Engineering Structures 48 (2013) 144–154

Contents lists available at SciVerse ScienceDirect

Engineering Structures

journal homepage: www.elsevier .com/locate /engstruct

Dynamics of modular expansion joints: The Martinus Nijhoff Bridge

0141-0296/$ - see front matter � 2012 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.engstruct.2012.09.016

⇑ Corresponding author. Tel.: +31 888664269; fax: +31 888663010.E-mail addresses: bruno.zuadacoelho@tno.nl (B. Zuada Coelho), adri.vervuurt

@tno.nl (A.H.J.M. Vervuurt), willy.peelen@tno.nl (W.H.A. Peelen), han.leendertz@rws.nl (J.S. Leendertz).

B. Zuada Coelho a,⇑, A.H.J.M. Vervuurt a, W.H.A. Peelen a, J.S. Leendertz b

a TNO, Van Mourik Broekmanweg 6, PO Box 49, 2600 AA Delft, The Netherlandsb Ministry of Infrastructure and the Environment, Utrecht, The Netherlands

a r t i c l e i n f o

Article history:Received 14 February 2012Revised 28 August 2012Accepted 13 September 2012Available online 24 November 2012

Keywords:Over rolling testMonitoringNumerical analysisDynamic behaviourDamage detection

a b s t r a c t

Modular expansion joints are structures that are submitted to severe fatigue load conditions. This maylead to unexpected premature damage of the structure which, besides the economic cost of repair,may limit the regular service of the bridge. To better understand the dynamic behaviour of modularexpansion joints, different over rolling tests have been performed at the modular expansion joint ofthe Martinus Nijhoff Bridge in the Netherlands. The tests were part of a research program for developingan early warning monitoring system for expansion joints. Strain measurements were performed duringthe passage of a vehicle at different speeds, for two different scenarios regarding the fixations (slidingsprings and sliding bearings) of the cross beams. At the same time, a numerical model was developedand validated by means of the experimental data. This article presents and discusses the measurementsand the numerical analysis. The results highlight on the effect of the cross beam fixation, and the effect ofthe vehicle speed on the strain distribution along the centre beam, together with the changes in modalproperties of the structure.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Expansion joints in large span bridges are designed to accom-modate the relative displacement due to temperature effects be-tween bridge decks and abutments, ensuring the serviceability ofbridges. Expansion joints are subjected to heavy traffic often lead-ing to fatigue damage and subsequent repair and maintenance ef-forts for responsible bridge administration. In the Netherlands it isestimated that approximately €20 million are spent yearly in themaintenance of expansion joints [1], while in Europe the mainte-nance of expansion joints represent between 8% and 20% of the to-tal maintenance costs for bridges [2].

Modular expansion joints are common for long span bridges.Modular expansion joints consist of centre beams welded to crossbeams, allowing the total gap to be divided into smaller parts inbetween the centre beams, making it possible to accommodate sig-nificant bridge movements (above 100 mm [3]). Additionally, mod-ular joints are designed to prevent the penetration of water andchemicals into the underlying structures, therefore reducing thecorrosion of the structural elements [4–8]. Despite the generalagreement on the good behaviour of modular joints, some prema-ture failures due to fatigue have been reported in literature [4,5,9].

The Martinus Nijhoff Bridge, located in the Netherlands, openedto traffic in 1996 and contains three modular expansion joints thatwere designed for a 20 years life expectation. Due to more and hea-vier traffic than projected the joints were subjected to heavier traf-fic than anticipated in the design. Therefore fatigue damage wasnoticed much earlier than expected. Consequently the joints havebeen repaired several times. The experience in the Martinus NijhoffBridge has shown that the damage of the modular expansion jointwas, in most cases, initiated by the lack of support of the crossbeam, due to the movement of the sliding bearings.

Since joints directly affect the availability of the road network,possible failure of such expansion joints is of major concern forinfrastructure managers. Because failure of the joints may lead toan unsafe situation, an early warning monitoring system (EWMS)was developed and implemented on the Martinus Nijhoff Bridge[10,11]. The goal of the system was to foresee possible failure ofthe joint in an early stage and take preventative measures accord-ing to plan, thus causing less hindrance for the traffic. As part of theearly warning monitoring system, two over rolling tests were per-formed, aiming to increase the knowledge of the dynamic behav-iour of modular expansion joints. In the second over rolling testthe primary goal was to examine the possibility of damage detec-tion on the cross beam fixations (sliding spring and sliding bearing)by monitoring the changes in strain magnitude and modal proper-ties of the modular joint. This paper presents the results of thisover rolling test. The effect of sliding bearing degradation was as-sessed together with the effect of speed and vehicle position. Atthe same time, a numerical model was developed and validated

B. Zuada Coelho et al. / Engineering Structures 48 (2013) 144–154 145

by comparison with the experimental data, which allowed thestudy of gradual degradation of the sliding bearings.

