic values of bridge responses for different surface


qualities and flowing and congested traffic

77

Figure 7-7:
(18)

Dynamic factors resulting from the characteristic values in fig. 7-6

In order to mirror traffic situations more realistically also the assumption of


different speeds of vehicles in various lanes was made (mixed traffic):
-

for 2 lane traffic:

lane 1 v = 80 km / h
lane 2 v = 10 km / h

for 4 lane traffic:

lane 1
lane 2
lane 3
lane 4

v=
v=
v=
v=

10 km / h
80 km / h
80 km / h
10 km / h

however comparative studies showed that either flowing or congested traffic


with equal speeds are relevant depending on span lengths.
7.2

Results of the simulations for LM1 and LM2

(1)

Characteristic values of uniformly distributed loads determined from dynamic


simulations and from evaluations for static influence lines are compared for
flowing traffic in fig. 7-8 and for congested traffic in fig. 7-9.

78

Figure 7-8:

Comparison of characteristic values of equivalent uniformly


distributed loads from dynamic simulations and static simulations
using influence lines for flowing traffic

Figure 7-9:

Comparison of characteristic values of equivalent uniformly


distributed loads from dynamic simulations and static
simulations using influence lines for congested traffic

(2)

The comparison demonstrates that the results obtained in different ways by


different authors are sufficiently accurate, differences result mainly from
dynamic effects.

(3)

Dynamic factors for 1 lane flowing traffic, with average surface quality are
given in fig. 7-10.
79

Figure 7-10: Dynamic factors for single lane bridges for average surface
quality and flowing traffic
(4)

In using the maximum value = 1.70 applied to the characteristic axle load 270
kN according to fig. 3-6 the extreme characteristic value for a single axle is for
LM2.
Q K = 2.70

(5)

279
= 402 kN
1.14

Other dynamic factors that have been used in the evaluation of static influence
lines are given in fig. 7-11 and fig. 7-12.

80

Figure 7-11: Dynamic factor for 4 lane bridges for average surface quality and
flowing traffic

Figure 7-12: Dynamic factor for 4 lane bridges for average surface quality and
congested traffic
(6)

The additional impact factor from a local irregularity is given in fig. 7-13.

81

Figure 7-13: Impact factor for a local irregularity (see fig. 3-9) in addition to
the dynamic effect from average surface quality
(7)

Comparisons of the characteristic action effects from the dynamic simulation


and the Eurocode Load-Model LM1 for the influence line DLT3 and DLT3
are given in fig. 7-14 and fig. 7-15.

Figure 7-14: Comparison of characteristic action effects from influence line


DLT 2 (fig. 7-1) from dynamic simulations and Load Model LM1
82

Figure 7-15: Comparison of characteristic action effects for influence line DLT
3 (fig. 7-1) from dynamic simulations and Load Model LM1
7.3

Determination of representative values

(1)

The combination factors to obtain representative values are determined from

= 1 2 3 4 5
where the factors i are related to

(2)

1
2
3
4

=
=
=
=

TR return period
surface quality of road
frequency of jam per year
p composition of traffic

v type of traffic (congested or flowing).

The factor 1 related to the return period

1 =

E Re turn period
E1000 years

may be taken from evaluations as given in fig. 7-16, that result in a range

1 = 0 .7 0 .8 .

83

Figure 7-16: Reduction factor 1 for a return period of 1 day


(3)

The full evaluation results in 1 -values as given fig. 7-17.

Figure 7-17: Results 1 for different return periods


(4)

The factor 2 for the surface quality of the pavement is defined by

2 =

E dyn ,good
E dyn ,average

as a good surface quality is presumed for serviceability verifications.


A mean value from the results of simulations for good and for average quality
is
2 = 0.89 .

84

(5)

The factor 3 accounts for the frequency of jams per year. In applying an
estimation of the frequency of jam per year (related to the highway BAB 9
between Nuremberg and Munich)
= 3 10 3 jams / km day

for traffic in one direction a factor of

3 = 0 ,95
may be taken.
(6)

The factor 4 for the composition of traffic is defined by


4 =

( p)

(100% )

where p is the percentage of lorries in the traffic and 100 % means traffic with
lorries only as presumed for the slow lane 1.
(7)

Fig. 7-18 gives the 4 -values depending on the percentage of lorries and the
return period, where p is 4 only and TR is the combined effect of 1 and
p.

