1 INTRODUCTION

In the abstract’s context the present work is focused on three levels: − Level I: Global behaviour – primary load bearing

members (traffic loading observation mainly based on laser-supported global deflection measurement)

− Level II: Cross-sectional behaviour (dynamic traffic-

weight registration system, which utilises laser-calibrated accelerometers reproducing vertical canti-lever deformations based on pattern recognition.

− Level III: Local systems – e.g. the bottom and top

joints of the bridge’s torsional bracings (verification with additional strain gauge measurements)

In each of these levels of analysis the consumption of the structure’s overall-capacity per year is to be determined. A continuative methodology was prepared, which uses the isolated hot spot areas for detailed fatigue analysis to determine, how grave the present situation is with regard to the remaining service lifetime of the analysed structural members. As the present lifetime calculations are performed in terms of stresses by means of damage-accumulation, comprehensive Finite Element Analysis is necessary to quantify the real fatigue-threat. The calculation’s loading-input is going to be strictly measurement based. This methodology of course implies additional strain gauge measurements for verification purposes.

2 METHODOLOGY

An extensive explanation of the present research work as well as an insight into its methodology in detail is given in [Wenzel et. al 2008] and is visualised in Fig 1. Even if the present methodology represents a tailor-made approach, it can be modified and transferred to other bridge structures. For reasons of a limited extension of the present contribution it is mainly going to be focused on showing the development of a tailor-made loading model for performance prediction.

3 TAILOR-MADE LOADING MODEL FOR PER-FORMANCE PREDICTION

Based on the hot spot areas isolated in [Veit-Egerer & Wenzel 2007], detailed fatigue analysis is demanded to determine, how grave the present situation of the relevant structural members is with regard to the remaining ser-vice lifetime – using the monitoring based impact. In the following an insight is given into the idea, how the results for a one week lasting measurement campaign are ex-tracted and implemented into the developed methodology for performance prediction. Firstly accompanying video recordings were undertaken to eliminate possible uncertainties and to provide com-plementary information. Extensive data mining lead to frequency distributions including all different kinds of occurring loading scenarios (Fig. 2). One of these loading configurations - two trucks simultaneously stressing the bridge’s side span by driving in the downhill direction -

Monitoring based performance prediction of steel bridges against traffic loading exemplified at the Europabrücke

R. Veit-Egerer & H. Wenzel VCE – Vienna Consulting Engineers, Vienna, Austria

ABSTRACT: Bridges are ageing and traffic is growing, which creates a demand for accurate fatigue life assessment. The Europabrücke – a well known Austrian steel bridge near Innsbruck, opened in 1963 - is one of the main alpine north-south routes for urban and freight traffic. It represents a bridge generation, where bridge designers acted on a maxim of building material economisation. A long-term preoccupation of VCE with BRIMOS® (BRIdge MOnitoring System) on the Europabrücke (since 1997) with regard to fatigue problems and possible damage led to the installation of a permanent monitoring system in 2003. Since that time a lot of investigations and additional special measurements were devoted to innovative, mainly monitoring-based fatigue assessment. As life-time predictions in modern standards depend on lots of assumptions, the emphasis is to replace those premises – referring to loading - by measurements.

was chosen to show its consequence for the three levels of performance analysis (Fig. 5), which have been intro-duced.

Raw Data (Measurement)

Data treatment & pre-conditioning

Data evaluation & verification based on Finite Element Analysis

Data processing & statistical analysis

Structural analysis by various constraint load cases(FE_Modelling leads to nominal & geometric hot spot stresses)

Fatigue analysis (High-Cycle Theory)& sensitivity studies (statistical scatter)

Comparison to prevailing standards

Laser Displacement Forced Balance Accelerometers

Strain Gauge Measurements

Baseline correction(Histogram); Outlier-elimination

DYGES - algorithm to reproduce absolute displacements from accel.via p :

x smoothing (weighted averaging)x unification of algebraic signx envelope utilisation (time window - secant method)

x histogram based new offset determination

attern recognition

x resampling

x calibration - function(truck tonnage, velocity)

X rainflow-cycle counting (individualised)x statistical analysis, adaptation & extrapolation

