Experimental Investigation and Finite Element Analysis on Electromagnetic Compression Forming Processed Aluminum Alloy Tubes

Simulation Modelling Practice and Theory 19 (2011) 1795–1810

Contents lists available at ScienceDirect

Simulation Modelling Practice and Theory

journal homepage: www.elsevier .com/ locate/s impat

Experimental investigation and finite element modelling of the effectsof flow velocities on a skewed integral bridge

M.M. Fayyadh ⇑, S. Akib, I. Othman, H. Abdul RazakDepartment of Civil Engineering, Faculty of Engineering, University of Malaya, Kuala Lumpur, Malaysia

a r t i c l e i n f o

Article history:Received 18 August 2010Received in revised form 2 March 2011Accepted 20 April 2011Available online 7 June 2011

Keywords:Finite element modellingIntegral bridgesPile scouringStructural behaviour

1569-190X/$ - see front matter � 2011 Elsevier B.Vdoi:10.1016/j.simpat.2011.04.010

⇑ Corresponding author. Tel.: +60 142224029.E-mail address: [email protected] (M.M

a b s t r a c t

This paper presents finite element modelling of the effects of different flow velocities onthe structural behaviour of a skewed integral bridge. Flow velocities affect the scour depthsat the piles of a bridge and thus affect its structural behaviour. Laboratory tests on a scaled-down hydraulic model along with numerical modelling were performed to simulate thestructural behaviour of the scoured integral bridge. A finite element package was usedfor the numerical modelling work, and the displacements and strains corresponding tothe measured locations on the physical model were extracted. The experimental andnumerical results for the case of maximum scour depths were compared.

� 2011 Elsevier B.V. All rights reserved.

1. Introduction

Over the past 30 years, there has been an increasing need to replace the current stock of bridges in Malaysia, since mod-ern bridge systems have a lower life cost. In particular, the huge maintenance cost incurred for the expansion joints andbearings of conventional bridges has been a major concern for local state councils and authorities. The government has alsoacknowledged the exceptional rise in maintenance costs, the concurrent decrease in highway revenues, and the serious im-pact on future highway construction projects. Bridges of total lengths less than 60 m are more economical and cost-effectiveif designed as integral bridges having full structural continuity and smaller numbers of expansion joints. In view of theserequirements, integral bridges have become feasible alternatives, and a dramatic increase has been noted in the constructionof such bridges in Malaysia.

However, since the use of integral bridges is still relatively new in Malaysia, design factors relating to the effects of naturalhazards and local weather as well as environmental conditions are unavailable and yet to be established. One of these factorsis the effect of floods on integral bridges, which is of prime concern to bridge designers. Since the 1920s, the country hasexperienced major floods during seasonal monsoons, causing a large concentration of surface-runoff that exceeds the capac-ities of most rivers. States located on the east coast of Peninsular Malaysia such as Kelantan, Terengganu, Pahang and Johorare badly affected by these massive seasonal floods.

It is only since the early 1990s that flash floods have become a concern in urban areas; these flash floods are perceived tobe the most critical of flood types. Hence, detailed investigations on the effects of floods on integral bridges are vital. Flashfloods are the most dangerous kind of floods, because they combine the destructive power of a flood with incredible speedand unpredictability. Examining flash floods and scouring effects, Tregnaghi et al. [25,26] found that clear-water boundaryconditions can be extended to sediment-supply tests if specific supply input conditions hold. Moreover, experiments show

. All rights reserved.

. Fayyadh).

1796 M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810

that the ratio between the final scour depth and the potential scour depth at a bed sill for a given hydrograph can be esti-mated as a function of the identified temporal parameter.

The general aim of the study was to investigate of the effects of floods and scour on a skewed integral bridge. The specificobjective of this study was to investigate the behaviour of scour and the relationships between scour depth and structuralbehaviour of skewed integral bridge. The other parameters include time evolution of scouring, loading, strain and displace-ment at different locations of the bridge. Although time evolution for scouring data was taken, the main focus of the exper-imental work and modelling investigation was to compare the maximum scour after 24 h.

The developed model consists of a single span integral bridge that focuses on the local scour on the abutments. Previousresearchers have categorized abutment as short and long based on the observed flow features. Kwan [14] investigated theeffect of local scour on short abutments. The study showed that the local scour on short abutments and piers are similar. Theprinciple features of the flow are the down flow ahead of the abutments, a principal vortex, and wake vortices. Many articleshave been published on matters pertaining to scour on conventional bridge foundations [6,7,15,16,19,20,22,23]. Local scourstudies focusing on the effects of time have been published by Kwan [14] and Melville and Chiew [21]. Akib et al. [1] con-cluded that local scour on a double-row pile integral bridge is higher than on a single-row pile integral bridge in a two-stagechannel. Scours on pier and pile groups were well researched and documented by Kambekar and Deo [13], Sumer et al. [24],Ashtiani and Beheshti [3], and Coleman [9]. Martin-Vide et al. [18] examined the problem related to the interaction of twowidths (pier and piles) that were set at different elevations with respect to the riverbed; a width-weighting method was rec-ommended due to greater scouring when the riverbed is closer to the base of the pier. Akib et al. [2] proposed a countermea-sure to reduce scour on a semi-integral bridge pier using Epipremnum aureum.