Figure 7-18: Reduction factors 4 = p and TR = 1 p for the percentage


of lorries in the traffic
(8)

The factor 5 for the type of traffic is defined by


5 =

TR ( p , flowing )
TR ( p , congested )

Results are given in fig. 7-19 using the values TR in fig. 7-18 as reference
values showing the dependence on the number of lanes.
85

Figure 7-19: Reduction factor 5 = v


(9)

Traffic measurements in Europe demonstrate, that the percentage of lorries is


between 9 % and 59 % (Fiano Romana, Piacenza, Sesso Marconi). The traffic
in Auxerre (33%) represents about the mean value and therefore is taken as
the reference model.

(10)

Conclusions are made separately for

short spans L < 40 m, where flowing traffic is relevant and = TR


longer spans L < 40 m, where for

jam situations = TR p

(11)

flowing traffic = TR , and v .

The results for short spans are given in fig. 7-20.

Figure 7-20: Reduction factors for short spans L < 40 m


(12)

The results for longer spans are given in fig. 7-21.

86

Figure 7-21: Reduction factors for long spans L > 40 m


(13)

The factors in EN 1991-2 for the uniformly distributed load of Load-Model


LM1 have been determined with the assumption that flowing traffic represents
the frequent load

1 0.45
whereas infrequent loads are related to jam situations

1 0.80
7.4
Dynamic simulations for fatigue assessments
7.4.1 General procedure
(1)

The fatigue assessment in general is two-dimensional as it uses both stressranges i and cycle numbers ni . Hence the verification format reads

D Ed =

i > D

(2)

ni i3
n D D3

i > L

ni i5
n D D5

D Rd

(7-1)

In several Eurocodes this format has been transferred into a one-dimensional


equation
Ff LM = Ff f

c
Mf

(7-2)

where
Ff

is the partial factor for the fatigue action side

Mf

is the partial factor for the fatigue resistance side


87

is the fatigue damage equivalent factor related to 2 10 6 cycles

LM is the stress range due to the fatigue Load-Model

(3)

is the fatigue damage equivalent stress range related to 2 10 6 cycles

is the fatigue resistance of detail related to 2 10 6 cycles (detail class)

The fatigue damage equivalent factor is defined by


= 1 2 3 4 max

(7-3)

where
1
2
3
4
max

is the span-factor taking into account the shape of influence line and the
span length and the type of traffic on which the damage calculation is
based
is the factor for the volume of traffic
is the factor for the service life required
is the factor taking account of traffic on more than 1 lane
is a limit resulting from the constant amplitude endurance limit D of
the fatigue resistance curve, see fig. 7-22.

Figure 7-22: Determination of max


(4)

For determining max the maximum value max of the stress range
spectrum, see fig. 7-22, which normally is expressed by frequent fatigue
loads, is used.
max D or equivalently

max LM C

(7-4)

that gives because of C = 1.357 D


88

max = 1.357
(5)

max

(7-5)

LM

Normally max obtained from the Auxerre-traffic data is determined for


flowing traffic, though congested traffic or jam situations may give higher
max values. Such higher values are however ignored due to the rare
occurrence of jams and their small contribution to the overall damage.

7.4.2 Span factor 1


(1)

For determining 1 the following assumptions are made:


-

reference traffic is the Auxerre traffic; other types of traffic would lead to
other 1 -values
the factors 2 , 3 , 4 are taken equal 1 for the following conditions

2 = 1.0

for N obs = 0.5 10 6 / year and direction

3 = 1.0

for Tservice = 100 years

4 = 1.0

for the traffic on the slow lane (lane 1) only

1 includes dynamic effects, so that no additional value fat needs to be

used
1 -values is composed of two components
1 = '1 '1'

where

(2)

'1

results from equality of damage of the Auxerre traffic and the Eurocode
fatigue load model D FLM = DiA

'1'

results from refering the damage equivalent value FLM to the


reference number of cycles 2 10 6 , to which also the fatigue resistance
is related.