2. Description of the modular joint

The modular expansion joint chosen for the implementation ofthe early warning monitoring system is part of the MartinusNijhoff Bridge. This bridge is located in the centre of the Nether-lands, on the highway A2 that makes the connection betweenthe cities of Utrecht and ’s Hertogenbosch. The bridge consists ofan approach section with ten spans and a main bridge with a spanof 256 m and two approach spans of 152 m. The main bridge is acable stay bridge with a reinforced concrete deck, supported byprestressed concrete cross beams, which are connected to twomain girders that are supported by a multiple stay system. The ap-proach bridge has a length of 492 m and consists of a concrete decksupported by longitudinal beams continuous over three spans. Thewidth of the bridge is 34 m. The bridge road has three traffic lanesin each direction. On the western side of the bridge there is a bicy-cle lane [12]. Fig. 1 shows an overview of the bridge.

The bridge comprises three expansion joints, one between thenorth abutment and the approach section, one between the ap-proach section and the main bridge and one between the mainbridge and the south abutment. All three joints are modular jointsmanufactured by Maurer Söhne in Germany. The early warningmonitoring system was implemented in the modular joint locatedon the south side.

The modular joint consists of three centre beams supported oncross beams. Each centre beam is welded to a corresponding crossbeam, and the cross beams are supported at the joist boxes by slid-ing bearings and prestressed elements (sliding springs and slidingbearings), to allow the expansion and contraction of the structure.

Fig. 1. Martinus Nijhoff Bridge.

Fig. 2. Modular expansion joint.

The gap between the different centre beams are kept constant bymeans of control springs. A rubber membrane is placed betweeneach pair of centre beams and edge beams to seal the joint system,and avoid the passage of water and other remains. Fig. 2 shows aschematic view of the modular expansion joint.

3. Experimental program – over rolling test

3.1. Overview

The over rolling test was performed as part of the early warningmonitoring system that was applied at the Martinus Nijhoff Bridgefor continuously evaluating the performance of the modular joint[11]. The goal of the early warning system is to detect damage inthe modular expansion joint by monitoring changes in the dy-namic behaviour. As a result, it is expected an extension of the ser-vice life of the joint, together with the reduction of traffichindrance, since maintenance can be planned due to the timelywarning.

The over rolling test was performed according to the ETAG rec-ommendations [13].

3.2. Instrumentation

The instrumentation consisted of strain gauges placed along themodular expansion joint. An overview of the device position is pre-sented in Fig. 3. Two distinct areas with sensors can be distin-guished, on the west and east sides. These areas are related tothe early warning monitoring system. The over rolling test wasperformed on the west side, at which a total of 27 strain gaugeswere installed, 16 on the cross beams and 11 on the centre beams(Fig. 3). Centre beams L1 and L3 have 2 strain gauges, one betweenjoist 5 and 6 and another between joist 7 and 8. Centre beam L2contains eight strain gauges: four at each midspan between joist4 and 8 and adjacent to the supports between joist 5 and 7.

The goal was to accurately characterise the strain variationalong the length of the centre beam. On five joists (joist 4–8), allthe cross beams (3 per joist) were instrumented with one straingauge. All the over rolling test was performed with an samplingfrequency of 2000 Hz.

3.3. Tests

The main objective of the research is to determine whether it ispossible to detect failure of the sliding bearings by monitoring thestrain level. This is related to the past experience of the mainte-nance manager, where it was found that the degradation of themodular joints is related to the lack of support of the cross beams,due to the movement of the sliding bearings. Therefore, two differ-ent test scenarios were performed:

� test 1, which corresponds to the state of serviceability;� test 2, where one sliding bearing on joist 6 cross beam T2 (see

Fig. 3) is manually removed, corresponding to a damagedsituation.

For each test, the vehicle crossed the joint at different speeds,and at two different positions (Table 1). Test A corresponds tothe vehicle passage with the right wheel crossing the centre beamat the midspan (crossing devices 1/3/5 – see Fig. 3), whereas test Bcorresponds to the vehicle passage with the right wheel crossingthe centre of joist 6 (crossing devices 24/25/26 – see Fig. 3). TestA focused on the behaviour of the centre beams, whereas test B fo-cussed on the cross beam behaviour.

Fig. 3. Device position for the over rolling test.

Table 1Over rolling test definition and vehicle speeds.

Test Expected speed (km/h) Measured speed (km/h)

A1 5 6.750 42.870 70.390 80.7

A2 5 4.350 51.670 69.890 80.5

B1 5 4.250 47.470 69.990 78.4

B2 5 6.350 49.470 71.090 82.1

146 B. Zuada Coelho et al. / Engineering Structures 48 (2013) 144–154

The load was applied by means of a three axle vehicle. The rel-evant vehicle geometry, as well as the axle loads, are presented inFig. 4 and Table 3. The axle load corresponds to the static weightdetermined prior to the test execution, while the wheel printwas determined by painting the contact area between the wheeland the asphalt. The variation of the wheel print with the speedwas not accounted in the analysis, which is a reasonable approxi-mation for the measured vehicle speeds [14], and is supported byprevious experimental results [15]. During the vehicle passagesthere was no traffic on the west side of the bridge.

Table 2 presents the information about the misalignment of thevehicle, regarding the two alignments (A and B – see Fig. 3). Sincethe vehicle crossed the joint at variable speeds, there were inaccu-racies relative to the defined alignments for the tests. Video re-cords of the vehicle passage were recorded, which allowed anestimation of the vehicle misalignment [16].

