Fatigue life estimates for a slender orthotropic steel deck

Journal of Constructional Steel Resea le S. , U iver 6; a tail ese eld because the random variation of local stresses is sensitive to many factors, in particular to those involved in the dynamic interaction between the vehicle’s tyres, the pavement and the steel structure. This paper presents and discusses the main causes for the observed cracks and the outcome of the ultimate fatigue life estimates for typical welded and geometrical details of a slender orthotropic deck with trapezoidal cross-section longitudinal ribs. This is the deck of an existing long- span steel bridge, which has been strengthened by adding a reinforced concrete layer fixed with shear studs to the steel deck top plate. This was done to avoid continued and extensive repair work and to enhance the fatigue life expectancy of the steel deck structure. The fatigue life estimation is done with the aid of refined numerical modelling and in situ experimental strain measurements and also by taking into account all the built-in structural alterations, changes in volume of traffic and in vehicles loading which have occurred during this bridge’s 32 years of service life. c© 2007 Elsevier Ltd. All rights reserved. Keywords: Steel bridge; Orthotropic steel deck; Fatigue cracks; Vehicle–structure interaction 1. Introduction Due to its light weight, slender steel orthotropic decks having thin-walled longitudinal stiffeners of trapezoidal cross- section are widely used in the construction of long-span steel bridges and also for deck replacement in existing bridges in many parts of the world. However, depending on the geometrical characteristics and relative slenderness of their components these structures may be quite susceptible to traffic- induced fatigue cracks. Fig. 1 refer to several details of the more slender stretch of the orthotropic deck of the steel twin box-girders of the Rio- Niteroi highway bridge which, before its recent rehabilitation used to be, under stochastic traffic loading, frequently damaged by fatigue cracks in the welded joints and geometrical details. longitudinal welded connection between the deck plate and rib web; (ii) the transversal field butt-welded connections of ribs and the welded splice plates; (iii) the welded connections between the ribs and the trapezoidal shape splice plates, and between these and the floorbeam web. Many relevant works reported in the technical literature have pointed out several aspects concerning fatigue cracks in slender orthotropic decks commonly used in bridge structures. Cracks of type (i) above are reported in Refs. [1,2] while cracks at the rib to floorbeam connection might occur for different cut out details [3]. Improving these geometrical details [3,4] in order to reduce stress variations has been an important task for the design engineer. Rehabilitation of the Rio-Nitero´i bridge orthotropic deck was carried out from the second half of the year 2000 to Fatigue life estimates for a s Ronaldo C. Battistaa,∗, Miche`le a Research Institute COPPE Civil Engineering bCivil Engineering Department, Un Received 26 March 200 Abstract Fatigue cracks in several types of welded joints and geometrical de decks of existing steel bridges in many parts of the world. Some of th into service. That is why it is said that fatigue life estimation for the w Cracks were first and more frequently observed in the welded details illustrated in Fig. 1(b), (c) and (d): (i) the ∗ Corresponding address: COPPE – Engenharia Civil, C. Postal 68506, CEP 21945-970, Rio de Janeiro, RJ, Brazil. Tel.: +55 21 2562 8477; fax: +55 21 2562 8484. E-mail address: battista@coc.ufrj.br (R.C. Battista). 0143-974X/$ - see front matter c© 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.jcsr.2007.03.002 rch 64 (2008) 134–143 www.elsevier.com/locate/jcsr nder orthotropic steel deck Pfeila, Eliane M.L. Carvalhob niversidade Federal do Rio de Janeiro, Brazil sidade Federal Fluminense, Brazil ccepted 21 March 2007 s have been reported to occur in a large number of slender orthotropic cracks are initiated very early, a few years after the bridge is brought ed joints of orthotropic decks is not an easy designing task. This is so the beginning of 2001, by implementing one of the two proposed [5–7] alternative solutions in which a reinforced concrete slab was fixed to the deck steel plate with shear stud connectors. The work was carried out without interrupting the traffic of vehicles on the lanes, but the one and the neighbouring lane where repair work was in progress. ion Fig. 1. Steel orthotropic deck of Rio-Niteroi bridge: (a) bridge steel box girder with orthotropic deck; (b) detail of slender trapezoidal ribs and deck plate; (c) longitudinal section of the deck, with transverse welded joint and splice; and details of rib to floorbeam connection: (d) as built detail (1974); (e) altered detail (2001) with scallop cutout according to the original design drawings. In