The factor '1 results from


D FLM =

N
DiA = sim
ND

'1 FLM

(7-6)

and reads

1 = m

N D DiA D

N sim
FLM

(7-7)

89

where

D , N D
N sim
FLM

(3)

represent the constant amplitude endurance limit


is the number of lorries used in the simulation of DiA
is the stress range due to the fatigue load model

The factor 1 results from

c = FLM 11

(7-8)

with 1 =

(7-9)

where results from


3
5
N (T ) (1 FLM ) = N D ( 1 FLM )

(7-10)

in the range of m = 5 and reads

=5

N (T )
ND

(7-11)

and results from


N D ( 1 FLM ) = N c ( 1 FLM
3

)3

(7-12)

in the range of m = 3 and reads

=3

ND
NC

(7-13)

The factor 1 therefore reads

1 = 5
(4)

N (T ) N D
3
ND
NC

(7-14)

In conclusion 1 is
1 = m

N D DiA D
N
N (T )

5
3 D
N sim
FLM
ND
Nc

(7-15)

7.4.3 Factors 2 ,3 , 4
(1)

The factor 2 for traffic volume is defined by


90

2 = 5

(2)

N obs

The factor 3 for service life TS reads

3 = 5
(3)

(7-16)

0.5 10 6

TS
100

(7-17)

The factor 4 fort he contribution of more than 1 lane reads

Ni
N1

LM ,i

LM ,1

Ni
N1

i

1

= 5 1+

N2
N1

i
N
+ .. n n
N 1 1
1

i =1

i =1

(7-18)

(7-19)

where

LM ,i

is the number of lanes with lorries


is the number of lorries per year on lane i
is the stress range form fatigue load model FLM on lane

is the value of the transverse influence line

n
Ni
i ( = f ( i ))

7.4.4 Results of simulations


(1)

In the following the results of simulations are given for the single fatigue
loading model FLM 3.

(2)

The definition of FLM is


FLM = max FLM min FLM

from extreme positions on the full influence line neglecting any secondary
stress range. (e.g. from waterpockets of the reservoir method)
(3)

Fig. 7-23 gives the equivalence factor 1 for sagging moments and hogging
moments.

91

Figure 7-23: Span factor 1 for sagging and hogging moments


(4)

The factor 2 is given in fig. 7-24

Figure 7-24: Factor 2 for traffic volume


(5)

The factor 3 may be taken from fig. 7-25.

92

Figure 7-25: Factor 3 for service life


(6)

For determining 4 it has to be considered that additional 10 % of N obs should


be applied to each lane other than the slow lane.

(7)

max values may be taken from fig. 7-25 and fig. 7-26.

Figure 7-26: max for sagging moments (comparison with EFT1)

93

Figure 7-27 max for hogging moments (comparison DLT 1)

94

8.
8.1

Braking and acceleration forces


Code provisions

(1)

A braking force, Qik , shall be taken as a longitudinal force acting at the


surfacing level for the carriage way.

(2)

The characteristic value of Qik , limited to 900 kN for the total width of the
bridge, should be calculated as a fraction of the total maximum vertical loads
corresponding to the Load Model 1 likely to be applied on Lane Number 1, as
follows:
Q1k = 0 ,6 Q1 (2Q1k ) + 0 ,10 q1 q1k w1 L

180 Q1 (kN ) Q1k 900 (kN )

where
L is the length of the deck or of the part of it under consideration.

Note 1: For example Q1k = 360 + 2 ,7 L ( 900 kN ) for a 3 m wide lane and for a
loaded length L > 1,2 m, if factors are equal to unity.
Note 2: The upper limit (900 kN) may be adjusted in the National Annex. The
value 900 kN is normally intended to cover the maximum braking force of
military vehicles according to STANAG6.
8.2.

Calculative model for braking forces on stiff bridges

(1)

Braking forces are determined mainly by the following factors:


1.
2.
3.
4.

(2)

mass G of the vehicle braking


number n of vehicles braking simultaneously
braking deceleration
correlation of braking processes in time.

To simplify the situation the following assumptions are made:


1.