4. Experimental results

4.1. Static results

The static response corresponds to the vehicle passage along themodular joint at very low speed (�5 km/h). Fig. 5 shows the straintime history for test A1, on (a) the centre beam L2, strain gauge 3and (b) on the box joist 6, cross beam T2, strain gauge 25. The max-imum strain is consistently observed at the time the vehicle wheelcrosses the monitored strain gauge. The strain level at the centrebeam is higher than the strain level at the cross beam. This is re-lated to the fact that the load is directly applied to the centre beam,and redistributed towards several cross beams.

The maximum strain level at the centre beam is approximatelyconstant for the three vehicle axles, while on the cross beam therear axles exhibit higher strain magnitude, which is in agreementwith the axle weight of the vehicle (see Table 3). The constantstrain level recorded for the centre beam is related to the vehicleconfiguration. During test A, the right front wheel is aligned withthe middle of the centre beam (alignment A), but since the rearaxles are dual wheels the point of load application is different forthe front and rear axles. When the front wheel is applied at middlespan, the load application point for the rear wheels is shifted to-wards the support (joist 6), causing a reduction of the maximum

Fig. 4. Geometrical vehicle description.

Table 2Misalignment of the vehicle passage. The values correspond to an estimation of the misalignment of the middle of the front wheelin relation to alignment A or B, based on video records. A positive value corresponds to a vehicle passage on the west side of thealignment, while a negative value to a passage on the east side of the alignment, in correspondence with Fig. 3. The estimationsare presented in m.

Speed (km/h) Test

A1 B1 A2 B2

sta

+0.05 0 +0.15 0

50

�0.15 +0.20 +0.05 +0.05

70

�0.15 �0.15 +0.05 0

90

+0.05 �0.20 0 0

B. Zuada Coelho et al. / Engineering Structures 48 (2013) 144–154 147

strain in the middle of the centre beam. For the cross beam, thisvariation of the load application point is of minor importance sincethe vehicle load is kept the same. Therefore, the same ratio be-tween the strain at the rear and front axle (�1.6) and the axleweight (see Table 3) is found.

Fig. 6 shows the strain influence line along centre beam L2 (seeFig. 3), for tests A1 and B1. For test A1 the vehicle is crossing themodular joint along the middle of centre beam L2 (right wheelon strain gauge 3 and right wheel on strain gauges 8 – alignmentA), while for test B1 the vehicle is running on top of the box joist6 (right wheel between strain gauge 4 and 6 and right wheel onstrain gauge 10 – alignment B). It follows that a distinct behaviouris found. For test A1 the maximum strain is observed under thepoint of load application causing tension on the bottom flange ofthe centre beam. For strain gauges 4 and 6 an inversion of thestrain diagram occurs, due to the presence of the cross beam thatacts like a support. For test B1 the maximum strain level is noticedin between the wheel loads (� strain gauge position 8), caused bythe centre beam top flange being in tension. The strain magnitudefor devices 0, 2 and 3 is approximately zero, because the vehiclecrosses the joint along alignment B and the wheel that is closestto these points is applied directly to the cross beam, which causesno deformation of the centre beam.

The effect of damage was simulated by manually removing thesliding bearing of cross beam T2 located on the north side of joist 6(where strain gauge 25 is installed, see Figs. 2 and 3). Fig. 7 showsthe influence line for centre beam L2, for both tests A and B. It

follows that if the sliding bearing is removed (the removed slidingbearing is located in between strain gauge position 4 and 6), thedeformation pattern of the centre beam changes. For test A a straininversion occurs for strain gauges 4, and 6, together with an in-crease of strain for strain gauges 3, 7 and 8. This shows that thestructural scheme of centre beam L2 has changed. Due to the lackof support, the strain that was negative on the support region(strain gauges 4 and 6) is now positive. This means that if the fix-ation is removed the centre beam has to resist to more load sinceits span is increased. The variation between 3 and 7 is not linear,which shows that the cross beam still imposes some restrictionto the system, acting like a spring.

For test B the strain increases at the gauges placed in the regionof load application (strain gauges 4 and 6) which coincides withthe place where the sliding bearing has been removed. These straingauges, for test B1, had no significant strain since the load is di-rectly transmitted to the cross beam. In test B2, the cross beamis no longer effective in transferring the load, therefore the strainincreases by factors of 6 and 3 for, respectively, strain gauges 4and 6. Since the sliding bearing is no longer present, a stress redis-tribution occurs towards the available supports (strain gauges 2and 8), where an increase of the strain level is registered.

Fig. 8 shows the maximum strain magnitude in the differentcross beams. During test B2 the sliding bearing of cross beam T2(strain gauge 25) is removed, which causes a significant reductionin the strain level of this gauge. Additionally, an increase of thestrain occurs for strain gauges 22 and 28. As the sliding bearing

0 1 2 3 4 5−20

0

20

40

60

80

100

120

140

160

Time [s]

Mic

roSt

rain

[−]

0 1 2 3 4 5−20

0

20

40

60

80

100

120

140

160

Time [s]

Mic

roSt

rain

[−]

(a)

(b)

Fig. 5. Strain during the passage of the vehicle at �5 km/h for test A1: (a) on centrebeam L2, strain gauge 3 and (b) on box joist 6, cross beam T2, strain gauge 25.

Table 3Vehicle characteristics.

