Resonance and cancellation phenomena in two-span continuous beams and its application to railway bridges

The objective of this study is to evaluate the vibratory response of two-span continuous beams subjected to moving loads and, in particular, to investigate the maximum resonance and cancellation of resonance phenomena. The main practical interest is the evaluation of the maximum acceleration response in railway bridges, which is one of the most demanding Serviceability Limit States for traffic safety according to current regulations. Two-span continuous bridges, in their simplest version (i.e. uniform identical spans), present antisymmetric and symmetric modes with closely spaced natural frequencies, leading to a more involved dynamic behaviour than that of simply-supported bridges. First, the free vibration response of a BernoulliEuler two-span beam after the passage of a single load at constant speed is formulated analytically, and non-dimensional speeds leading to cancellation or maximum response in free vibration are obtained for each mode. Then, these conditions are equated to resonant speeds induced by equidistant load series, and span length-to-characteristic distance ratios causing cancelled out resonances, or remarkably prominent ones, are obtained. Based on the previous derivations, a methodology for detecting which could be the most aggressive trains for a particular structure based on pure geometrical considerations is discussed. Finally, the applicability of the theoretical derivations is shown through the numerical analysis of two real bridges belonging to the Swedish railway network.


Introduction
The progressive increase in operating speeds on railway lines constitutes a challenge for administrations, rolling stock manufacturers and engineers as vibration levels that are admissible for passengers, vehicles, infrastructures and surrounding structures must be guaranteed. In this regard, railway bridges have received considerable attention during the last decades. The periodic excitation caused by the axle loads crossing a bridge or a viaduct at constant speed may induce significant levels of vertical oscillations on the deck, which can lead to adverse consequences such as ballast deconsolidation, track misalignment, passenger discomfort or even wheel-rail contact loss and associated risks [1], [2], [3].
Among the longitudinal typologies of railway viaducts, both bridges with continuous decks resting on multiple supports and bridges composed of simply-supported (SS) spans coexist. The former, common 10 in countries such as Spain, Sweden or France, are structurally more efficient and able to transmit the horizontal break and acceleration forces to the ground with the collaborative action of the substructure elements. The latter, frequently found in countries like Germany or China, may be constructed in a rather systematic way, allow possible prefabrication, partial replacement of the SS decks and facilitate continuous rails. Nevertheless, the higher number of joints and supporting devices increases the maintenance costs and 15 these structures are usually appropriate only when piers have a limited height [4], [5]. Simply-supported bridges have received considerably more attention than continuous structures. This work is devoted to twospan continuous bridges, as the simultaneous contribution to the transverse vibrations of antisymmetric and symmetric modes constitutes a more complex problem and these structures may still experience important amplifications under railway traffic [6], [7]. 20 The basic phenomenon governing the level of vibrations induced in a bridge by a railway convoy is the amplitude of the free vibrations that each axle leaves on the structure after its passage, as these free vibration waves accumulate and may add in phase for certain speeds. Depending on the ratio between the travelling time of the load and the natural period of the structure, the amplitude of the free vibrations in that particular mode may be maximum or may be cancelled out, implying that the structure will remain 25 at rest under certain ideal conditions (i.e. in the absence of structural damping). This basic problem has been analysed in detail by authors such as Yang et al. [8],[9], Savin [10], Museros et al. [11] and Kumar et al. [12] for simply-and elastically-supported beams. The interest of knowing these conditions a priori is that when resonance, caused by series of loads, takes place at a maximum free vibration velocity or, on the contrary, close to a cancellation condition, the train will induce either a very prominent response or an 30 almost imperceptible one, respectively.
The problem of a continuous uniform beam with two equal spans traversed by a concentrated force moving at constant speed was first solved by Ayre et al. [13], who described the moving load as a series of pulsating forces. The early works on the problem of moving loads acting on simple structures is described in detail in the classic reference by Frýba [14]. In the 1990s, Zheng et al. [15] and Cheung et al. [16] analysed 35 the vibrations of multi-span beams subjected to moving forces and oscillators, respectively, using modified beam vibration functions as the assumed modes. The authors showed fast convergence of the method with a small number of unknowns when compared to Finite Element (FE) solutions. Yau [7] investigated the effect of the number of spans in the dynamic response of multi-span uniform beams subjected to train loads.
The author concluded that the increase in the number of spans results in the appearance of more resonant 40 peaks as the speed increases, but also in a reduction in the impact response. This is attributed to the transmission of the vibration energy to the neighbouring spans and to the higher restraining effect of the supports on the transverse displacement. Dugush and Eisenberger [17] obtained the exact solution of the moving load problem on multi-span non-uniform beams for any polynomial variation in the cross-section properties using the exact element method. Johansson et al. [18] derived a closed-form exact solution for 45 evaluating the dynamic behaviour of a general multi-span Bernoulli-Euler (BE) beam under constant moving loads, considering stepped sections and elastic boundary conditions. Publications devoted to resonance or its cancellation in two-span beams or bridges are rather scarce and recent. Yau [7] investigated the impact response of continuous beams and, considering moving loads, detected multiple resonant peaks due to the coincidence of the excitation frequency with the beam frequencies. Kwark  [20] analysed the resonant response of two-span bridges modelled as BE 55 uniform beams subjected to trains of moving oscillators. The authors focused on the vertical response in terms of displacements, and on the appearance of two critical speeds causing resonance associated to the first antisymmetric and first symmetric modes of vibration. Wang et al. [6] numerically investigated the vertical acceleration of two-span continuous bridges with long spans (40 and 45 m) and uniform cross-sections under the action of High-Speed trains, modelled as equally spaced 2-degrees-of-freedom mass-spring-damper units.