2.

there is a convoy of n loaded heavy vehicles with equal weight, the


same silhouette, same speed and equal spacing l d between the
vehicles.
The smallest possible intervehicle spacing is used resulting from
l d = v0 t R

(8-1)

where
v0 =

vehicle speed [m/sec] before braking

STANAG: Military STANdardisation AGreements (STANAG 2021)

95

tR =

(3)

reaction time for the driver after observing the braking lights of
the vehicle travelling in front of him.

The static value F0 of the braking force is determined in good approximation of


the time-functions as measured, see fig. 8-1, by

F (t ) = F0 = G a = constant
(8-2)
where
G=
A=
A=

mass of vehicle
breaking deceleration assumed to be
5 m/sec2.

Figure 8.1:
(4)

Braking deceleration of a heavy vehicle dependant on time

The number of vehicles braking simultaneously is depending on the speed v0


of the convoy
For the speed assumed for the convoy

v0 = 90 km / h = 25 m / sec .
the deceleration time is
tB =

v0
a

and with the reaction time t R = 1,0 sec . the number of vehicles is

n = t b / t R = (25 / 5 ,0 )1,0 = 5 vehicles.

(8-3)

96

This means that there is an upper boundary value for the accumulation of
braking forces Fi from various vehicles braking simultaneously depending
on their number.
(5)

The time correlation of the braking processes resulting in F (t ) is given in fig.


8-2

Fig. 8-2:
(6)

Time process of a vehicle convoy braking

The length of the bridge, along which n vehicles brake simultaneously, is


L = n l F + (n 1)l d

(8-4)

where

l f = length of the vehicle [m]


l d = spacing between the vehicles [m]
8.3

Dynamic effects from braking

(1)

For determining the dynamic braking actions on a fixed bearing (F.B.) the
behaviour of the bridge in the longitudinal direction may be modelled according
to fig. 8-3.

97

Figure 8-3:

Dynamic model of a bridge for longitudinal actions F (t )

(2)

In fig. 8-3 C L is the reaction force at the stiff bearing and C u the reaction force
in the flexible bearing.

(3)

The vibration period T for the longitudinal vibration is:


T = 2

(4)

M
Cu

(8-5)

In the following the dynamic amplification factor (ratio between the dynamic
and the static bearing reaction) is determined by numerical simulations of the
dynamic horizontal movements of the bridge under the impulses F (t ) from the
various vehicles.
The parameters controlling the results are:
-

(5)

the impulse-length: t B
the time gap between individual impulses: t R = 1,0 sec.
the number n of impulses acting simultaneously
the eigen-period T of the bridge including abutments and piers
the damping coefficient: = 0 ,07 .

The figures 8-4 to 8-18 illustrate the time-histories of the dynamic and static
reaction forces. The eigen-period of the longitudinal bridge system is the
dominant parameter controlling the results.

98

For high eigen-frequencies (bridges, which are horizontally fixed) the dynamic
effect of the n-th vehicle has died down, when the (n + 1) th vehicle starts
with braking ; hence the dynamic amplification is governed by the dynamic
impact from a single vehicle only. For low eigen-frequencies (bridges with tall
piers and fixed bearings on the middle piers) the worst situation occurs when
the gap between impulses t R corresponds to the eigen-period T. Then there is
an accumulation of dynamic effects of various vehicles. The maximum is
obtained when for t R = 1 sec. the eigen-period is T = 1,0 sec .

99

T = 1,0 sec.

T = 3,0 sec.

T = 5 sec.

8-4

8-8

8-12

8-5

8-9

8-13

8-17

8-6

8-10

8-14

8-18

8-7

8-11

T = 0,1 sec.

8-16

8-15

Figure 8-4 to 8-18: Comparison of static and dynamic response


100

Table 8-1:

Dynamic amplification factors

(6)

Table 8-1 gives the dynamic amplification factors calculated for bridges which
are horizontally flexible or stiff. The maximum value for any system is = 1,80 .

(7)

A further damping can be effected by friction in movable bearings which may


reduce the peak values of the reactions in the fixed bearings.
The friction forces depend on the number of movable bearings, the vertical
and transversal bearing forces and the friction coefficient . Table 8-2 gives
the results of a parameter study: the amplification factors are given in
dependence of the friction forces expressed as a percentage of the bearing
forces:

Table 8-2:
(8)

Dynamic amplification factors taking account of the friction forces


in movable bearings

In the following the amplification factors given in table 8-3 are used

Table 8-3:

Amplification factors used for the simplified determination of


braking forces

8.4.