Axle Axle load (kN) Wheel print (m)

Width (m) Length (m)

Front axle 73 0.335 0.275Rear axle 1 115 0.231 0.242Rear axle 2 109 0.223 0.246

0 2 3 4 6 7 8 10−150

−100

−50

0

50

100

150

200

Strain gauge position [−]

Max

imum

Mic

roSt

rain

[−]

A1B1

Fig. 6. Strain influence line during the passage of the vehicle at �5 km/h at twodifferent alignments (A1 and B1), along centre beam L2. The results concern thevehicle first axle.

0 2 3 4 6 7 8 10−150

−100

−50

0

50

100

150

200

Strain gauge position [−]

Max

imum

Mic

roSt

rain

[−]

A1A2

−50

0

50

100

150

200

xim

um M

icro

Stra

in [−

]

B1B2

(a)

(b)

148 B. Zuada Coelho et al. / Engineering Structures 48 (2013) 144–154

in cross beam T2 is removed, the load that was received by thissupport needs to be transferred, therefore a stress redistributionoccurs towards the adjacent cross beams (strain gauges 22 and 28).

0 2 3 4 6 7 8 10−150

−100

Strain gauge position [−]

Ma

Fig. 7. Strain influence line of centre beam L2, during the passage of the vehicle at�5 km/h for: (a) test A (b) test B. The results concern the vehicle first axle.

4.2. Dynamic results

The dynamic tests were performed at three different vehiclespeeds. The goal was to evaluate the effect of the speed, as wellas changes in dynamic properties due to damage of the modularjoint.

Fig. 9 shows the time history strain during the passage of thevehicle at 70 km/h along alignment A. The same conclusions asdrawn before still apply for the joist (Fig. 9b). Again, a factor of�1.6 is found for the ratio between the rear axle and front axlestrain on the joist. For the centre beam (Fig. 9a) it is found that

the strain amplitude is higher for the first axle than for therear axles. This is attributed to inaccuracies during the vehicle

21 22 23 24 25 26 27 28 290

20

40

60

80

100

Strain gauge position [−]

Max

imum

Mic

roSt

rain

[−]

B1B2

Fig. 8. Maximum strain during the passage of the vehicle at � 5 km/h for twodifferent tests (B1 and B2), on box joists. The results concern the vehicle first axle.

0 0.1 0.2 0.3 0.4 0.5 0.6−20

0

20

40

60

80

100

120

140

160

Time [s]

Mic

roSt

rain

[−]

0 0.1 0.2 0.3 0.4 0.5 0.6−20

0

20

40

60

80

100

120

140

160

Time [s]

Mic

roSt

rain

[−]

(a)

(b)

Fig. 9. Strain during the passage of the vehicle at �70 km/h for test A1: (a) oncentre beam L2, device 3 and (b) on box joist 6, device 25.

0 2 3 4 6 7 8 10−10

−8

−6

−4

−2

0

2

4

6

8

Strain gauge position [−]

Dyn

amic

am

plifi

catio

n fa

ctor

[−]

50 km/h70 km/h90 km/h

0 2 3 4 6 7 8 10−10

−8

−6

−4

−2

0

2

4

6

8

Strain gauge position [−]

Dyn

amic

am

plifi

catio

n fa

ctor

[−] 50 km/h

70 km/h90 km/h

(a)

(b)

Fig. 10. Dynamic amplification influence line for centre beam L2, during thepassage of the vehicle at several speeds for: (a) test A1 and (b) test B1. The resultsconcern the vehicle first axle.

B. Zuada Coelho et al. / Engineering Structures 48 (2013) 144–154 149

travelling along the joint (see Table 2). During the test the vehicledid not cross the joint along alignment A, but the right wheel

crossed the joint towards joist 6 (west). This causes the point ofload application of the rear wheels to be closer to the cross beam,causing a reduction of the strain magnitude.

After the vehicle has crossed the joint a free vibration takesplace, due to the dynamic effect of the interaction between thevehicle and the modular joint.

Fig. 10 presents the influence line of the dynamic amplificationfactor (DAF) for all vehicle speeds. This is defined as:

DAF ¼ 1þ edyn � esta

esta; ð1Þ

where edyn is the maximum dynamic strain, and esta is the maximumstatic strain. For test A1 (Fig. 10a), with exception of strain gauges 7and 8, no noticeable dynamic amplification is found, which is inagreement with previous analytical results, regarding this expan-sion joint [17]. For strain gauges 7 and 8 a significant dynamicamplification occurs, especially on strain gauge 8. This is relatedto the misalignment of the vehicle passage. Since the left wheel iscrossing the joint along strain gauge 8, which is located next to asupport (joist 7), a slight misalignment of the vehicle causes a sig-nificant increase of the strain on this point (see Table 2).