60
This is one of the few publications to focus on the acceleration response of the bridge rather than on the displacement. The authors concluded that the resonant acceleration response in bridge and vehicle may be amplified to a fairly high degree, especially for the first two resonant speeds that may fall within the range of operating speeds of High-Speed trains. Moreover, due to the presence of sub-resonant speeds of higher modes, the maximum acceleration may occur at sections other than mid-span.
In the authors' opinion, what it is presented herein is useful, novel and contributes to the knowledge on the dynamic performance of railway bridges as (i) it provides a comprehensive study on the cancellation and maximisation of resonance in two-span continuous bridges based of the free vibration response of the structures; (ii) previous publications generally address specific structures and trains and, in the opinion of the authors, the problem should be formulated non-dimensionally, in order to reach general conclusions; (iii) 70 most of the previous contributions focus on the displacement response of the bridge, generally governed by only a few modes [21]- [22], and not on the acceleration, which is a far more restrictive Serviceability Limit State for these structures according to current regulations [23]; and (iv) being able to predict velocities leading to maximum free vibration or cancellation is of practical interest as not only the most and least aggressive trains may be detected for a particular structure but, in addition, this information could also be 75 useful when planning experimental campaigns on bridges with the aim of identifying amplitude-dependent magnitudes (e.g. modal damping).
The objectives of this study are to (i) investigate analytically the problem of free vibrations in two-span continuous beams; (ii) verify whether maximum free vibration and cancellation conditions take place and, if so, to determine their value for any longitudinal bending mode and for any structure; (iii) obtain geometrical 80 ratios leading to maximum resonance and cancellation of it for symmetric and antisymmetric modes and to prove their applicability when ideal conditions are not met; and (iv) apply the former theoretical derivations to the application of two-span bridges under High-Speed traffic by proposing a methodology to detect which could be the most and least aggressive trains for a particular design speed of the line, and what kind of resonance (order and mode) is responsible for it.

85
The contents of the manuscript are organised as follows. In section 2 the free vibration response of a BE two-span continuous beam under a single moving load is formulated analytically, and maximum free vibration and cancellation non-dimensional speeds are presented and obtained for each mode. In section 3 the excitation caused by trains of equidistant loads is considered and length-to-characteristic distance ratios are derived leading to maximum or cancelled different order resonances of any mode of the beam.