Scouring around a submerged vertical cylinder in a steady current was studied both experimentally and numerically byZhao et al. [30]. A three-dimentional finite element model was developed for local scouring simulation. They found that thesimulation modelling results are smaller than the experimental results (around 20%). Huang et al. [12] investigated the scaleeffect on turbulent flow and sediment scour using the three-dimentional computational fluid dynamics model. The physicalscale and the boundary velocities were setup for the small-scale model based on Froude similarity law. The results werecompared for two cases: small-scale model and full-scale model. The study shows that ignoring Reynolds similarity in phys-ical modelling may result in errors for scouring in large bridge piers. They show also that not all of the physical quantitiescould be provided; some of these quantities in a turbulence flow are difficult to measure such as the vortex, which is themajor factor responsible for base scouring. Moreover, in comparison in physical modelling, either Reynolds or Froude sim-ilarity has to be ignored due to difficulties in meeting both similarity laws.

Therefore, perfect results are difficult to obtain in view of many factors involved that cannot be modelled directly usingnumerical simulation. The effect of the turbulent flow on the local scour around a single spur dyke was investigated by Zhanget al. [29]. They simulate the complex local flow field around the scour area using a three-dimentional nonlinear model thatemploys the finite volume method. They found that the simulation results are reasonably consistent with those of the exper-imental measurements. Down-flow, horse show vortex, and wake vortex are important parameters affecting the local scouring.

Bateni and Jeng [5] predicted scouring for group of piles using an adaptive neuro-fuzzy inference system model. The mod-el used two combinations of input data to predict the scour depth: the first input combination involved dimensional param-eters such as wave height, wave period, and water depth, while the second combination contained non-dimensionalnumbers including the Reynolds number, the Keulegan–Carpenter number, the Shields parameter, and the sedimentnumber. The results show that the model better predicted the scour depth with the original dimensional rather than thenon-dimensional numbers. The sensitivity analysis showed that the scour depth is governed mainly by the Keulegan–Carpeneter number, and wave height has a greater influence on scour depth than the other independent parameters. Leeet al. [17] applied the Back-Propagation Neural Network (BPN) to predict the scour depth in order to overcome the problemof exclusive and the nonlinear relationships. They verify the observations obtained from thirteen US states and found thatthe scour depth could be efficiently predicted using the BPN compared with conventional methods.

The present research fills the gap of literature, where there is limited experimental research on the effect of scouring onthe structural behaviour of the integral bridges. Moreover, the present study also explores the simulation effect of flowvelocities.

2. Experimental work

The flow velocities effects on the structural behaviour of skewed integral bridge were determined from nine tests (threefor three velocities; followed by six repeated tests) performed in the re-circulating flume at the Hydraulic Laboratory of theUniversity of Malaya. The flume was 16 m long, 0.6 m wide and 0.57 m high. The model was tested in trapezoidal floodplainsof 0.6 m wide, 2 m long, and 0.184 m high. Three parts of both side walls had clear Perspex panels, which are useful for clearand direct observation of the flow, scouring, and sediment transport process. A re-circulating pipe system was used: a cen-trifugal pump raised the water from a sump to the upstream end of the flume, from which it went back to the sump througha return pipe. The discharge was measured using a v-notch weir in the tank placed at the outlet of the flow. The bed sills/floodplains used in all experiments were 142.2 mm thick by 18.4 cm high Perspex plates. The two floodplains were located atthe longitudinal abscissa (starting from the inlet) x = 4.00 m and x = 6.00 m; the distance between the floodplains, L, was con-stantly equal to 2.00 m. The floodplains were used to observe the scouring and eroded sediments.

Table 1Bridge part dimensions (units in mm).

Slab Inclined length Width Thickness597 172 6.5

Abutment Length Width Thickness172 36 26.7

Piles Depth Diameter170 8

Table 2Scaling factor for each quantity based on similarity law.

Quantities Scaling factors Prototype Model

Length 1:75 44.78 m 597 mmVelocity of water 1:8.66 2.165 m/s 0.25 m/sAcceleration (gravity) 1:1 9.81 m/s2 9.81 m/s2

Mass (four wheels) 1:860.683 1.028 � 106 kg 1.1944 kgForce 1:860.683 450 kN 5.228 � 10�4 kN

Fig. 1. Skewed integral bridge setup in the flood plain.

M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810 1797

The riverbed material selected was uniform fine sand with a median particle size of d50 = 0.13 mm in diameter. The sandswere sieved in the range of 2 mm to 63 lm in order to obtain a uniform size for the bed sediment and according to BritishStandard 1377: Part 2: 1975. The sand that was trapped between these series of sieves had a range of 2 mm, 1.8 mm,600 mm, 425 mm, 300 lm, 212 lm, 150 lm, and 63 lm. In addition, the specific gravity for the sand particles was deter-mined for each type of sand. The sand diameter was chosen approximate to the average scale of all types of soils at the fieldsite. This type of sand did not represent all types of soils for the actual rock and gravel in the case of the prototype structure,since it is impossible to scale down from the real soil conditions at the site.

Flow shallowness y was fixed at 1.5 cm and constant to serve as a control after being scaled down from the real waterlevel in the actual flood. All runs used a skewed integral bridge model built on a deck fixed into an abutment, which wassupported by a set of piles on both sides. Each side contained seven circular piles embedded at the base of the flood channeland fixed into holes 1 cm in depth at the bottom of the flow channel. There was a tolerance of 1 mm in the diameter that wasfilled with high strength adhesive material to model the fixed support conditions and to prevent transitional and rotationaldisplacement in three dimensions.