Determination of braking forces in dependence of the bridge length

(1)

Further input values for the determination of braking forces are the following:

=
=

tR
G

=
=

lF

(2)

12,0 m
5 m/sec.2
see table 8-3
1,0 sec.
200 400 kN per vehicle.

From the speed v0 for the vehicles the number n of vehicles and the bridge
length are determined in table 8-4.

Table 8-4: Bridge length L in relation of the number of vehicles breaking


simultaneously
(3)

Table 8-5 gives the braking forces F for the various weights of vehicle and the
number n of vehicles.

102

Table 8-5:
(4)

Braking forces F in dependence of the mass of vehicles

Fig. 8-19 illustrates the braking forces versus the bridge length and the mass
of vehicle

Figure 8-19: Braking forces in dependence of bridge length and mass


of vehicle
8.5

Specification of braking forces in EN 1991-2, 4.4.1 and conclusions

(1)

Fig. 8-20 demonstrates the results from the fig. 8-19 together with the
specification of the characteristic values of braking forces in EN 1991-2, 4.4.1.

(2)

For bridges with small spans the bearing forces are caused by single vehicles.
This is reflected by the minimum breaking force F = 180 kN = 180 kN for L <
1,20 m and F = 360 kN for L < 1,20 m resulting form a vehicle mass of 400 kN.

103

Figure 8-20: Comparison of braking forces in EN 1991-2 with braking forces


from convoy of vehicles
(3)

For larger spans the code-specification covers the simultaneous braking of


convoy of 300 kN-vehicles, a situation that may occur frequently. The rare
situation of a convoy of heavy vehicles with F > 300 kN is covered by the
partial factor Q = 1,35 . For a bridge length L > 200 m the design value is

Q F = 1,35 900 = 1,35 900 = 1215 kN , which covers a convoy of 400 kN


vehicles.
(4)

In considering the assumptions and simplifications made in the model it is


clear that the probability of occurrence of the braking forces as specified
decreases with longer bridge spans; however a model uncertainty cannot be
quantified.

(5)

Because of the conservations of the assumptions for braking of a convoy it is


not necessary to combine deceleration forces with acceleration forces from
vehicles travelling on other lanes.

(6)

The braking forces in EN 1991-2, 4.4.1 have been specified as a portion of the
traffic loads Q1k and q1k to install an automatic system of adaption where other
characteristic are values Q1k and q1 chosen than those recommended.

104

BIBLIOGRAPHY
[1]

JACOB, B., BRULS, A., SEDLACEK, G.: European Traffic Samples. Report at
Working-Group 2 of Eurocode 9, Part. 12, " Traffic Loads for Road Bridges ".

[2]

BATHE, K. J., WILSON, E. L. : Numerical Methods in Finite Element Analysis.


Prentice-Hall, Inc., 1976.

[3]

MITSCHKE, M.: Dynamik der Kraftfahrzeuge, Band B: Schwingungen. 2.


Auflage, Springer-Verlag, Berlin 1984.

[4]

BRAUN, H. : Untersuchungen von Fahrbahnunebenheiten und Anwendungen


der Ergebnisse. Dissertation, TU Braunschweig 1969.

[5]

CANTIENI, R.: Dynamische Belastungsversuche an der Bergspurbrcke


Deibel. EMPA Forschungs- und Arbeitsberichte, Bericht Nr. 116/4, 1988.

[6]

CANTIENI, R.: Dynamische Belastungsversuche an Straenbrcken in der


Schweiz - 60 Jahre Erfahrung der EMPA. EMPA Forschungs- und
Arbeitsberichte, Abt. 116, Bericht Nr. 116/1, Juli 1983.

[7]

DROSNER, St.: Beitrag zur Berechnung der dynamischen Beanspruchung


von Brcken unter Verkehrslasten. Dissertation RWTH Aachen 1989.

[8]

SEDLACEK, G., KNIG, G., BRULS, A. and others: Background Document of


Eurocode 9, Part. 12: "Traffic Loads on Road Bridges", Aachen 1988

105

También podría gustarte