90
In section 4 a methodology is proposed so as to be able to compare the effect of different trains on the maximum acceleration response of the bridge admitting a maximum operational speed for the line. Finally, two bridges from the Swedish railway network are evaluated in order to show the applicability of the former theoretical derivations. Conclusions are presented in section 5.
The derivations and conclusions presented herein are limited to the effect of the geometry of the trains 95 on the bridge response. Additional phenomena conforming the train-induced vibration problem may affect the maximum acceleration response, such as vehicle-track-bridge ([24], [25], [26]) or soil-structure interaction ([27], [28]). The influence of these effects on the maximum response of the bridge may or may not be relevant depending on the level of coupling between the subsystems. This study pretends to constitute a firm theoretical base on which additional interaction effects, as those previously mentioned which entail an 100 important level of uncertainty, may be evaluated in subsequent investigations.

Amplitude of free vibrations in undamped case
The partial differential equation governing the undamped transverse vibrations of a BE beam, neglecting shear deformation and rotatory inertia, traversed by a constant-valued load P moving at constant speed 105 (see Fig. 1(a)) is given by where w(x) is the transverse displacement of a generic section x at time t, ρA(x) is the mass per unit of length of the beam and EI(x) represents the cross-section bending stiffness. In Eq. 1, δ and H stand for Dirac Delta and Heaviside unit functions, respectively. The solution to Eq. 1 may be expressed as a linear combination of the beam normal modes of vibration φ i (x) as per For the simplest case of a two-equal-span uniform beam with total length 2L (see Fig. 1(a)), applying appropriate boundary conditions and performing a free vibration analysis [29], the analytical normal modes are obtained for antisymmetric and symmetric modes, φ a i and φ s i , respectively, and can be expressed as where bold numbers correspond to antisymmetric modes frequencies.

120
By substitution of Eqs. 3a, 3b and 2 into 1, multiplication by the n-th mode, integration along the beam length, and in virtue of the orthogonality condition of the modes, the uncoupled equation governing the n-th modal amplitude is obtained: In order to determine the amplitude of the free vibrations once the load P leaves the structure, the previous equation and its first derivative are solved by convolution for the Finally, the amplitude of the free vibrations in each mode is obtained and non-dimensionalised by the modal static displacement as per Eq. 8. By doing so, the so-called normalised amplitude of the free vibrations, R n , can be obtained for antisymmetric and symmetric modes in terms of a single non-dimensional velocity, K n . This speed parameter has also been used by previous authors [9]. In Eq. 9 the particular closed form expression for the antisymmetric modes is provided. For the symmetric case the analytical expression 130 is rather involved and it is not included for the sake of conciseness. Nonetheless, R n may be computed for any mode either analytically or by numerical evaluation of Eq. 7. Notice that the amplitude of the free vibrations does not decay, as damping is not considered in this section.
In Figures 2(a) and 2(b), the evolution of R n is presented versus the speed parameter corresponding to the fundamental mode, K 1 , for the first antisymmetric (n = 1) and the first symmetric (n = 2) modes of the two-span uniform beam, respectively. Fig. 2(c) also represents R n versus K 1 but for the first three antisymmetric modes (i.e. n = 1, 3, 5). As both R n and K n are non-dimensional, these representations and the conclusions derived hereafter are applicable to any two-equal-span uniform beam. Representing the response with respect to the same speed parameter (K 1 instead of K n ) allows direct comparison in terms of the speed V for the different modal responses. The relation between any modal speed parameter K n and the one referred to the fundamental mode can easily be deduced for antisymmetric and symmetric modes as From the analysis of Fig. 2 the following can be concluded: • Depending on the travelling speed, the amplitude of the free vibrations that a particular beam under-135 goes in a certain mode once the load leaves the structure can be maximum or cancelled out, taking into account that no damping is present in the system. The speed parameters for these cancellation and maximum free vibration conditions, K ci 1 and K mi 1 , can be obtained analytically and are pointed out in Figures 2(a)-(b) for the first antisymmetric (n = 1) and the first symmetric (n = 2) modes, respectively. Notice that index i refers to a particular cancellation or maximum free vibration event, 140 and that i = 1 corresponds to the event taking place at the highest velocity.
• The velocities cancelling out the free vibrations of the fundamental mode also cancel out the response of the remaining antisymmetric modes (see Fig. 2(c)). This does not occur among the symmetric modes or among these and the fundamental mode.
• When damping is present and for moderate levels of it, like those usually identified in railway bridges 145 [1], R n presents a similar evolution in terms of the speed parameter K n . The main differences in the damped case are that (i) amplitudes at local maxima are lower, and that (ii) the response is not completely cancelled out at cancellation speeds, although the response is remarkably low. The effect of damping will be accounted for in sections 3 and 4.
Finally, it is of interest to note that, due to the selected normalisation and to the relation between displacement and acceleration amplitudes in the undamped case, the dimensional vertical displacement and acceleration amplitudes in a particular section of the two-span beam in the free vibration phase induced by a load P is related to R n according to From Eq. 11 it is seen that the relative differences between R n values in Fig. 2 are proportional to 150 the relative differences between the acceleration amplitudes in free vibration in a certain mode and in a particular section.