The dimensions of each part of the bridge are shown in Table 1, and the scaling quantities are shown in Table 2. The modelwas made using perspex material having a density ranging from 1144 to 1250 kg/m3, a modulus of elasticity of 2050–2300 kN/m2, and Poisson’s ratio of 0.39. The model was set up in the flood plain as shown in Fig. 1. Fig. 2 shows the planview and the skew angle of the model. The skew angle is 34� against the centre line of the flood channel. According to BritishStandards BS 4296, the maximum skew angle for integral bridge should be 30�; a skew angle greater than this precludes theuse of integral bridge construction. However, a survey of recent practices revealed that designers are creating fully integralbridges with skews up to or slightly above this value. Yannotti et al. [28] reported that the maximum skew angle of integralbridge is 45�. Therefore, the skew angle for this study, 34�, is allowable and safe to be designed. Moreover, the velocities ef-fects on the structural behaviour of skewed integral bridge would not be similar if the angles were different due to the dif-ferent angle of attack on the water.

Each side of the piles was embedded into the flood plain. The elevation of the model is presented in Fig. 3, which showsthe depth of the bridge model and the bridge dimensions as perpendicular to the flow flume. Sand was poured into the floodplain until half of the abutment was covered. The channel was also filled 5 cm high with sand. The water level was set tosubmerge the bridge’s slab by 5–12 mm for all experiments in order to reach the effect of high water flow pressure onthe bridge model and to be sure that all bridge parts came under the effect of water pressure.

Since the stabilization of the water level and the velocity were important in this experiment, several pre-experimentalworks were conducted to ensure a specific velocity upstream, and precise water levels were achieved before beginning each

Fig. 2. Plan view of the skewed integral bridge.

Fig. 3. Elevated view of the skewed integral bridge (units in mm).

Fig. 4. Actual bridge setup during the test.

1798 M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810

experiment. The velocities chosen in this study were referred to the frequent real-life flooding at Sg. Lebir, which is situatedat Tualang, Kelantan. The flood velocity was 2.165 m/s (0.25 m/s after scaled down). The velocities chosen include slowerthan flooding (0.19 m/s), flooding (0.25 m/s), and higher than flooding (0.31 m/s). For all experiments, local scour depthswere developed from flat bed conditions (ds = 0 for t = 0), with scour depths measured using a vertical scale depth positionedat all piles. The pile diameters were 8 mm to simulate the actual bridge with the dimension scale of 1:75, which was chosenaccording to the flume’s limitation.

During the experiment, changes in scour depth on both sides of the abutment and piles were recorded. The experimentwas repeated three times with three different velocities (0.19 m/s, 0.25 m/s, and 0.31 m/s). For each velocity, the experimentwas repeated three times, and the average values were taken. The effect of these three different velocities on the scour depthwas recorded. For the first 100 min of water flow, the scour readings were taken at 10-min intervals. Subsequent readingswere recorded at intervals of 100 min for a continuous duration of 8 h and 20 min (500 min). The final readings were takenafter 24 h of running the experiment.

Fig. 5. Locations of strain gauges and LVDTs on bridge slab.

Fig. 6. Locations of strain gauges and LVDTs on the Q side piles.

M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810 1799

The set of piles on each side are named Q and P, where Q is the set of piles located upstream. Therefore, Q was the first tocome in contact with the water flow and thus the first to be affected by it; P was downstream and the last to be affected bythe water flow. The main data recorded, in addition to scour depth, were strain and displacement on the bridge slab and onthe Q set of piles. The actual bridge setup during the test is shown in Fig. 4. For strain, we used an electric resistant straingauge, which measures sensitivity in microns to any change in the strain. Strain gauges were coated with bituminous mate-rial to prevent water intrusion, which acted as a waterproof covering to the gauges. To measure displacement, we used theLinear Variable Displacement Transducer, which measures displacement in mm units and is sensitive up to 0.01 mm. Cali-bration was performed before starting the experiments. The strain gauges and LVDTs for recording the strain displacementon the bridge slab and Q piles were positioned as shown in Figs. 5 and 6.

3. Finite element modelling

Finite element modelling is widely used for the simulation of engineering problems. There are many studies related to theuse of finite element modelling as a simulation tool for structural elements such as the aerodynamic derivatives for thebridge girder [11]. It is also widely used for modelling the behaviour of composite materials, such as in the study microcracks effects on composite material properties [8]. Finite element modelling is a powerful technology that can be usedfor modelling dynamic loading such as the wind loading effect on long-span bridges [27]. Furthermore, finite element mod-elling can be used for modelling the dynamic response of structural elements [4,10].

Fig. 7. Geometric modelling of integral bridge with FE software.

1800 M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810

The paper addresses modelling and simulation aspects in hydraulic structure engineering, specifically, the effects ofvelocities and scouring on the structural behaviour of a skewed integral bridge. There is no literature comparing the effectsof velocities and scouring effects on the structural behaviour of a skewed integral bridge using finite element modelling. Boththe experimental work and modelling are essential in this study. Experimental results were applied to simulate the scouringeffects on the boundary constraints of the piles. In the case of no scouring, the full embedded length of the pile was sup-ported by the elastic constraints; in the case of scouring, there was no elastic constraint on the scoured part of the piles.

