Академический Документы
Профессиональный Документы
Культура Документы
M. OLSSON
In this paper the dynamic problem of a simply supported beam subjected to a constant
force moving at a constant speed is discussed. Analytical and finite element solutions to
this fundamental moving load problem are presented. The results of this paper, provided
by the author and other investigators, are intended to give a basic understanding of the
moving load problem and reference data for more general studies. Some computational
aspects are also discussed.
1. INTRODUCTION
Structures in the field of transportation, e.g., bridges, guideways, overhead cranes, cable-
ways, rails, sleepers, roadways and runways, are subjected to moving loads. In contrast
to other dynamic loads, the moving loads vary in position. This makes the moving load
problem a special topic in structural dynamics. Structures subjected to moving loads have
been analyzed ever since the first railway bridges were built early in the 19th century.
Since then the moving load problem has become more dynamic in character mainly due
to increased vehicle speed and structural flexibility. This trend also highlights the structure-
vehicle interaction phenomenon. A large number of studies of the moving load problem
is referred to in the excellent monograph by Fryba [l]. More recent developments and
results can be found in state-of-the-art reviews [2-81.
Many analytical methods have been proposed in the past to solve simple moving load
problems (cf. [l]). However, for general analyses numerical methods have to be used.
Although varying positions of the present dynamic loads need some special considerations,
the finite element method (FEM) is especially powerful due to its versatility in the spatial
discretization; see, for example, references [9, lo]. To the authors knowledge, the finite
element method was first applied to the moving load problem by Yoshida and Weaver
[ 111. It has since been used by many other investigators [ 12-221.
This paper is concerned with the fundamental moving load problem of a uniform,
simply supported Euler-Bernoulli beam subjected to a constant vertical force moving at
a constant speed. The basic purposes of the study are to discuss assumptions inherent in
the present problem and to give a basic understanding of the moving load phenomenon.
A comparison of finite element results with analytical results is also included; this indicates
discretization errors introduced. In addition, the results of this paper should serve as
reference data for more general moving load studies in which structure-vehicle interaction
is considered and where the finite element method is clearly favourable. Finally, some
computational aspects are discussed.
2. PROBLEM DEFINITION
The moving load problem of interest here is illustrated in Figure 1. The equation of
motion of the transverse beam displacement uZ= u, (x, t), for 0 < x c L and 0 =Gt < L/ v,
299
0022~460)</91/050299+09!03.00/0 @ 1991 Academic Press Limited
300 M. OLSON
Figure 1. Uniform, simply supported Euler-Bernoulli beam subjected to a contant vertical force P moving
at constant speed u. Beam length L, cross-sectional area A, moment of inertia I?, elastic modulus E and mass
density p. The beam is undamped and initially at rest.
can be written as
pA a2u,/at2+ EI, a4U,/ax4= 6(x - vt)P, (1)
where 6 is the Dirac delta function and the time t is set to zero when the moving force
enters the beam. The boundary conditions are
Many simplifying assumptions have been introduced in the problem as defined here.
Some of the structural assumptions are: initially straight beam, linear elastic material,
small structural deformations and negligible damping effects. Shear deformation and
rotatory inertia effects are neglected and this implies that the height: length ratio of the
beam is small and that the higher beam vibration modes are not excited significantly.
For the moving load or vehicle, the assumption of small contributions from the higher
vibration modes indicates that the speed of the moving vehicle must not be too high. The
vehicle length is also assumed to be short compared to the beam length so that a one-axle
vehicle model is sufficient. Furthermore, a constant force magnitude implies that the
inertia forces of the vehicle are much smaller than the dead weight of the vehicle. For a
vehicle moving along a straight path at a constant speed, these inertia effects are mainly
caused by structural deformations (structure-vehicle interaction) and structural
irregularities. Hence factors that contribute in creating vehicle inertia effects include: high
vehicle speed, flexible structure, large vehicle mass, small structural mass, stiff vehicle
suspension system and large structural irregularities. Finally, the vehicle speed is assumed
to be known in advance and is thus not dependent on structural deformations. However,
this traditional assumption is not adopted in references [19-221.
One can notice that the dynamic effects of the present problem are caused only by the
varying position of the force since its magnitude is assumed constant. It should also be
noted that in traditional structural design, as in reference [23], only quasi-statically moving
(or crawling) vehicle loads are considered. Thus all dynamic vehicle effects on structures
are represented by dynamic magnification factors specified in code formulas.