Cancellation and maximum free vibration speeds
From the solution of R n = f (K n ) (Eq. 8) cancellation and maximum free vibration non-dimensional speeds may be calculated as From now on, K ci n and K mi n stand for the speed parameters leading to the i-th cancellation and i-th maximum free vibration conditions in mode n, respectively. In Table 1 the values for the first four cancellation and 155 maximum free vibration speeds are included for the first two antisymmetric and the first two symmetric modes. As should be expected, the values for the fundamental mode coincide with those of the SS beam for the second bending mode if L is the span length [11].  Table 1: Values for the first four cancellation and maximum free vibration speed parameters for the first two antisymmetric (n = 1, 3) and first two symmetric (n = 2, 4) modes.
As the main practical interest of this research is the evaluation of railway-induced vibrations in two-span bridges, it is worth mentioning that a realistic upper limit for K 1 , which will always present the highest value 160 of K n according to Eq. 10, can be estimated. Admitting an average bridge fundamental frequency in terms of the span length as per [30], considering a maximum operational speed of 500 km/h and a minimum span length of 15 m, a maximum value K 1 0.5 is obtained. Therefore, the overall first maximum corresponding to i = 1 will never be reached in a realistic situation. or is close to a cancellation condition). Ideally, a train of equidistant loads with characteristic distance d induces a j-th order resonance of the n-th mode when travelling at V r nj , as per Eq. 13, [9]. Moreover, the resonant speed may be expressed non-dimensionally, according to the speed parameter definition in Eq. 8,

Forced vibrations under equidistant loads: maximum resonance and cancellation
By equating Eq. 13 to the cancellation or maximum free vibration speed parameters included in Table   1, for any mode n and cancellation or maximum event i, L/d ratios leading to the cancellation or maximum of the j-th resonance of the n-th mode of the two-span continuous beam are obtained according to In Tables 2 and 3 these ratios are given for the first two normal modes n = 1, 2, i.e. first antisymmetric 180 and first symmetric modes, which will have the highest participation in the transverse acceleration, as will be shown later on. Ratios for higher modes could also be obtained analytically by applying Eq. 14a. These ratios are valid for any beam or bridge due to the non-dimensional nature of the formulation.
j    Table 3: L/d ratios leading to cancellation of resonance and maximum resonance of the second mode (symmetric).