This study promotes an integrated approach by using Standard finite element software (DIANA), where standard finiteelement modelling provides a simple and time-consuming method to predict the behaviour of integral bridges under differ-ent flooding and scouring conditions. Finite element modelling allowed us to model the experimental procedure in the lab-oratory, the material used for constructing the integral bridge model, different flow velocities, and the scouring effect on thepiles’ support constrains. The study took advantage of the finite element software facilities to model the flow velocities andthe loading applied on it. Moreover, the study suggests a method to model the flow velocity effect on the scouring by con-trolling the support condition statues of the piles.

A three-dimensional model was used to simulate the skewed integral bridge. The bridge slab was modelled as a two-dimensional flat shell element on the x–z axis (the plan of the flood cross section and flow streamline). The abutmentwas modelled as two-dimensional flat shell elements on the y–z axis (the plane of the water depth and the flow streamlines).The flat shell elements can support in-plane and bending stresses. Piles are modelled as one-dimensional beam elementswith an axis in the y-direction (the water depth direction). The beam elements are defined as Beam Class II, which can sup-port compression, tension, and bending load. Fig. 7 shows the geometric modelling of the skewed integral bridge.

The applied loading included the bridge self weight in the �(ve) y-direction (the vertical and the water depth direction),water pressure above bridge slab in �(ve) y-direction, uplift water loading at the bottom of the slab in the +(ve) y-direction,uplift water pressure at the bottom of the abutment in the +(ve) y-direction, and water velocity pressure on the immersed partsof the bridge in the�(ve) z-direction (direction of the flow water streamlines). All of these loadings, which were related to thewater flow velocity, weight, and uplift pressure, were calculated manually and applied in the FE software as forces on a lines orsurfaces.

The bridge was modelled with vehicle loading to simulate the traffic loading. Since it is rather complex to simulate vehi-cles moving on the bridge, the vehicle loading was modelled as a non-moving truck in the mid-span of the bridge. For thepurpose of this investigation, the selection of loading on the model is based on type HB (abnormal vehicle load) as defined byBritish Standard 5400: Part 2 with the maximum 45 units of HB. One unit is equivalent to 10 kN per axle; therefore, load peraxle is equal to the number of units multiplied by the unit load (45 � 10 = 450 kN), and load per wheel equal to load per axledivided by four wheels (450/4 = 112.5 kN). In order to apply the load onto the model, reduction of the entire load to anappropriate scale is needed. In order to compare various quantities in the prototype and the model, load ratio is derivedto be equal to the sheer force ratio (SFR):

SFR ¼Wp=Wm ð1Þ

where Wp is the weight of the prototype, and Wm is the weight of the model.Since the weight of the prototype is 1.028 � 106 kg, and the weight of the model is 1.1944 kg, the SFR is equal to

8.607 � 105. After applying the scale factor to the loading at each wheel, the load per wheel on the model equals 13.4 g.The wheel loads were applied onto the mid-span of the bridge slab.

Assumptions adopted in the modelling of the experimental work are as follows:

– Boundary conditions were modelled as fixed components at the base of the piles at the bottom of the flow channel. In theexperimental work, the piles were embedded in holes of 1 cm depth made in the bottom of the flow channel, where there

Fig. 8. Effect of scouring on the elastic constraint of the piles.

Fig. 9. Modelling the scouring effect on elastic constraints for slow velocity flow.

M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810 1801

was a tolerance of 1 mm in the diameter. The tolerance was filled with high strength adhesive material to model the fixedsupport conditions and to prevent transitional and rotational displacement in three dimensions.

– The connection of the piles to the abutment, and the abutment to the bridge’ slab were modelled as fixed components. Inthe experimental work, bridge substructures were connected using high strength adhesive material referring to the inte-gral bridge concept (fixed and acted as a portal frame).

The sub-grade reaction of the soil where the bridge piles were embedded was modelled as elastic constraints with con-stant stiffness equal to the sub-grade constant of the experimental soil ranging between 11 and 21.4 N/mm. The scour effect

Fig. 10. Modelling the scouring effect on elastic constraints for fast velocity flow.

1802 M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810

was inserted in the model by changing the depth of the boundary elastic constraints. In the case of no scouring, the fullembedded length of the pile was supported by the elastic constraints; in the case of scouring, there was no elastic constrainton the scoured part of the piles. Fig. 8 shows the effect of the scour on the elastic constraints of the piles embedded in thesoil. The left picture shows the case of no scouring, where the full embedded length of the pile was supported by the sur-rounding soil. The right picture shows the case of scouring, where a part of the piles had already been scoured. The scouringeffect was included by decreasing the supporting length of the elastic constrains on the piles, where it was reduced corre-sponding to the scour depth.

The effect of different velocities on scouring was inserted similarly by changing the depth of the elastic constraints. Figs. 9and 10 show the elastic constraints on the piles for the slow and fast velocity flow, respectively; both figures show the dif-ference in scouring and its effect in the model.

The flat shell elements and the beam elements were meshed using fine sizes, where each element was divided into 100small elements, (see Figs. 9 and 10). Since the loading was within the linear limit of the experimental model material, struc-tural linear analysis was adopted. The experimental model was modelled using isoparametric linear elastic material usingthe perspex material properties as described previously.