3. ANALYTICAL SOLUTION
The problem stated in the preceding section belongs to the very few moving load
problems that can be solved analytically. Different solution methods exist and most of
ON THE FUNDAMENTAL MOVING LOAD PRORLEM 301
them are referred to by Fr$ba [l]. The analytical solution can for instance be found by
introducing the separation of variables
where Y,(t) and sin (nrx/ L) are the modal displacements and modal functions (eigen-
functions) respectively. Substituting equation (4) in equation (1) then yields
i;,(t)+wf,Y,,(t) = (ZP/PAL) sin c&t, (5)
for n = 1,2,. . . , 00 and 0~ t c L/v, and where
o, = n4T4EI/pAL4, (5, = nm/ L (6a, b)
are the squared (circular) eigenfrequencies and the loading frequencies respectively; see
also references [ 1,241. It is interesting to note that the present moving load is represented
by a half sine wave for n = 1 in the modal equation (5).
With the initial conditions Y(O) = Y(O) = 0, which represent equation (3), one obtains
the solution of equation (5) as
(2P/pALw~){l/(l -p,)} (sin c&t--& sin u,,t), Pn f 1
, Ua, b)
Y,(t)= 1 (2P/pALwi)i(sin w,t - w,t cos co,&), P = 1 I
where & is the frequency ratio, defined as
Pn = %I%. (8)
Equation (7) can be compared with expressions given, for example, in reference [25].
The analytical solution to equations (l)-(3) can now be obtained by introducing equation
(7) into equation (4). In this way one obtains the series solution
a # n, (94
+ 96[
%dL/2) r4 &(sin(crnt/~)-a7rt/7cos(a77t/~))sin(a7rx/L)
I
,
a = n, (9b)
in which u,,( L/2) = PL3/48EI is the static mid-span displacement for the force P placed
at mid-span, and T = L/v is the traversing time of the moving load. Moreover, the
non-dimensional parameter (Yis defined by
(Y= np,, = m/w, L = T,/2r, (10)
where equations (6) and (8) have been utilized and where T, is the period of the lowest
vibration mode of the beam.
From equation (9) one can see that the response ratio u,(x, t)/u,(L/2) depends on
only three non-dimensional parameters: x/L, t/r and (Y. Quantities such as mid-span
dynamic magnification factors are thus functions of a alone. Consequently, the velocity
parameter a plays an important role in understanding the dynamic character of the
fundamental moving load problem. Typically, (Y= 1 corresponds to a vehicle speed of
302 M. OLSSON
u = 400 - 1500 km/h, depending on the structural flexibility (cf. equation (10)). Therefore,
(Yvalues in the range 0~ (Ys 1 are of main interest in most applications today. On the
basis of equation (9), time histories for mid-span displacement and mid-span bending
moment (MY = -ELVd2u,/dx2) are given in Figure 2 for some values of (Y in the range
0~ (YG 1. The static mid-span moment M,,(L/2) equals PL/4, in analogy with the
displacement u,,( L/2).
As the parameter (Yincreases one can clearly see that the time histories, or influence
lines, deviate more from the quasi-static ones ((Y= 0). The pattern is similar for both
displacement and moment. However, the moment histories are more irregular, and
noticeable negative mid-span moments are obtained when the moving force enters the
beam. In Figure 2 one can also note that the curves for (Y= O-125 and (Y= O-25 oscillate
about the quasi-static influence lines. For these two cases the traversing time T is as much
as four and two, respectively, times longer than the lowest vibration period T, ; thus the
important lowest vibration mode has enough time to complete four and two vibration
cycles respectively (see Figure 2 and equations (7a) and (10)). In Figure 2 one also can
see that, except for (Y= 1, the mid-span displacement and moment are zero when the
force exits from the beam (t/7 = 1). In fact, this is true when (Y is less than unity and
when l/a is an integer, as shown by equation (9a).
For (Y= 1, equation (9b) or equation (7b) has to be used. The response for (Y= 1 and
t/7 = 1 in Figure 2 is not zero but reaches its maximum at that instant. For (Y> 1 the
maximum response occurs after the moving force has passed the structure. In such
extraordinary cases the moving load problem tends to show more resemblance to an
impact dynamics problem. One may recall that the modal load for n = 1 is a half sine
wave and that the pulse ratio r/ T, is only 0.5 for a = 1.
The maximum response ratios (dynamic magnification factors) in Figure 2(a) are 1.121,
1.258, 1.705 and 1.548 for (Y=0*125, 0.25, 0.5 and 1 respectively. The corresponding
data for the bending moment, a vital design quantity, is 1.027, 1.089, 1.389 and l-273
respectively. To visualize further the dynamic character of the fundamental moving load
-0.5
1.5
2.01
0.00 0.25 050 0.75 1.00 000 0.25 OdO 0.75 ICCI
Figure 2. Time historiks for normalized mid-span displacement and mid-span moment for the problem defined
by Figure 1, 0~ t/TS 1. (I = 0, 0.125, 0.25, 0.5 and 1. Analytical solution; see equation (9). (a) Displacement
u,(L/2, t); (b) moment M,.(L/2, 1).