situ measurements taken afterwards [8] confirmed the expected reduction in local stresses in the most critical welded details. In spite of that cracks could still be observed in the as built ribs to floorbeam connections, as illustrated in Fig. 1(d). Because of this, scallop cutouts were made to reduce concentration of stresses in these regions as illustrated in detail in Fig. 1(e). A continued monitoring of the most critical and slender stretches of the original and composite orthotropic deck has been carried out in order to collect data and to better understand the observed complex random dynamic stress field variation and better estimate the consequent remaining fatigue life of the original and recently altered critical welded joints and geometrical details. Refined numerical results from parametric studies, obtained with experimentally calibrated finite element models of the subjected to heavy vehicle loading [5–7,10] and also to better estimate the remaining fatigue life of the strengthened orthotropic deck of the bridge. With regard to the vibration response spectra to the traffic loading [5] no changes in bandwidth were produced by the implemented composite deck structure, although amplitudes of displacements, accelerations and consequent stresses have been substantially reduced. 2. In situ experimental measurements Traffic on this bridge, brought into service in 1974, has raised according to a 8.7% average rate on the first 15 years, well beyond the initial 1.5% rate estimate at design stage [5]. The histograms of frequency of heavy vehicles per number of axles in two distinct periods of time of the bridge service life R.C. Battista et al. / Journal of Construct orthotropic deck, were used to better understand the static and dynamic behaviour and sensitivity of this structural component al Steel Research 64 (2008) 134–143 135 are shown in Fig. 2. They reveal the following traffic changes along the years: a decrease in frequency of buses and trucks of two axles and a clear increase for trucks with three and five axles. The classification of typical commercial vehicles in terms of number of axles and weight range is presented in Table 1. In situ experimental strain measurements [8,9] as well as on a prototype scale physical model [5,6] were of fundamental value for understanding the dynamic behaviour and for estimating the fatigue life of welded joints. Two in situ measurement campaigns were carried out on slender stretches of the deck: the first in 1997, in which the original steel deck with asphaltic pavement was subjected to both normal traffic of heavy vehicles and a controlled loading given by a weighed Fig. 3. Schematics of deck instrumentation. (a) Longitudinal and transverse strain g pavement (1997 measurements); (b) strain gauge rosettes at hotspots on the compo truck under controlled low speed on a relatively new asphaltic pavement surface (1997 measurement campaign). What can be readily observed in these figures is the sharp effect of wheels on the resultant stresses. These local stresses resulting from the interaction between pneumatic tyres, flexible asphaltic pavement and the thin-walled steel structure, particularly those related to transverse bending moments (Fig. 4(b)), are most sensitive to: • transversal location of tyres’ contact area with respect to the rib webs; and dependence on tyres’ radius, hardness and pneumatic pressures and, of course on the varying loading wheel axle. 136 R.C. Battista et al. / Journal of Constructional Steel Research 64 (2008) 134–143 Fig. 2. Histograms of frequency (%) of heavy vehicles per number of axles. Table 1 Categories of heavy vehicles, number of axles and weights Category Number of axles Description Total weight (kN) range 2 2 Buses and trucks 30–130 4 3 Buses and trucks 130–200 6 4 Trucks 200–250 7 5 Trucks 250–300 8 6 Trucks 300–400 three-axle truck [9]; the second in 2002 on the new composite deck subjected to normal traffic of heavy vehicles [8]. Fig. 3(a) shows the instrumentation schematic for the first campaign, which focused on the welded connections between ribs, webs and deck plate, also on the transverse field butt- welded joints of ribs and their welded splice plate connections (see Fig. 1(b), (c)) where cracks started to be detected as early as in 1980. Fig. 3(b) and (c) show the instrumentation schematic and a detail of it for the second measurement campaign, in which attention was drawn to the rib to floorbeam welded connections, where cracks on rib web, splice plate without scallop cutout and floorbeam web (see Fig. 1(c), (d)) started to be also detected as early as 1980. Strain measurements were made at points R3, R4 and R5 for the deck floorbeam with scallop cutouts. Typical stress versus time responses obtained from longitudinal and transverse bending strains measured on ribs (see Fig. 3(a)) are shown in Fig. 4(a) and (b) respectively, where one can note the peak amplitudes associated with the passage of each of the