4. Results and discussion

This section presents the results of this study and is divided into two parts. The first part discusses the effect of differentscour depth and velocities on the structural behaviour and the modelling of this effect using finite element software. Thesecond part discusses the effect of different velocities on the scour depth with time based on experimental results.

5. Comparison of finite element modelling with experimental results

A comparison is made between experimental and finite element modelling results. The main data compared is the strainat the locations of strain gauges SG1-SG8 with the displacement at locations of LVDT1–LVDT8 as shown in Figs. 5 and 6. Thenumerical modelling was done based on the final scour results, which were at the maximum scour after 24 h.

The values of the sub-grade stiffness, water level above the bridge slab, bridge model material mass, and bridge modelmaterial modulus of elasticity were adjusted for the purpose of refining and getting a more accurate representation ofthe finite element model compared with the physical model. The values chosen in order for the FE results to match as closeas possible with the experimental results were 11 N/mm for the soil sub-grade stiffness, perspex mass of 1250 kg/m3, per-spex modulus of elasticity of 2300 kN/m2. In addition the water depth above the bridge slab according to each flow velocitywere fixed at 12 mm, 5 mm and 9 mm for slow, medium and fast flow velocities, respectively.

The strain behaviour on the slab with the longitudinal strain in the X-axis was analyzed. Fig. 11 shows the experimentaland FE results of the strain gauges (SG1–SG5) located on the slab of the skewed integral bridge to measure the flexural strainof the slab.

Experimental results show that strain gauges SG1 and 2, which were located at the forehead of the mid-span of thebridge, showed the same trend: a decrease in the strain values when the flow velocity increased from 0.19 to 0.25 m/sand then an increase when the flow velocity increased from 0.25 to 0.31 m/s. For SG3, located at the back of the bridge slabmid-span, there was always a decrease when the flow velocity increased; this could be due to the effect of the backwardwater flow at the back of the slab. For strain gauge 4, located at the back side of the piles Q set, there was increment inthe strain when the flow velocity increased from 0.19 to 0.25 m/s, followed by a decrease when the flow velocity increasedfrom 0.25 to o.31 m/s. This could be due to the turbulent flow of water at the back of the bridge. For SG5, there was incre-ment in the strain values with the increase in the flow velocity due to the forward effect of flow velocity at the location ofSG5.

The finite element (FE) results were in good agreement at SG1, SG2 and SG5 where the trend was the same for the strainvalues obtained experimentally for different the flow velocities. The maximum difference was 4% for SG2 at fast velocitywhere FE gave higher strain than the experimental results. There were discrepancies in the FE results for the strain valuesat SG3 and SG4. For SG3, the experimental results showed a decrease in the strain values with increase in the flow velocity,while for the FE results the converse. The experimental results for SG4 showed an increase in the strain values for the med-ium flow velocity followed by a decrease for the fast flow velocity. However for the FE results, the strain values increasegradually as the flow velocity increases. The difference in the strain values between FE and experimental results at SG3and SG4 can be due to the effect of the back water flow and the turbulent flow of water at the back of the bridge. In orderto improve the accuracy of the FE model to get better compatibility with the experimental results at SG3 and SG4, the backwater flow force was manually inputted into the model where it was calculated as a percentage of the forward flow force.The back flow force was initially estimated within a range of 5–50% of the forward flow force and finally the value of 25%gave results very similar to the experimental results without affecting the strain values at SG1, SG2, and SG5. The forcewas applied as a line force at the level of the bridge slab and the updated strain gauge values at SG3 and SG4 after inputtingthe back flow force are shown in Fig. 12 and the difference in strain values is about 6%.

The vertical strains on piles on the y-axis were compared. Fig. 13 shows the experimental and FE results from straingauges (SG6–SG8) located on piles on the Q side of the skewed integral bridge. The numerical modelling was done based

Fig. 11. Experimental and FE results for strain gauges located on the slab.

M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810 1803

on the final scour results, which were at maximum scour after 24 h. The range of values for the soil sub-grade stiffness wasused to update the FE model, and a value of 11 kN/mm was finally adopted in order to obtain close agreement with theexperimental results.

Experimental results show that for SG7 and SG8, which are located at piles 4 and 6 on the Q side, the strain values chan-ged from positive when flow velocity was 0.19 m/s to negative when flow velocity increased to 0.25 and 0.31 m/s. Moreover,at 0.25 m/s flow velocity had higher strain values than 0.31 m/s. This behaviour could be due to the vortex at the back andsides of the piles as well as the uplift pressure effect on the bridge slab and abutment, which influenced the pile strains. ForSG6, located at pile 2 on the Q side, the experimental results were always negative, and 0.25 m/s flow velocity was higherthan 0.31 m/s strains. This could be due to the position of the strain gauge, which was located close to the front of the bridgeand caused forward flow higher than the effect of the vortex and the uplift force.

FE results for SG7 and 8 shown are similar to the experimental results, that is, a change from positive to negative strainvalues with the increase in flow velocity. This could be a good indicator for the FE model to simulate the strain at the piles.For SG6, both the FE and experimental results were negative. The FE model was not able to simulate the pattern when theflow velocity increased from 0.25 m/s to 0.31 m/s, whereby the vortex at the piles became active and the effect was moresignificant than the forward flow forces. The model was updated by inputting the vortex forces as a back flow force. The back

Fig. 12. Experimental and updated FE results for strain gauges located on the slab.