Maximum displacement and acceleration numerical response: contour plots.
In the previous sections, the cancellation and maximum free vibration phenomena for two-span B-E  (a)  (see vertical dashed lines in Fig. 3(d)). The first two correspond to first resonances of the fundamental and 220 the second mode. The latter is not a resonant condition for either of the two modes. It is worth noting that as L/d increases, the amplification at resonance reduces, since shorter characteristic distance trains excite the same resonance at a lower velocity, and lower velocities imply lower amplitude local maxima of the free vibration response, as shown in section 2. Fig. 4(b) shows the same maximum acceleration curves but computed with 6 modes. It can be seen how the contribution of higher modes is not very relevant for 225 these speed ratios, as will be shown in more detail in what follows.
The actual value of the maximum response in Fig. 3 is not relevant for this study, as no upper limit for the train speed is considered. In all plots and for the first resonance order (j = 1) of the first two modes, corresponding to V /f 1 d = 1 and 1.56, L/d analytical ratios for maximum resonance and cancellation of resonance have been highlighted in thick and discontinuous horizontal segments along with the specific 230 value, which is that included in Tables 2 and 3. The analytical predictions are accurate, even though the two modes contribute simultaneously to the total response and despite the presence of structural damping.
It is also worth mentioning that these ratios predict cancelled and maximum resonances adequately both in the displacement and in the acceleration response. In order to evaluate the effect of higher modes, the same analysis is performed but including the con-235 tribution of the first six modes of vibration, N = 6, (first three antisymmetric and first three symmetric), and it is represented in Fig. 5. Again, L/d analytical ratios for maximum resonance and cancellation are accurate estimates for real values. The effect of higher modes is visible far from resonance, for moderate values of the response, especially in the acceleration plots. The displacement response is mildly affected by modes higher than n = 2, which is to be expected and consistent with previous works [21], [22]. The overall    that is, the remaining equidistant trains will also induce this resonance (at different speeds) but leading to lower acceleration amplitudes. Fig. 6(b) represents the time-history for the corresponding resonant velocity marked with a discontinuous vertical trace in Fig. 6(a). Figs. 6(c) and 6(d) represent the acceleration associated to a train with a characteristic distance d such that when inducing a first resonance of the second symmetric mode, this also takes place at the most prominent maximum (i = 1). Again, no other train will   From the previous analysis it can be concluded that when resonance is induced on a two-span continuous beam or bridge by a train of equidistant loads, its amplification will depend on the level of free vibrations 265 associated with the particular velocity. Resonances of either the first (antisymmetric) or second (symmetric) mode are prone to be responsible for the overall maximum vertical acceleration of the bridge. The same train will require a higher speed in order to induce the same resonance of the second mode (when compared to that of the fundamental), with a higher level of free vibrations left by each axle load. Therefore, the second mode can be the one responsible for the maximum response as long as the train speed is sufficiently 270 high. In other words, the mode causing the maximum overall response of the bridge will depend on the maximum train speed.
Depending on the ratio between the length and the train characteristic distance, the response at resonance may be rather prominent or almost imperceptible. Furthermore, the analytical predictions of these L/d ratios leading to maximum resonance or cancellation of it, which were obtained in closed form in the absence of 275 damping and admitting separate modal contributions, show themselves to be good estimates of the real values. This is due to the moderate damping values in railway bridges and also to the fact that, at resonance, the contribution of modes other than the one undergoing resonance is very limited.

Case studies
In the previous section, the conditions for maximum resonance and its cancellation have been analysed upgraded is a topic of major interest for the Swedish railway administration [32], [33].
The structures under study are single-track and, in a first approach, the contribution of modes other than

Case 1. Bridge over River Lögde in Västerbotten, Sweden
The first structure under study is a bridge crossing the River Lögde on the Bothnia railway line, between the cities ofÖrnsköldsvik and Gimonäs. It is a continuous bridge with two 43 m identical spans and a uniform steel-concrete composite deck, as shown in Fig. 7. The deck accommodates a single ballasted track.