Baseline correction(Histogram); Outlier-elimination

Level I Level II Level III

Fig. 1 Detailed flowchart of the methodology

Fig. 2 Typical, exemplary loading scenario – two trucks caus-ing unsymmetrical traffic impact - highlighted in Fig. 4

Fig. 3 Video based observation of relevant loading scenarios

Sing

le T

ruck

Nor

th

Sing

le T

ruck

Sou

th

Two

Truck

s Nor

th

Two

Truck

s So

uth

Thre

eTru

cks Nor

th

Thre

eTru

cks So

uth

FourT

ruck

s Nor

th

FourT

ruck

s So

uth

One T

ruck

Eac

h Si

de

Two

Truck

s Ea

ch S

ide

Thre

e Tr

ucks

Eac

h Si

de

2 Tr

ucks

Nor

th +

1 T

ruck

Sou

th

2 Tr

ucks

Sou

th +

1 T

ruck

Nor

th

3 Tr

ucks

Nor

th +

1 T

ruck

Sou

th

3 Tr

ucks

Sou

th +

1 T

ruck

Nor

th

3 Tr

ucks

Nor

th +

2 T

ruck

Sou

th

3 Tr

ucks

Sou

th +

2 T

ruck

Nor

thWE, 23.05.2007

TH, 24.05.2007

FR, 25.05.2007

SA, 26.05.2007

0

20

40

60

80

100

120

140

160

180

200

Frequency of occurence

Fig. 4 Video supported classification of loading scenarios and resulting frequency distribution

LEVEL I

v=48 / 45 km/h

Deformation 4,55 / 2,17 mm

LEVEL I

v=48 / 45 km/h

Deformation 4,55 / 2,17 mm

LEVEL I

v=48 / 45 km/h

Deformation 4,55 / 2,17 mm

DMS_RFBB_mo_1 DMS_RFBI_mo_2 DMS_RFBI_mu_3 DMS_RFBB_mu_4DMS_RFBI_u_5 DMS_RFBB_u_6 DMS_RFBI_o_7 DMS_RFBB_o_8DMS_RFBI_ml_9 DMS_RFBI_mB_10

09:47:00 09:47:05 09:47:10 09:47:1526.5.2007

-100

-75

-50

-25

0

25

50

75

100

Fig. 5 Randomly selected loading scenario and the corre-sponding structural response – expressed by means of the Three Level Approach

Global response in terms of vertical deflection due to the chosen scenario was recorded using a transmitter unit at the bridge abutment and a receiver unit located at the southern side span (laser supported global deflection measurement) In the course of [Wenzel & Veit-Egerer 2007] the main feature of the permanent monitoring system – the DY-GES algorithm (Dynamic Freight Traffic Classification) - was introduced. This feature explicitly provides the truck loads corresponding with the observed global bridge span deflection. Finally the effective axial forces and bending moments acting on the bracings as well as the occurring stress cy-cles can be extracted from the monitored strain data re-corded at the directly stressed bracings. In the following it is explained, how the established, tailor-made loading model (Fig. 2) is utilised for performance analysis of Level I (Global Impact):

Fig. 6 Traffic induced response in terms of vertical displace-ment (red) – isolated from the overall impact (blue) for the bridge‘s side-span V; 22.05.2007 19:00 - 27.05.2007 10:00 Monitoring based global impact data for a certain, repre-sentative time period (one day, one week – e.g. Fig. 6) are assessed using Rainflow Analysis. To express the Damage Rate for this loading function each of the ob-served, typically occurring loading scenarios (Fig. 4) is implemented one by one into a Structural Finite Element Model using certain truck loads (DYGES). This leads to a corresponding structural stress cycle Δσ. As it has to be dealt with truck traffic causing High Cycle Fatigue, linear elastic material behaviour can be assumed. Stresses and strains are assumed to remain elastic. This facilitates the calculations, as a single load case due to the chosen sce-nario is to be calculated in terms of stresses, while all other occurring truck loads are automatically included within the Rainflow Matrix. The latter, the calculated stress cycle value and the stress life curve corresponding with the analysed relevant structural detail lead to a Damage Matrix. This assumes only the chosen loading scenario to be occurring. Based on the frequency distribu-tion of the observed loading scenarios (Fig. 2) the chosen entry gets its weighting within all other relevant loading cases to quantify the consumed loading capacity related to a certain observation period. The same procedure is repeated for all other entries of the observed loading sce-nario distribution leading to an accumulated damage per determined observation period.