Fig. 13. Experimental and FE results for strain gauges located on the piles at the Q side.

1804 M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810

flow force was initially estimated within a range of 5–50% of the forward flow force and finally the value of 11% gave resultsin agreement the experimental results. The backward force was applied as a line force on the piles in opposite direction tothe flow direction. Fig. 14 shows the updated FE strain values at SG6–SG8.

The results show good compatibility between the updated FE strain values and the experimental results. The used of theback flow force helped to improve the FE model’s accuracy to model the actual operating condition of the integral bridgeunder different flow velocities. The maximum difference between FE and experimental results was 7.2%, and this differenceis probably due to the assumptions made during the construction of the model regarding the support conditions and thebridge model substructures connections.

For the deflections on the bridge slab and piles, the same updated values for the sub-grade stiffness, bridge material massand elasticity modulus, and the water level above the bridge slab corresponding to each flow velocity were used. Moreover,the back flow force was applied at the level of the deck slab and on the piles with percentage of 25% and 11% of the forwardflow force, respectively. Fig. 15 compares the deflection of the bridge slab obtained from FE and from the experiment. Theexperimental results were acquired from LVDT1, LVDT2, LVDT3 and LVDT4 located on the slab surface. The numerical mod-elling was performed based on the final scour results, which were at maximum scour after 24 h.

Fig. 14. Experimental and updated FE strain values of the piles at the Q side.

M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810 1805

The experimental results for LVDTs 2, 3, and 4 showed that, as expected, there was always an increase in the deflectionvalues in the y-direction. For LVDT1, located at the forehead of the bridge slab mid-span, the deflection showed unnoticeablechange when the flow velocity increased from 0.19 to 0.25 m/s. This could be due to an error with the equipment used or dueto the test conditions. The deflections of LVDT1 and LVDT2, which were located at the bridge slab mid-span, were the firstand second highest deflection, respectively, while LVDT4 was the lowest. Variations in the deflection values along the slabcould be due to the fact that the hydrodynamic pressure induced at the bottom of the bridge slab was different at each spe-cific location. The FE model was able to model the variation of the hydrodynamic pressure on the bridge slab. The FE resultswere similar with the experimental results, showing that there was always an increase in the deflection with the increase inthe flow velocities.

The deflection behaviour of the piles for the bridge where deflection occurred on the x-axis is also compared. Fig. 16shows the experimental and FE results of LVDT5 through 8, located on piles on the Q side.

Experimental results show conflicting behaviour in the values of deflection on the x-axis (along the floodplain cross sec-tion) corresponding with the increase in flow velocity; this could be due to the vortex at the back and the sides of the pilesand the uplift effect on the slab and the abutment, which influences the pile deflection along the x-axis (along the floodplaincross section). The deflection values shifted from negative to positive corresponding to the increase in the flow velocity.

The FE model gave compatible results with the experimental results where updating of the FE model with the materialproperties and the modelling of the back water flow effect as a percentage of the forward flow improved the accuracy of theFE results. The difference between FE and experimental results was reduced to within 10%.

The results can be summarized as follows:

– The experimental results show various behaviours of bridge substructures in terms of strains and deflections. Variation inresults between the experimental and modelling could be due to the backward, turbulent flow, effect of water and thehorseshoe vortex at the back of the bridge, the test conditions or equipment accuracy used during the test record, thevortex at the back and sides of the piles as well as the uplift force effect on the bridge slab and abutment. Moreover,the hydrodynamic pressure induced at the bottom of the bridge slab also caused the discrepancy. The results were alsodifferent at each specific location.

– The FE model was slightly different from the experimental model due to the lack of available information on the effect ofthe backward and turbulent flow of water at the back of the bridge and the vortex at the back as well as at the sides of thepiles. Since the FE model assumed ideal equipment accuracy, the results differ within less than 10%.

– Assumptions made when constructing the FE model led to different results. These assumptions include the ideal integralconnections of the bridge substructures, which were constructed using high strength adhesive material in the experi-ment. The another assumption was the fixed components of the pile support at the bottom of the flood plain, which were

Fig. 15. Experimental and FE results for LVDTs located on the slab.

1806 M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810

formed by inserting the pile end into a hole at the base of the flood plain that was filled with high strength adhesive mate-rial for the experiment.

6. Effect of flow velocities on scour with time (experimental results)

This section compares the effects of different velocities on the scour depth for 24 h of flooding. Figs. 17 and 18 illustratethe effects of scour over time for slow velocities on the Q and P sides.

The results show that in the first 100 min, there were rapid increases in scouring depths for all piles. Between 100 minand 600 min, the scour depth pattern started to subside; after 600 min, there seems to be a constant scour depth exceptfor pile Q1, which was the first pile affected by the water flow. The results also show that piles on the Q side generallyhad higher scour depths compared to the P side. Figs. 19 and 20 show the effect of medium flow velocity on the scour depthover time on the Q and P sides.

The results show that for the first 70 min, there was a rapid increase in scouring depth on both pile sides. From 70 to600 min, the scour depth seemed to be constant for the Q side. After 600 min, there was a gradual and smooth increase untilthe 24-h mark. From 70 to 500 min, the scour depth varied on the P side. After 600 min, it seemed to subside and remainconstant until the 24-h mark. The results also show that, in general, piles on the Q side had greater scour depth, whereQ1 had the highest scour of 78 mm in 24 h. Figs. 21 and 22 shows the effect of the fast velocity flow on the scour depth overtime on both the Q and P sides.