In test B1 (Fig. 10b), the right wheel is crossing the joint on topof joist 6, therefore any misalignment of the vehicle position hassubstantial effect on the strain gauges adjacent to it (strain gauges

150 B. Zuada Coelho et al. / Engineering Structures 48 (2013) 144–154

3, 4, 6 and 7). The left wheel of the vehicle is crossing at midspan ofthe centre beam (along strain gauge 10), therefore its effects is ofminor importance for the strain gauges located next to it. The fig-ure exhibits significant DAF and two distinct influence lines, fortest at 50 km/h and tests at 70 and 90 km/h, respectively. This dif-ferences are caused by the vehicle misalignment (see Table 2).When comparing the vehicle passages at the different speeds it fol-lows that, the vehicle crossed the joint on the west side of crossbeam for the test at 50 km/h, while for the other two speeds thevehicle crossed on the east side of the cross beam. This causesthe inversion of the strain distribution along the centre beam,and is responsible for the high values of the DAF.

The dynamic tests were also performed for the scenario inwhich the sliding bearing of cross beam T2 was removed (testsA2 and B2). Fig. 11 shows the variation of the strain with timefor a vehicle passage at �70 km/h. From the figure it follows thata significant free vibration occurs on the centre beam. The ampli-tude of this free vibration is significantly higher, than the one re-corded for the normal situation (test A1 – Fig. 9a). The higherstrain is expected to lead to early fatigue problems, since the num-ber of cycles with high amplitude increases. A peak to peak strainlevel of �300 microstrains is observed, which corresponds to a ten-sion of approximately 63 GPa. Depending on the detailing of thewelded joint, this may lead to damage. Nevertheless, it should be

0 0.1 0.2 0.3 0.4 0.5 0.6−200

−150

−100

−50

0

50

100

150

200

Time [s]

Mic

roSt

rain

[−]

0 0.1 0.2 0.3 0.4 0.5 0.6−200

−150

−100

−50

0

50

100

150

200

Time [s]

Mic

roSt

rain

[−]

(a)

(b)

Fig. 11. Strain during the passage of the vehicle at �70 km/h for test A2: (a) oncentre beam L2, strain gauge 3 and (b) on box joist 6, cross beam T2, strain gauge25.

noted that this strain measurements were not located at thewelded joint. The wheel peak is no longer clearly identified, sincethe maximum and minimum peaks are similar in magnitude. Thisclearly shows a different behaviour of the centre beam, in the caseof sliding bearing degradation. Also the period of the free vibrationis increased. The strain level at the cross beam (Fig. 11b) is reduced,and, as for the centre beam, a significant amplitude of the freevibration is recorded.

The effect of the vehicle speed on the strain amplitude is pre-sented in Fig. 12 for the scenarios without sliding bearing (testA2 and B2). On test A2 (Fig. 12a) a significant amplification is no-ticed for strain gauges 4 and 6. These are located next to the slidingbearing that has been removed. For the other strain gauges no sig-nificant dynamic amplification has been found, for the differentvehicle speeds. For the points where the wheel loads are applied(strain gauges 3 and 8) no significant amplification occurs, sincethe strain magnitude is governed by the quasi-static component.The maximum dynamic amplification factor registered for straingauges 3 and 8 is, respectively, �1.2 and �1.4.

On test B2 the largest dynamic amplification is observed forstrain gauges 3 and 7. A similar behaviour to test A2 is found.The strain gauges that are under the location of wheel passage(in between strain gauges 4 and 6 and strain gauge 10), exhibit asmall dynamic amplification (<1.4), due to the quasi-static

0 2 3 4 6 7 8 100

1

2

3

4

5

6

7

Strain gauge position [−]

Dyn

amic

am

plifi

catio

n fa

ctor

[−] 50 km/h

70 km/h90 km/h

0 2 3 4 6 7 8 100

1

2

3

4

5

6

7

Strain gauge position [−]

Dyn

amic

am

plifi

catio

n fa

ctor

[−] 50 km/h

70 km/h90 km/h

(a)

(b)

Fig. 12. Dynamic amplification influence line for centre beam L2, during thepassage of the vehicle at several speeds for: (a) test A2 and (b) test B2. The resultsconcern the vehicle first axle.

0 50 100 150 2000

1

2

3

4

5

6

Frequency [Hz]

Mic

roSt

rain

am

plitu

de [H

z−1]

A1A2

Fig. 14. Strain during the passage of the vehicle at �70 km/h for two different tests(A1 and A2), device 3 in the frequency domain. First axle.

Table 4Eigenfrequencies (Hz) of a continuous simple supported beam with constantproperties along the longitudinal direction.

Mode Normal scenario Without fixation

1 116.6 45.52 127.8 125.03 144.3 143.7

B. Zuada Coelho et al. / Engineering Structures 48 (2013) 144–154 151

character of the loading. In both tests it is found that, with theexception of the locations where the load is applied, a significantdynamic amplification takes place for the spans that are adjacentto the removed sliding bearing. The reason for such amplificationis discussed later on.

Previously it was found that the misalignment of the vehicledictates the strain distribution along the centre beam for the differ-ent speeds. This does not apply for tests A2 and B2, since the spanof the centre beam is larger due to the lack of support, hence theinfluence of the vehicle alignment is diminished.

The response on box joist 5 (strain gauge 22) is presented inFig. 13. A clear change in the strain magnitude is found betweentests B1 and B2. This is related to the strain redistribution, due tolack of support, as found for the centre beam (Fig. 7). Again, a sig-nificant amount of free vibration takes place in the absence of fix-ation, that is not present for test B1.