laser_gesamt laser_verkehr

23.5, 0h 23.5, 8h 23.5, 16h 24.5, 0h 24.5, 8h 24.5, 16h 25.5, 0h 25.5, 8h 25.5, 16h 26.5, 0h 26.5, 8h 26.5, 16h 27.5, 0h 27.5, 8h2007

-17.5

-15.0

-12.5

-10.0

-7.5

-5.0

-2.5

0.0

2.5

5.0

7.5

10.0mm

-17.5

-15.0

-12.5

-10.0

-7.5

-5.0

-2.5

0.0

2.5

5.0mm

LEVEL III 35 t

Bracing Innsbruck N = 54,3 kN; Mx = -0,26 kNm

Bracing Brenner N = -50,7 kN; Mx = -0,05 kNm

22 t Bracing Innsbruck

N = 37,7 kN; Mx = -0,18 kNm Bracing Brenner

N = -35,8 kN; Mx = -0,03 kNm

LEVEL II v= 48,16 km/h; 44,93 km/h Dyges = 5,9721 mm ; 4,0974 mm => Scaling Factor = 0,749 ; 0,758 Dygeskal = 4,47 mm ; 3,10 mm Tonnage classif. 35 t; 22 t Velocity classif. 45-50 km/h; 40-45 km/h

In the course of the performed Finite Element calcula-tions - following the consequence for the fatigue resis-tance of structural details relevant for Level I, a parallel determination for those structural details relevant for Level III can be done. Additionally there is always the possibility to make a crosscheck for those stresses from analytical calculations by comparing them to the ones given by explicit strain measurements at representative positions, which shows the capability of this integral methodology (Fig. 7).

DMS_RFBB_mo_1 DMS_RFBI_mo_2 DMS_RFBI_mu_3 DMS_RFBB_mu_4 DMS_RFBI_u_5 DMS_RFBB_u_6 DMS_RFBI_o_7 DMS_RFBB_o_8 DMS_RFBI_ml_9 DMS_RFBI_mB_10

5/25, 12h 5/25, 15h 5/25, 18h 5/25, 21h 5/26, 0h 5/26, 3h 5/26, 6h 5/26, 9h 5/26, 12h 5/26, 15h 5/26, 18h 5/26, 21h 5/27, 0h 5/27, 3h 5/27, 6h 5/27, 9h2007

-80

115

-100

0

100

-135

95

-90

105

-190

200

-1000

100

-100

0

100

-1000

100

-50

050

-85

80

Fig. 7 Successively added strain gauges - installed along an assembly of two corresponding torsional bracings; 25.05.2007 11:00 - 27.05.2007 11:00

dygeskal dygeskal_2 dygeskal_3

25.5, 12h 25.5, 15h 25.5, 18h 25.5, 21h 26.5, 0h 26.5, 3h 26.5, 6h 26.5, 9h 26.5, 12h 26.5, 15h 26.5, 18h 26.5, 21h 27.5, 0h 27.5, 3h 27.5, 6h 27.5, 9h2007

-15

-14

-13

-12

-11

-10

-9

-8

-7

-6

-5

-4

-3

-2

-1

0

1mm

Fig. 8 Corresponding pattern of the Dynamic weight registra-tion based on cantilever acceleration measurements (repro-duced cantilever deformations via pattern recognition) – a fea-ture implemented into permanent monitoring; 25.05.2007 11:00 - 27.05.2007 11:00 Performance analysis for Level II (Cantilever impact) seems to be less complex, as the consequence on fatigue relevant structural details in terms of principal structural stresses occurs perpendicular to the stress cycles - previ-ously analysed for Level I. The DYGES based truck loading impact (Fig. 8) is again implemented into struc-tural models by means of a deformation constraint load case affecting the cantilever’s outer edges.