300
The main properties of the beam model for this bridge are those also used in [33], and are summarised in Table 8. A modal damping ratio of 0.5% is admitted as recommended in [30] for composite bridges of the aforementioned span length. The first two bending frequencies of the bridge are 2.34 and 3.66 Hz. Table 6 presents the critical speeds for the ten HSLM-A trains' characteristic distances for the lowest resonance order attainable given the design 305 speed, and for the first two modes, which will be the ones that contribute most to the acceleration response, as shown later on. From now on, d k stands for the characteristic distance of the k-th train. The dimensional resonant speeds are computed applying Eq. 13, and the non-dimensional ones refer to the fundamental mode for both resonant velocities of the first and the second mode, for convenience. All the trains induce first  resonance of the fundamental mode under 300 km/h. Nonetheless, only the first five trains, with smaller 310 characteristic distances, are able to do so in the case of the second mode, due to its higher frequency. The lowest resonance order attainable for this second mode for trains A6 to A10 is then second order (j = 2).
Tr  In order to compare the level of free vibrations associated to each train at the resonant speeds, the values of K r 1j are superimposed to the normalised amplitude of the free vibrations for the first two modes, R 1 and R 2 in Fig. 8. The vertical black lines stand for the resonant speeds of the fundamental mode and 315 the grey vertical ones to those of the second mode. The intersection of each vertical line with either R 1 (K 1 ) or R 2 (K 1 ) (only intersections of traces of the same colour should be considered) provides an estimation of the level of acceleration experienced by the bridge due to the accumulation of free vibrations in a certain mode at resonance. Moreover, as per Eq. 11, the amplitude of either R 1 or R 2 is multiplied by the factor F P k = P k /P 1 in order to account for the different axle load modulus of the HSLM-A trains, P k being the 320 axle load of the k-th train and P 1 = 170kN, which is the minimum value. This corrected product is shown with a red circle that has a black border in the case of resonances of the first antisymmetric mode and a grey border in the case of resonances of the second mode. Admitting that the number of loads is sufficient and that the resonance state has reached a constant amplitude due to the presence of damping, this may be used to compare the relative amplitudes of the acceleration at resonance induced by different trains on the 325 first two modes of vibration. This is of course an estimation that only takes into account the geometry of the compositions (i.e. the lengths of the passengers' coaches) and admits similar modal damping ratios for both modes, but it allows a preliminary prediction of which train will induce the most detrimental resonance and which mode will be the one undergoing it, taking into consideration all cancellation and maximum free vibration situations. In Table 6 the values of R n F P k and R n F P k /ω 2 n are included as well for each train and 330 resonance speed. Notice that the first is proportional to the amplitude of the accelerations in free vibration and the latter to the amplitude of the displacements as per Eq. 11. Also, the overall maximum value for each of these ratios is highlighted in bold. According to this, the maximum displacement could occur when train A8 induces first resonance of the fundamental mode (j = 1, n = 1), while the maximum acceleration may take place at first resonance of the second mode induced by train A4. This train should be one of 335 the most aggressive trains for the particular structure and speed limit according to the Serviceability Limit State for traffic safety. The overall maximum acceleration reaches 6.01 m/s 2 for N = 2, exceeding the limit for ballasted tracks according to standards [23]. Therefore, this bridge may need to be improved in order to allow increased train speeds. These results are consistent with those presented by Andersson [33]. As predicted, for the admitted design velocity the maximum response in terms of accelerations is due to a first resonance of the 350 second mode (V /f 1 d = 1.56), and it is induced by the HSLM-A4 train (red trace). As per the displacement, train A8 (green trace), together with A7 and A9, lead to the maximum displacement at first resonance of the fundamental mode. Notice in Table 6 that the three trains (A7, A8 and A9) present a very similar value of R 1 F P k /ω 2 1 . In Fig. 9 it can also be observed that train A2 does not induce first resonance of the symmetric mode. For this train L/d k = 2.26, very close to the theoretical value 2.266 for cancellation of the 355 second mode first resonance (see Table 3). Finally, it should be noted that the effect of modes higher than the second one is very low, especially at resonance. In the displacement response, the difference is almost imperceptible.