One of the main observations from the previous analyses con-cerns the changes in the dynamic properties of the centre beam,when the cross beam is not supported. A significant free vibrationtakes place, with a different period and with higher amplitudes.Also a significant dynamic amplification takes place on the centrebeam for the spans next to the removed sliding bearing. This canbe better explained by considering the free vibration representedin the frequency domain (Fig. 14). The figure shows a peak inamplitude at a frequency of �55 Hz for test A2 where the slidingbearing has been removed (the peaks occur at the same frequen-cies for all the vehicle speeds). If the sliding bearing is available(test A1), the main frequency peaks occur at �115 and �130 Hz.Table 4 contains the first three eigenfrequencies computed for acontinuous simple supported beam, following the analytical solu-tion presented in [18]. Comparing the experimental results(Fig. 14) with the analytical solution, a good agreement for testA1 is found. For the case where the sliding bearing is removedthe agreement is less (55 Hz experimental and 45.5 Hz analytical),likely because the cross beam is not completely detached from thecentre beam (due to welding and sliding spring), therefore acts likea spring. This is in agreement with the strain influence line(Fig. 7a), as shown before, and the eigenfrequencies are in agree-ment with results from literature regarding this type of modularexpansion joints [17]. Thus, the lack of fixation causes a reductionof the first natural frequency for the centre beam.

The frequency of load application, is defined as [8,17,19]:

f ¼ vlv þ lb

; ð2Þ

0 0.1 0.2 0.3 0.4 0.5 0.6−60

−40

−20

0

20

40

60

Time [s]

Mic

roSt

rain

[−]

B1B2

Fig. 13. Strain during the passage of the vehicle at �70 km/h for two different tests(B1 and B2), on box joist 5, cross beam T2, device 22.

where v is the vehicle speed, and lv and lb the length of, respectively,the wheel print and the top flange, varied between 38 and 69 Hz forvehicle passages at, respectively, 50 and 90 km/h. Therefore, fortests without sliding bearing, the loading frequency is within thesame range as the eigenfrequency of the centre beam. For testsA2 the vehicle excites the joint with frequencies of 36, 49 and55 Hz for the passages at 50, 70 and 90 km/h. Especially the pas-sages at 70 and 90 km/h have input frequencies that are very closeto the first natural frequency of the joint, which causes the reso-nance of the structure. This is the justification for the higher DAFfound in Fig. 12 for these two vehicle passages. Similar results havebeen experimentally observed by others [6], for this type of expan-sion joints.

5. Numerical analysis

5.1. Model description

The numerical analysis was performed by means of a Finite Ele-ment Model (FEM) built in Matlab [20]. The model consists of threedimensional Euler beam elements for the centre and cross beams.The cross beam fixations (sliding springs and sliding bearings) aresimulated by means of a spring/damper system. The vehicle ismodelled through moving loads of constant magnitude, travellingat a constant speed. The timestep used for the analysis is5 � 10�4 s, smaller than the limit of Tmin/10, where Tmin representsthe minimum period of interest, which was considered to be0.005 s [21]. The system of equations is solved implicitly by meansof the Newmark method. Table 5 shows the properties of both cen-tre and cross beams. An overview of the mesh used for the analysisis available in Fig. 15. The damping values of the modular joint, aswell as the properties of the fixations, were derived from labora-tory tests on the centre beam [22], being the summarised in Tables5 and 6.

Table 5Properties of the modular expansion joint.

Property Value

Centre beam Cross-beam

Length (m) 18.15 0.80Span (m) 1.785 0.51Height (m) 0.12 0.13Width (m) 0.090 0.040Linear mass (kg/m) 46.60 40.82Density (kg/m3) 7850 7850Area (cm2) 59.3 52Shear area (cm2) 35.6 31.2Moment of inertia along y-axis (cm4) 1158 732Moment of inertia along z-axis (cm4) 399 64Torsional inertia (cm4) 67.3 202Vertical fixation stiffness (kN) – 1 � 105

Horizontal fixation stiffness (kN) – 1 � 105

Damping of the fixation (kNm/s) – 4.3 � 102

Fig. 15. Finite element mesh of the modular expansion joint.

Table 6Damping ratio of the modular expansionjoint.

Frequency (Hz) Damping ratio (%)

0 050 1

100 5118 40

3000 40

0 0.01 0.02 0.03 0.04 0.05 0.06−20

0

20

40

60

80

100

120

140

Time [s]

Mic

roSt

rain

[−]

ExperimentalNumerical

Fig. 16. Vertical strain at the middle point of the centre beam (point 1), during thepassage of an axle at 70 km/h – test A1.

0 0.05 0.1 0.15 0.2−100

−50

0

50

100

150

200

Time [s]

Mic

roSt

rain

[−]

ExperimentalNumerical

Fig. 17. Vertical strain at the middle point of the centre beam (point 1), during thepassage of an axle at 70 km/h – test A2.

152 B. Zuada Coelho et al. / Engineering Structures 48 (2013) 144–154

5.2. Validation with experimental data

The validation of the FEM model is made by comparing thenumerical with the experimental results. Fig. 16 shows the com-parison of the strain for point 1, located at the middle span ofthe longitudinal beam (Fig. 15), during the passage of one axle(36.5 kN/wheel) at 70 km/h. The distance between the two wheelsis 2.10 m, which corresponds to the vehicle used for the tests (seeFig. 4). It is assumed that one wheel is crossing the joint along thealignment that contains point 1 and the other wheel 2.10 m to-wards the positive side of Y axis (Fig. 15). From Fig. 16 it followsthat a close agreement exists between experimental and numerical

results, which is a good indication of the suitability of the numer-ical analysis.