The results show that for the first 30 min, there was a rapid increase in the scour depth for all piles. For piles on the Q side,there was a rapid increase in scouring until 300 min, after which scouring rates slowed to a constant rate for the remainderof the 24 h. On the P side, there was a high rate of increase in scour depth until 500 min, after which it reached a constantrate for the remaining 24 h. Generally, piles on the Q side showed greater scour depths than piles on the P side.

From the aforementioned results, different velocities had different effects on the scour depth. The slowest velocity hadthe lowest scour depth values, where the maximum scour depth was 28 mm for pile Q1. The highest velocity had the

Fig. 16. Experimental and FE results for LVDTs located on piles on the Q side.

Fig. 17. The effect of slow velocity on the scour depth on Q side over time.

M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810 1807

maximum effect on scour depth, where the maximum scour depth was 93 mm for pile Q1. The slowest velocity rapidly in-creased the scour depth within the first 100 min and reached a slow, consistent rate from 100 min until the remainder of the24 h. The medium velocity had an intermediate effect on scour depth compared to the slowest and highest velocities. Thehighest velocity induced a rapid increase in scour depth within the first 30 min followed by an exceptionally high butsmooth increase in the scour rate until 300 min and then a nearly constant or very low increase for the remainder of the 24 h.

The results show that the highest rate of velocity induced maximum scouring at the piles, followed by the medium ratevelocity and then the slowest rate velocity, which caused only minor scouring at the piles. In addition, the highest rate veloc-ity also induced a higher rate of scouring within the first 100 min compared to the medium and slowest rate velocities. Fromthese observations, it can be concluded that the development of scour depends on the rate of the water flow velocity. This issupported by the fact that different velocities produced different intensities of downward flow due to the obstruction of thebridge model. The more intense the downward flow, the greater the ability of the flow to remove sediment at the base.

Fig. 18. The effect of slow velocity on the scour depth on P side over time.

Fig. 19. The effect of medium velocity on the scour depth on Q side over time.

Fig. 20. The effect of medium velocity on the scour depth on the P side over time.

1808 M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810

Additionally, it was also observed that clear water scour only occurred under the slowest and medium velocities. It shouldbe noted here that the local scour developed at the piles was influenced by the contraction scour at the front of the floodplain. Contraction scour caused loosened sediments to be continuously deposited into the scour hole during the 24-h exper-iment. This was observed during the slow rate velocity, where the scouring process tended to be stable after the first100 min. At this point, maximum scour depths had been achieved, as the amount of sediment supplied to the scour holeswas nearly equal to the amount of sediment transported out.

For highest rate velocity, both the contraction scour and live bed scour occurred at the same time. The live bed conditionand contraction scour at the flood plain caused a lot of sediment to be transported. Though the amount of sediment suppliedwas large compared to the slow velocity, the fast velocity was relatively stronger, and the force generated was large enoughto transport the sediment before it was deposited at the scour hole. Simultaneously, the large magnitude of downward flow

Fig. 21. The effect of fast velocity on the scour depth on the Q side over time.

Fig. 22. The effect of fast velocity on the scour depth on the P side over time.

Fig. 23. Actual scour at pile P after 24 h under slowest, medium, and highest velocities (from left to right, respectively).

M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810 1809

also transported a large amount of sediment away from the piles. Hence, this compounded phenomenon resulted in very fastscouring during the initial stage and the highest scour depth in 24 h. Fig. 23 shows the actual scour that occurred at pile Pafter 24 h under all velocities.

7. Conclusions

Based on the results of the experiments and finite element analysis, the following conclusions can be made:

1. Scour depth had a direct effect on the structural behaviour, such as strain and displacement of the bridge substructure.2. The experimental results showed different behaviour of the bridge substructures in terms of strains and deflections.

1810 M.M. Fayyadh et al. / Simulation Modelling Practice and Theory 19 (2011) 1795–1810

3. The FE model was able to model the conditions of the experimental model after updating the values of the sub-grade stiff-ness, perspex mass, perspex modulus of elasticity, and the water level above the bridge slab.

4. The modelling of the back water flow effect at the rear of the bridge and the vortex phenomenon at the piles improved theaccuracy of the FE model and hence made it more compatible with the experimental model results.

5. Assumptions made when constructing the FE model led to discrepancy between the FE model and the experimentalresults, whilst the maximum difference in values was less than 10%.

6. An increase in flow velocity increased the scour depth.7. The first 100 min of the test showed rapid increase in the scour depth for different velocity values.8. The scour depth showed a high increase for the first 500 min for all velocity values.9. The tests results from 500 min to 24 h showed a small increase in some cases and no increase in other cases.

The tests results from strong differences with respect to the experimental evidences as follows:

1. The results are better compared at the deck, because there is no sediment transport or scouring at that position (onlywater structure interaction).

2. The comparison of results showed high discrepancy at the piles because of the complex interactions of hydrodynamicwater flow, structure, and fluctuating scouring that occurred in the experiments. The software was unable to simulatethose conditions.