4 REMAINING SERVICE LIFETIME BY MEANS OF EXISTING TRAFFIC DATA

The demonstrated damage calculations refer to a sufficient amount of workdays and weekend-days, that enable a representative extraction of a damage rate per week, which consequently enables its extrapolation to enlarged time periods. Thus the assessment of an analysed detail via Damage Matrix calculated for the measuring time of a whole year ( => ‘Damage-per year

effect’ ) is possible to determine the consumption of the overall loading-capacity per year for the corresponding detail. The detailed knowledge about the progression of the prevailing traffic from the very beginning up to these days and the implementation of published future trend studies with regard to the next ten years (until 2015) are used for fatigue analysis (variation of traffic volume & variation of the notional truck-weight). Fig. 9 shows the increase of the freight traffic volume at the Europabrücke. According to [Verkehrsentwicklung in Tirol 2006], traffic volume in 2006 increased to an amount of 472 % in relation to 1964 and is expected to grow 2.9% per year until 2015 [Verkehrsprognose 2015]. To derive every considered year’s Damage Matrix affected by the variation of traffic volume, fatigue analysis approximately demands a uniform adaptation of the number of occurrences for all elements of the derived Rainflow Matrix before extrapolation techniques of the measured impact for the whole lifetime - discussed in [Wenzel et. al 2008] - are applied.

Fig. 9 Progression of freight traffic volume at the Europ-abrücke from 1964 until 2015 [trucks/day] Fig. 10 shows the increase of the effective amount of transported goods compared to a calculated cargo per no-tional truck. The calculations showed that this truck weight in 2006 increased to an amount of 503 % in rela-tion to 1964 and is assumed to have already reached a maximum. This means, that a further increase of trans-ported goods is likely to be a consequence of the still growing traffic volume. An adaptation for fatigue analy-sis due to the variation of the notional truck-weight is re-alised by re-scaling the Rainflow Matrice’s subdivision. The information included in both of these functions can be broadened crucially, as the DYGES based freight traf-fic classification delivers exact truck weight data for the analysed bridge object – starting with the summer of 2004.

Fig. 10 Effective amount of transported goods on the Brenner route compared to a calculated cargo per notional truck

5 CONCLUSIONS AND FUTURE WORK

The present contribution explicitly deals with measure-ment data and the procedures, how they were provided and conditioned for continuative performance analysis. The methodology provides strictly in-situ based loading input parameters for continuative fatigue analysis leading to a “tailor-made” performance prediction. In addition strain gauge measurements in each of the considered lev-els of analysis (Level I – Level III) are done for verifica-tion purposes of the Finite Element based stress-life ap-proach. Thus the main goal, the substitution of the standard’s premises – referring to loading – has been reached in a quite innovative manner with regard to de-termine the consumption of the structure’s overall-capacity per year by means of a three level approach. One of the main innovations within the present method-ology is the introduced DYGES Algorithm, which is nec-essarily based on a pattern recognition algorithm by Wenzel & Veit-Egerer [Wenzel & Veit-Egerer 2007] in order to reproduce vertical cantilever deformations from accelerometers (major feature of the permanent monitor-ing system). The development of the DYGES Algorithm was nominated for the Austrian Award for Telematics, by the Federal Ministry of Transport, Innovation and Tech-nology. Besides the traffic induced impact, additional monitoring campaigns revealed, that there is a strong influence due to sun-radiation (Fig. 11 & Fig. 12). In addition to the de-scribed laser supported global deflection measurements the progression of the offset of the bridge’s reference ac-celerometer in the lateral direction was transformed into an angle of inclination. The resulting temperature gradi-ent function induces additional axial forces and deforma-tions explicitly into the outer bridge spans. Approximate analysis indicates that this constraint’s intensity level can occur up to the range of the traffic-impact itself. The af-fection due to temperature and radiation impact needs to be analysed seperately, before it is necessarily going to be superimposed with the strictly traffic induced fatigue impact. Contrary to the general doctrine in conjunction with structural performance analysis of welded components

the authors assume, that the influence of temperature leading to varying mean stresses will have to be consid-ered. The authors are convinced that shrinkage effects that lead to residual stresses in the welds are minimized or already vanished in this almost 45 year old and con-stantly stressed structure. In that case varying mean stresses become relevant even for welded structural members and demand the application of appropriate cor-rection rules.