The comparison for the scenario where the sliding bearing ofthe cross beam is removed is given in Fig. 17. The removal of thesliding bearing is simulated by setting the properties of thespring/damper system that supports the cross beam to zero. Agood agreement is found for the peak value. However, the freevibration is less accurately predicted by the numerical model. Thisis most likely explained by changes in the properties of the crossbeams. From visual inspections it was concluded that the crossbeams have been strengthened during the years, therefore no exactquantification of the mass and stiffness can be made, as well as forthe damping parameters. This is more important for the case with-out fixation, since a resonance phenomena is present.

5.3. Parametric study – effect of cross beam fixation degradation

The effect of fixation deterioration is evaluated by changing theproperties of the spring/damper system that supports the crossbeam. Several scenarios were evaluated in order to simulate thegradual degradation, by reducing the stiffness in the three direc-

115

120

125

130

135

Max

imum

Mic

roSt

rain

[−]

k=105

k=104

k=103

k=102

k=0

B. Zuada Coelho et al. / Engineering Structures 48 (2013) 144–154 153

tions of the cross beam fixation (Table 7). In all the simulations thedamping value of the spring/damper system was considered to bezero. From preliminary analyses this showed to have no influenceon the results.

Fig. 18 shows the strain for two points (points 1 and 2) due tothe passage of an vehicle axle at 70 km/h. It follows that the fixa-tion degradation can be retrieved. As the degradation increases,by reducing the stiffness of the fixation, the strain level increases,together with the development of a significant free vibration, dueto resonance of the centre beam. The strain increase is more signif-icant for point 2 than for 1. This is in agreement to the experimen-tal results from Section 4.1. Since point 1 is located under point ofload application the major part of the strain is due to static defor-

Table 7Vertical fixation properties for damage detection.

Stiffness (kN) Note

105 Good support104 –103 –102 –0 No support

0 0.02 0.04 0.06 0.08 0.1−50

0

50

100

150

200

Time [s]

Mic

roSt

rain

[−]

k=105

k=104

k=103

k=102

k=0

0 0.02 0.04 0.06 0.08 0.1−140

−120

−100

−80

−60

−40

−20

0

20

40

Time [s]

Mic

roSt

rain

[−]

k=105

k=104

k=103

k=102

k=0

(a)

(b)

Fig. 18. Vertical strain during the passage of an axle at 70 km/h – test A. Effect offixation degradation. (a) point 1 and (b) point 2.

0 20 40 60 80 100110

Speed [km/h]

Fig. 19. Effect of fixation damage on the dynamic amplification factor during thepassage of the vehicle at several speeds along alignment A, for point 1.

mation. Point 2 is not directly submitted to the vehicle loading,therefore the dynamic effects are more important for this point.

Fig. 19 presents the maximum strain magnitude in point 1,function of the vehicle speed, for different degradation scenarios.Depending on the fixation degradation level, the shape of the straincurve changes. This is related to the changes in the modal proper-ties of the centre beam. As the cross beam fixation starts to de-grade, the cross beam starts acting like a spring rather than afixed support, causing a variation in the eigenfrequencies of thecentre beam. Therefore, for each different vehicle speed the struc-ture behaves differently, depending on the fixation stiffness.

For vehicle speeds up to 40 km/h no significant dynamic ampli-fication is found in the numerical analyses. For higher speeds thedynamic amplification becomes more important, especially forsmall values of fixation stiffness. It is found that for stiffness valuesof 105 and 104 kN/m the results coincide, though the stiffnesschanges by a factor 10. The same applies for the small stiffness re-sults, 0 and 102 kN/m. These correspond to the scenarios with andwithout damaged fixations.

6. Conclusions

This article presents the results of a detailed study on the dy-namic behaviour of a modular expansion joint, during one overrolling test. The test was part of a research program regardingthe development of an early warning monitoring system for theMartinus Nijhoff Bridge, aiming to evaluate the effect of degrada-tion on cross beam fixations. Moreover a numerical model of themodular joint was develop and validated.

The experimental results have shown that both centre and crossbeams behave differently, depending on the effectiveness of thesupport of the cross beam. As the cross beam sliding bearing is re-moved, the dynamic behaviour of the centre beam changes, whichhas consequences on the strain influence lines and eigenfrequen-cies of the centre beam. If the sliding bearing is effective the crossbeam behaves as a support of the centre beam. In the case of re-moval, the cross beam works as a spring, and causes a drop ineigenfrequencies (116.6–45.5 Hz for first eigenfrequency). Thisshows that is possible to implement systems which detect damagebased on variation of the modal properties.