3. Finite element modelling is a good tool to simulate water/structure interaction for the low-flow condition.

References

[1] S. Akib, F. Othman, I. Othman, Scour Behaviour on singly and doubly row pile integral bridges, in: Proceeding of United Kingdom Malaysia EngineeringConference, University College London, 2008.

[2] S. Akin, F. Othman, I. Othman, M. Solicit, Semi-Integral Bridge Scour Prevention by using Epipremnum aureum 32nd Hydrology & Water ResourcesSymposium, Newcastle, Australia, November 30, December 3, 2009.

[3] B.A. Ashtiani, A.A. Beheshti, Experimental investigation of clear-water local scour at pile groups, Journal of Hydraulic Engineering 132-10 (2006) 1100–1104.

[4] A. Baratta, O. Corbi, Dynamic response and control of hysteretic structures, Simulation Modelling Practice and Theory 11 (2003) 371–385.[5] S.M. Bateni, D.S. Jeng, Estimation of pile group scour using adaptive neuro-fuzzy approach, Ocean Engineering 34 (2007) 1344–1354.[6] H.N.C. Breusers, A.J. Raudkivi, Hydraulic Structures Design Manual, A.A. Balkema, 1991.[7] H.N.C. Breusers, G. Nicollet, H.W. Shen, Local scour around cylindrical piers, Journal of Hydraulic Research 15-3 (1977) 211–252.[8] D. Bruno, F. Greco, P. Lonetti, P.N. Blasi, Influence of micro-cracking and contact on the effective properties of composite materials, Simulation

Modelling Practice and Theory 16 (2008) 861–884.[9] S.E. Coleman, Clearwater local scour at complex piers, Journal of Hydraulic Engineering 131-4 (2005) 330–334.

[10] J.H. Haido, B.H. Abu Bakar, A.A. Abdul-Razzak, J. Jayaprakash, Dynamic response simulation for reinforced concrete slabs, Simulation Modelling Practiceand Theory 18 (2010) 696–711.

[11] L. Huang, H. Liao, B. Wangb, Y. Li, Numerical simulation for aerodynamic derivatives of bridge deck, Simulation Modelling Practice and Theory 17(2009) 719–729.

[12] W. Huang, Q. Yang, H. Xiao, CFD modelling of scale effects on turbulence flow and scour around bridge piers, Computers and Fluids 38 (2009) 1050–1058.

[13] A.R. Kambekar, M.C. Deo, Estimation of pile group scour using neural networks, Applied Ocean Research 25 (4) (2003) 225–234.[14] T.F. Kwan, A Study of Abutment Scour. School of Engineering, the University of Auckland, Auckland, New Zealand, Report No. 451, 1988.[15] E.M. Laursen, Analysis of Relief Bridge scour, Journal of the Hydraulics Division 89 (HY3) (1963) 93–118.[16] E.M. Laursen, A. Toch, Scour around Bridge Piers and Abutments, Bulletin No. 4, Iowa Highways Research Board, Ames, Iowa, USA, 1956[17] T.L. Lee, D.S. Jeng, G.H. Zhang, J.H. Hong, Neural network modeling for estimation of scour depth around bridge piers, Journal of Hydrodynamics 19 (3)

(2007) 378–386.[18] J.P. Martin-Vide, C. Hidalgo, A. Bateman, Local scour at piled bridge foundations, Journal of Hydraulic Engineering 124 (4) (1998) 439–444.[19] B.W. Melville, A.J. Sutherland, Design method for local scour at bridge piers, Journal of Hydraulic Engineering (1988) 1210–1226.[20] B.W. Melville, S.E. Coleman, Bridge Scour, Water Resources Publication, 2000.[21] B.W. Melville, Y.M. Chiew, Time scale for local scour at bridge piers, Journal of Hydraulic Engineering, ASCE 125 (1) (1999) 59–65.[22] A.J. Raudkivi, Functional trends of scour at bridge piers, Journal of Hydraulic Engineering, ASCE 112 (1) (1986) 1–13.[23] H.W. Shen, V.R. Schneider, S.S. Karaki, Local scour around bridge piers, Journal of the Hydraulics Division, ASCE 89 (1969).[24] B.M. Sumer, K. Bundgaard, J. Fredsøe, Global and local scour at pile groups, International Journal of Offshore and Polar Engineering 15 (3) (2005) 204–

209.[25] M. Tregnaghi, A. Martion, S. Coleman, Scouring at bed sills as a response to flash floods, Journal of Hydraulic Engineering 135 (6) (2009) 466–475.[26] M. Tregnaghi, A. Marion, S. Coleman, S. Tait, Effect of flood recession on scouring at bed sills, Journal of Hydraulic Engineering 136 (4) (2010) 204–213.[27] G. Vairo, A numerical model for wind loads simulation on long-span bridges, Simulation Modelling Practice and Theory 11 (2003) 315–351.[28] P. Yannotti, S. Alampalli, H.L., New York State Department of Transportation’s Experience with Integral Abutment Bridges, International Workshop on

the Bridges with Integral Abutments, 2006, pp. 5–12.[29] H. Zhang, H. Nakagawa, K. Kawaike, Y. Baba, Experiment and simulation of turbulent flow in local scour around a spur dyke, International Journal of

Sediment Research 24 (2009) 33–45.[30] M. Zhao, L. Cheng, Z. Zang, Experimental and numerical investigation of local scour around a submerged vertical circular cylinder in steady currents,

Coastal Engineering 57 (2010) 709–721.