Case 2. Förslöv bridge in Skåne, Sweden
As a second example, the case of a pre-stressed concrete railway bridge from the West Coast line located 360 between the cities of Gothenburg and Copenhagen is presented. A modified version of the real structure is analysed, with two identical spans of 23.5 m and a uniform cross-section with the properties listed in Table   8. The bridge is composed of two structurally independent single-track decks as shown in Fig. 10. A modal damping ratio of 1% is assigned to each mode as per [30]. In this case study the natural frequencies are higher than in the previous one. For this reason, the critical velocities leading to first resonance of the first two modes exceed the maximum design speed of 300 km/h assumed for all the HSLM-A trains. Again, in Table 7 the highest attainable resonant velocities for the first and second modes have been included, along with the ratios R n F P k and R n F P k /ω 2 n , proportional to the 370 acceleration and displacement amplitudes in free vibration, respectively. The highest values for these two ratios are highlighted in bold. In this case when train A10 induces a second resonance of the fundamental mode, for that particular speed the free vibration amplitudes both for the displacements and the accelerations are maximum. Therefore this train could be one of the most aggressive.
In Fig. 11 the highest attainable non-dimensional resonant speeds for each train are represented with 375 vertical solid traces for modes n = 1 (black) and n = 2 (grey). Again, the non-dimensional amplitude of the free vibrations, R n for each train in each mode, is marked with a circle after applying the correcting factor F P k . In this case, all the trains in the HSLM model are capable of inducing a second-order resonance of the first antisymmetric mode, but only the first four have a sufficiently low characteristic distance to induce second-order resonance of the first symmetric mode below 300 km/h. It can be verified graphically that train 380 HSLM-A10 is the one leading to a highest value of R n F P k , in particular for the first mode (n = 1), as its associated second resonance speed coincides with the third local maximum of the free vibrations for n = 1.
In what follows, the response of the bridge is obtained numerically under the ten HSLM-A trains.     Table 2) and also consistent with what is shown in Table 7 and Fig. 11. This train is also the one responsible for the maximum displacement, which is also caused by the second resonance of the fundamental mode. Only a few trains are able to induce second resonance of the second mode (i.e.

400
V /f 1 d = 0.78) and the response in terms of accelerations is much lower than that induced by the most aggressive train in the first mode. In Fig. 12 it can also be observed that train A2 does not induce second resonance of the fundamental mode. For this particular train and bridge L/d k = 1.24, very close to the n = 1  Table 2). It is important to state, again, that the maximum response is mainly governed by the first two modes of vibration of the two-span 405 beam, and that the contribution of higher modes in the acceleration response at resonance is negligible.
Finally, in order to summarize the steps applied in this section to propose the particular train leading to the maximum displacement and acceleration of the bridges at resonance a flow chart is included in Fig. 13.  Eq. (8) Figure 13: Steps applied to identify trains leading to the maximum displacement and acceleration at resonance. theoretical derivations are applied to two case studies of two bridges from the Swedish railway network and 425 their applicability is shown.

Conclusions
The main conclusions derived from the research conducted are: • When a load moving at constant speed travels on a two-span continuous beam, the level of free vibrations left by the load in a certain mode may be maximum or negligible, depending on the ratio between the load velocity and the beam frequency. These conditions for maximum free vibration and 430 cancellation can be obtained analytically in a non-dimensional format.
• Linear velocities that cancel out the free vibrations in the first antisymmetric mode also cancel out the response of the remaining antisymmetric modes. This does not occur among the symmetric modes or among these and the fundamental one.
• When resonance is induced on a two-span continuous beam or bridge by a train of equidistant loads, 435 its amplification will depend on the level of the free vibrations associated to the particular velocity.
Moreover, depending on the ratio between the length and the train characteristic distance, the response at resonance may be rather prominent or almost imperceptible.
• The analytical predictions of the L/d ratios leading to maximum resonance or its cancellation, obtained in the absence of damping and admitting separate modal contributions, show themselves to be excellent 440 estimates of the real values when several modal contributions and modal damping are considered.
• The maximum acceleration response in a two-identical-span railway bridge is mostly governed by the first antisymmetric and first symmetric modes. If a bridge undergoes resonance of the first two modes, the one leading to the maximum acceleration will depend on the maximum design speed. The non-dimensional free vibration amplitudes at the actual resonant speeds may be used to estimate 445 the particular train, resonance order and mode number leading to the overall maximum acceleration response in the structure.
The previous conclusions constitute a theoretical basis limited to the effect of the train axles distribution.
Additional phenomena may modify the maximum acceleration response, such as the effects of vehicle-trackbridge or soil-structure interaction, which are beyond the scope of this study.   Table 8: HSLM-A train axle loads (P k ), coach lengths (d k ) and bogie axle spacings (b).