It was shown that the vehicle alignment has severe implicationson the interpretation of the dynamic tests. The vehicle speedcauses no significant dynamic amplification on the response of

154 B. Zuada Coelho et al. / Engineering Structures 48 (2013) 144–154

the modular joint for the test with sliding bearing. This is due tothe quasi-static character of the response. For the tests in the stateof serviceability (A1 and B1) the maximum DAF was �1.4, valuethat is below 1.7 commonly used by designers of expansion joints[6]. In the case of sliding bearing degradation a significant increaseof the DAF was measured (values up to 6). A resonance phenomenawas identified for the scenario without sliding bearing which isassociated with high number of stress cycles. These two combinedeffects have severe implications on the resistance to fatigue, beinglikely to cause premature fatigue failure.

The effect of the vehicle speed is more pronounced for the testwithout sliding bearing, since the loading frequency matches theeigenfrequency of the centre beam, causing its resonance. Fromthe numerical analysis it was found that there is a cut off speed(40 km/h) up to which no significant dynamic effects occur dueto the moving loads.

The numerical results have shown that it is possible, with a sim-ple beam model, to accurately capture the general behaviour of thestructure, in particular the maximum strain level on the centrebeam. It also allows to evaluate the gradual degradation of thecross beam fixations. This is a main asset that can be incorporatedinto continuous monitoring systems, in order to evaluate the levelof degradation of cross beam fixations. By combining the strainmonitoring (and changes in modal properties) together withnumerical simulations it is possible to estimate the degradationstate of modular expansion joints, and prevent premature failureby defining appropriate criteria for intervention.

References

[1] Pijpers R. Modular expansion joints. The effect of static and fatigue trafficloading according Eurocode 1, (EN 1991-2, traffic loads on bridges), on varioustypes of modular joints. MSc. Thesis, Delft University of Technology; 2005.

[2] Lima J, Brito J. Inspection survey of 150 expansion joints in road bridges. EngStruct 2009;31(5):1077–84.

[3] Dexter RJ, Mutziger M, Osberg C. Performance testing for modular bridge jointsystems, Tech. Rep., National Cooperative Highway Research Program – Report467; 2002.

[4] Roeder C. Fatigue and dynamic load measurements on modular expansionjoints. Construct Build Mater 1998;12(2–3):143–50.

[5] Chang L, Lee Y. Evaluation of performance of bridge deck expansion joints. JPerform Construct Facil 2002;16(1):3–9.

[6] Ancich E, Chirgwin G, Brown S. Dynamic anomalies in a modular bridgeexpansion joint. J Brid Eng 2006;11(5):541–54.

[7] Ravshanovicha K, Yamaguchi H, Matsumoto Y, Tomida N, Uno S. Mechanism ofnoise generation from a modular expansion joint under vehicle passage. EngStruct 2007;29(9):2206–18.

[8] Ghimire J, Matsumoto Y, Yamaguchi H, Karahashi I. Numerical investigation ofnoise generation and radiation from an existing modular expansion jointbetween prestressed concrete bridges. J Sound Vibr 2006;328(1–2):129–47.

[9] Crocetti R, Edlund B. Fatigue performance of modular bridge expansion joints. JPerform Construct Facil 2003;17(4):167–76.

[10] Kalkman I, Lentzen S, Courage W, Napoles OM, Galanti F. Performanceindicator of a bridge expansion joint. In: Papadrakakis, M., Fragiadakis, M.,Plevris, V. (Eds.), Computational methods in structural dynamics andearthquake engineering, ECCOMAS; 2011. p. 1–20.

[11] Peelen W, Vervuurt A, Zuada Coelho B, Leendertz H. Early warning monitoringsystem of modular expansion joints based on dynamic behavior. in:MEMSCON – towards intelligent civil infrastructure; 2012. p. 1–11.

[12] Leendertz J, Galanti F, Butz C. Dynamic testing and analysis of existing modularexpansion joint. In: 35th Annual symposium of IABSE, taller, longer, lighter -meeting growing demand with limited resources: IABSE; 2011. p. 1–8.

[13] ETAG no. 032, Guideline for European technical approval of expansion jointsfor road bridges. Part 8: modular expansion joints, Tech. Rep.; 2010.

[14] Gong S. A study of in-plane dynamics of tires. Ph.D. thesis, Delft University ofTechnology; 1993.

[15] Vervuurt A, Galanti F, Courage W, van Holsteijn C, van der Meer P. Resultatenoverroll test Martinus Nijhoffbrug 12 September 2010 (in Dutch). Tech. Rep.,TNO; 2010.

[16] Vervuurt A, Zuada Coelho B, Dieteren C, van Holsteijn G, Middeldorp F.Resultaten overroll test Martinus Nijhoffbrug (in Dutch). Tech. Rep., TNO;2011.

[17] Steenbergen M. Dynamic response of expansion joints to traffic loading. EngStruct 2004;26(12):1677–90.

[18] Timoshenko S. Vibration problems in engineering. 3rd ed. Van Nostrand; 1955.[19] Ancich E, Brown S, Chirgwin G. The role of modular bridge expansion joint

vibration in environmental noise emissions and joint fatigue failure. In:Proceedings of acoustics 2004 conference; 2004. p. 135–40.

[20] The Mathworks. Matlab R2009b; 2009.[21] Bathe KJ, Wilson EL. Stability and accuracy analysis of direct integration

methods. Earthq Eng Struct Dynam 1973;1(3):283–91.[22] Lentzen S, Galanti F. Bridge expansion joint – Lab test (in Dutch). Tech. Rep.,

TNO; 2010.