three wheel axles of the 200 kN weight auges at deck midspan between floorbeams of the original structure with flexible site deck’s floorbeam with scallop cutouts. ion nse Fi • roughness, weariness, flaws and overall geometrical irregu- elements and also by axially rigid elements to prevent the larities of the pavement surface, which cause dynamic vari- ations in the contact area and applied pressure, and in- duce amplification of dynamic loading and therefore of local stresses; • multi-mode vibration behaviour, with associated clustered natural frequencies, typical of these easy-exciting slender orthotropic decks. Fig. 4(c) shows the variation with time of the maximum principal stress registered (2002 measurement campaign) by rosette strain-gauge R4 (see Fig. 3(c) for location) on the composite deck under normal traffic load. A short time interval of this signal corresponding to the stress variation due to the passage of a truck is depicted in Fig. 4(d). Also noticeable are the higher vibration frequency components of this stress history related to the many stress peaks in the frequency range (12–15 Hz) of the first two vibration modes of the composite deck structure (see Table 2 and Fig. 5(d)). 3. Numerical modelling interpenetration of steel and concrete layers. This second model was also calibrated in terms of natural frequencies with the experimental data obtained from the 2002 campaign (see Table 2). Additional refinement was further provided to these two finite element models in the locations where the wheel load pressure was to be applied and where stresses were to be observed. Fig. 5(c) shows a detail of the deck model in which both the floorbeam web plate and the trapezoidal splice plates (see Fig. 1(d)) were represented by shell elements connected along the weld lines. Also seen in this figure is a superimposed drawing of other structural elements which compose the finite element models. In these models the effect of the weld chord is not considered. All analyses were performed using the program SAP2000 [11]. 4. Points of measured strains and calculated stresses Fig. 6(a), (b) show for the original deck floorbeam without cutout details of the rib to floorbeam connection where fatigue R.C. Battista et al. / Journal of Construct Fig. 4. Typical stress responses. (a), (b) Longitudinal and transverse stress respo low controlled speed; (c) maximum principal stress time history at point R4 (see of the same stress time history. Table 2 Numerical and experimental vibration modes frequencies (Hz) obtained from in-situ measurements Mode Frequencies (Hz) of the original steel deck Frequencies (Hz) of the composite deck Theoretical Experimental Theoretical Experimental 1 13.7 – 12.6 11.3± 0.4 2 16.5 17.2± 0.4 15.0 14.6± 0.4 The stress histories shown in Fig. 4(a) and (b) appeared quasi-static, with negligible dynamic effect, due to the good conditions of the pavement. However, these stresses can also be most sensitive to: Fig. 5(a) shows the finite element model of the original steel orthotropic deck with three panels spanning on the transverse al Steel Research 64 (2008) 134–143 137 s, respectively, at midspan of the deck panel due to a three axles truck passing at g. 3(c)) due to normal traffic load, (d) zoom of a short interval with a large peak floorbeams. This model was initially calibrated in terms of natural frequencies and associated vibration modes with the experimental data obtained from in situ measurements (see Table 2). Four node shell elements that combine separate membrane and plate bending behaviour were selected to model the deck components. To model the concrete layer and the stud connectors in the actual composite deck, shell elements and spatial frame elements were respectively added to the model of Fig. 5(a). The shear connectors were considered embedded in the concrete layer and therefore, restrained against bending and shear deformations. The shell elements representing the concrete layer were disposed at its midplan and connected to the shell elements representing the steel plate by the stud cracks occurred (see Fig. 1(c), (d)). Fig. 6(c) shows for the composite deck a detail of the same connection with scallop ruc Fig. 5. (a) FEM e deck model at rib to floorbeam Fig. 6. Location of points for calculated stresses. cutouts on the trapezoidal splice plate. The spots in these figures indicate either points that display higher stresses or are located at a fatigue prone welded detail. Stress variations at points P5 to P9 were used to perform fatigue life estimates. Stress peaks at points P5 to P8 were used for comparisons between theoretical results obtained for the original and composite deck models. Moreover, stress peaks at points P3 to P5 were used for comparisons between theoretical and experimental results. It should be noted that points P5 and P6 on Fig. 6(a) are located along an imaginary edge line of the scallop cutout drawn on the trapezoidal splice plate. Stresses at these points were used for comparisons between the theoretical results obtained for the original and composite decks and to evaluate the effect of the scallop cutouts. It should also be noted that points P3, P4 and P5 on Fig. 6(c) share the same locations of points R3, R4 and R5 on Fig. 3(c), where strain measurements were taken. Stress peaks and experimental results. Points P3 and R3 are located on the floorbeam web, while points P4 (R4) and P5 (R5) are located on the trapezoidal splice plate. 