Fig. 11 Integral Assessment of environmental conditions: Steel-temperature and Air-temperature along on a certain cross section (pier V): 23.05.2007 19:00 - 27.05.2007 13:00

Fig. 12 Integral Assessment of Traffic impact and comparison with environmental conditions: Global vertical deformations vs Air-temperature vs. Radiation efficiency vs. Humidity: 23.05.2007 20:00 - 26.05.2007 12:00 Starting with May 2007 the in-situ distribution of tem-perature impact as well as the sun-radiation itself is as-sessed via permanent monitoring using a profile of tem-perature sensors over a certain box-girder’s cross section (Fig. 11). An explicit temperature load case is going to be created and again implemented into the Finite Element calculations. The implementation of strictly measurement based loading impact into Finite Element calculations strongly supports quantitative estimation of the service-lifetime via separation of fatigue relevant loading cycles from the randomly occurring overall traffic. The Palgrem-Miner based accumulation of all calculated Damage Matrices

gesamt_durchbiegungen MQ2_InnenTemp_Pt100 Strahlungsleistung Luf tfeuchte

23.5, 0h 23.5, 8h 23.5, 16h 24.5, 0h 24.5, 8h 24.5, 16h 25.5, 0h 25.5, 8h 25.5, 16h 26.5, 0h 26.5, 8h2007

-10

0

10mm

15.017.5

20.0

22.5

25.0

27.5°C

0.00

0.25

0.50

0.75

1.00kW/m^2

40

60

80

100%rF

st_A_o_gesamt st_A_u_gesamt st_RFBB_a_gesamt st_RFBB_m_gesamt st_RFBB_o_gesamtst_RFBB_u_gesamt st_RFBI_a_gesamt st_RFBI_m_gesamt st_RFBI_o_gesamt st_RFBI_u_gesamt

24.5, 0h 24.5, 12h 25.5, 0h 25.5, 12h 26.5, 0h 26.5, 12h 27.5, 0h 27.5, 12h2007

12

14

16

18

20

22

24

26

28

30

32

34

36

38°C

from the very beginning of the bridge’s existence up to now leads to the remaining capacity of loading cycles for the analyzed detail. As a superior conclusion in addition to a quantitative estimation of the service-lifetime another key figure FR (Fatigue Relevance) is also derived (eq. (1)). It separates remaining, fatigue relevant loading cycles ni (registered by sensors and taken from the Damage Matrix) from the randomly occurring traffic (ADTV = average daily traffic volume). It is obvious that the investigation’s results are going to be improved by progressive stages; the longer the observation period lasts.

(1) This whole conception assures the determination and ob-servation of slowly progressing processes in the structure, which might lead to local damage or to deterioration of the structure’s operational integrity. The results of these hot spot analysis calculations with regard to the fatigue resistance itself are going to be undertaken and discussed in the course of further publications.

6 REFERENCES

Veit-Egerer, R. & Wenzel H. 2007. Monitoring based weak point determination of a steel bridges’s torsional bracings with regard to fatigue threat in “Proceedings of the 2nd In-ternational Conference - Experimental Vibration Analysis for Civil Engineering Structures (EVACES'07)”. Portugal: ISBN 978-972-752-095-4

Verkehrsprognose 2015 - vorläufige Ergebnisse hochrangiges

Straßennetz Österreich. BMVIT-Abteilung II/A/1. Wien. 2000

Verkehrsentwicklung in Tirol – Berichte 1984-2006. Amt der

Tiroler Landesregierung-Abteilung Gesamtverkehrs-planung. Innsbruck.

Wenzel H. et al. 2008. Structural Health Monitoring of

Bridges. England: J. Wiley and Sons Ltd. (under examina-tion)

Wenzel H., Veit-Egerer R.: Measurent based traffic loading

assessment of steel bridges – a basis for performance prediction. International Journal of Structure and Infrastructure Engineering. Taylor & Francis Group. New York. 2007 (submitted and already scientifically reviewed)