5. Numerical results Refined numerical results were then obtained and used to better understand the behaviour and sensitivity of the original steel deck subjected to traffic loading. The very localized effect of the wheels’ contact pressure on the transverse deformed shape of the original steel deck is clearly illustrated in Fig. 5(b). This same effect can be observed in Fig. 7(a), which depicts the influence line of the transverse bending moment at the rib web to the deck plate welded connection (point r ). It can be noticed that a transverse shift, smaller than the ribs spacing, on the contact area location of the wheel loading can raise the bending moment, and therefore the resultant transverse stress, at these point model of the slender orthotropic deck; (b) local transverse deformation of the orthotropic deck under wheel loading; (c) detail of th connection; (d) second transverse bending vibration mode. 138 R.C. Battista et al. / Journal of Const s were used for comparison between theoretical tional Steel Research 64 (2008) 134–143 from zero to its maximum absolute value (Mr in Fig. 7(a)). on ec f t varied longitudinal position of truck axle on the border lane (see Fig. 8(b)). The transverse influence line for principal stress at points P8 (see Fig. 6(a)) on the original steel deck is shown in Fig. 7(b). It can be observed that the sensitivity of these stresses to the transverse position of the wheel load is not so high as in the case of the rib web to deck plate connection (see Fig. 7(a)). The passing vehicles cause ribs to deflect and rotate at the supports, generating an out-of-plane movement of the floorbeam webs. The out-of-plane bending stresses combine with the membrane stresses so that, in this structure, the out- of-plane bending effects account for about 50% of the principal stresses which occur for the truck axle located at a distance of approximately 1000 mm from the floorbeam. This can be seen in Fig. 7(c) where the longitudinal influence line of principal stress at point P8 is shown. Table 3 presents a comparison between the theoretical static stresses at points indicated in Fig. 3(a) and Fig. 6(a), (c), obtained with the numerical models for the original steel deck and for the composite deck. The stresses are due to an 80 kN axle load for which the positions of the wheels are indicated in Fig. 8(a), (b). What can be readily seen in this table is the substantial reduction of the transverse (T20, T21, T22) and longitudinal (L8) bending stresses at midspan between floorbeams (Fig. 3(a)), thanks to the composite deck stiffness properties as compared to the original steel deck (neglecting any composite action due to the asphaltic pavement). The principal stresses at points P5 to P8 (see Fig. 6(a), (c)) also after the construction of the concrete slab of the composite deck) one can make the following comments: (i) point P8, the most critical on the original deck, was eliminated; (ii) values of maximum and/or minimum principal stresses at points P5 and P6 increased in relation to the situation without scallop cutouts; (iii) values of maximum and minimum stresses at point P7 decreased. The dynamic effects originated from the vehicle–structure interaction were evaluated with a simplified approach. Departing from a typical profile of pavement irregularities and modelling a vehicle axle as a two degrees of freedom system [12,13] one can obtain the spectral density function of the random load on a structure, which has a dominant peak at 15.1 Hz. This exciting frequency is close to the natural frequencies of both original steel and composite decks (see Table 2). Moreover, this frequency may correspond to the sequential impulsive load induced by two (or more) close rear axles of a typical heavy vehicle crossing the deck. For example, for rear axles spacing s = 1.3 m and speed v = 70 km/h one has 1/T = v/s = 15 Hz. The impulsive loads are represented in Fig. 9 by a function given by a series of unit impulses representing the action of each truck axle, where time intervals T1 and T2 are set according to the axles spacings and speed. Dynamic responses of the numerical model were obtained by applying this dynamic loading at a certain vehicle position and for speeds ranging from 50 to 100 km/h. The R.C. Battista et al. / Journal of Constructi Fig. 7. Influence lines (a) of the transverse bending moment at the rib web to d Fig. 6(a)) for varied transverse location of truck axle located on the floorbeam o exhib