ON THE FUNDAMENTAL MOVING LOAD PROBLEM 303
problem, dynamic magnification factors are presented for the whole range 0.1 S LYS 1-O
in Figure 3. One can see that the main increases in the factors Dd and D,,,, defined in
Figure 3, occur only in the intervals 0.20 C (Yc 0.62 and O-20 s (Yd 0.37 respectively. The
analytical solution of equation (9) gives max ( Dd ) = 1.732 for (Y= O-619 and max (0,) =
1449 for LY= O-369. For lower values of the ratio (Y,that is (Y~0.20, the magnification
factors in Figure 3 both increase and decrease with increasing (Y.This phenomenon is
associated with the oscillations discussed in connection with Figure 2 and has been
investigated in more detail in references [24] and [26]. For higher (Yvalues, (Y> O-62 and
a > O-37 respectively, Dd and D, decrease in the present problem. Also note that the
moment factor D,,, is, in this specific problem, smaller than the corresponding displacement
factor Dd for all (Y,and that D,,, is even less than unity for certain (Yvalues.
The series solution of equation (9) converges very fast and may therefore be truncated
after a few modes. However, as is well known, the convergence of the bending moment
(M,, = -EI$*u,/~x*) is much slower. Reference [24] gives some further reference data
in terms of truncated series solutions, also for points outside mid-span (x # L/2).
Among other results, max ( Dd) = 1.743 for u = O-617 when only the fundamental vibration
mode is used. This result is close to the exact analytical solution. Furthermore, a
one-mode solution gives max (D,) = 1.434 for (Y= O-617, which differs significantly from
the exact solution due to slow convergence. Reference [24] also gives
max [u,(x, t)]/uZ,(L/2) = l-738 for x/L = 0.53 and (Y= O-625 in a five-mode solution, and
max [M,,(x, t)]/M,,(L/2) = 1.550 for x/ L = 0.636 and (Y= O-525 in a seven-mode solution.
Note, however, that this seven-mode solution is not directly based on equation (9), but
the convergence is improved by means of the Williams method. Reference [27] also gives
truncated solutions for the present problem. After correcting for the factor 96/ rr4 = 0.986,
which is excluded, the values D, = 1.704 and 1.548 for cy= 0.5 and 1 respectively are
reported for a five-mode solution. These results are in good agreement with the exact
ones above.
The fundamental moving load problem, defined in the previous section and solved
analytically in the present one, involves a paradox sometimes referred to as the Timoshenko
paradox. This paradox says that the force P leaves the beam in a state of (free) vibration
after traversing although the net work done by P is zero (the beam supports are located
- eo-
l-25- - I.25-
Figure 3. Dynamic magnification factors for mid-span displacement and mid-span moment for the problem
defined by Figure 1, a =O.l-1.0 on logarithmic scale (OS I/T s 1). Analytical solution; see equation (9).
(a) Displacement dynamic magnification factor D, = max [ u,( L/2, r)]/ uI,( L/2); (b) moment dynamic
magnification factor D,,, =max[M,(L/L r)l/M,,(L/2).
304 M. OLSSON
at the same horizontal level). As pointed out in reference [28], the paradox can be resolved
by introducing a rolling circular disc of negligible mass. In this way it is found in [28]
that a torque must be applied to the disc to maintain the prescribed constant speed u
and that the net work done by this torque exactly corresponds to the energy stored in
the freely vibrating beam. Thus in this complete description of the problem a zero torque
results in a varying speed. This speed variation, caused by structural flexibility, is not
considered in this paper. However, for low (initial) speeds the percentage variations can
be significant and for high speeds large-scale effects are important (see references [ 19-221).
1 (a)
201
000 0.25 050 0.75 I.00
f/T
Figure 4. Time histories for normalized mid-span displacement and mid-span moment for the problem defined
by Figure 1 (a = 0.25), 0~ t/r~ 1. -, Analytical solution, see equation (9); 0, finite element solution
according to reference [16], one mode and 100 time steps (a) Displacement u,( L/2, 1); (b) moment M,.(L/2, I).
Finite element solutions to the fundamental moving load problem are also provided
in references [ll] and [13]. In both studies conventional beam elements were used but
in reference [ll] a modal formulation was then used, whereas in reference [13] the
geometric co-ordinates were retained. Numerical results from references [l l] and [ 131
are given in Table 1 together with analytical results from the previous section and finite
element results by the author [16]. From Table 1 one can conclude that the errors of the
finite element solutions are all less than 1%. Note that eight DOF and 400 time steps
were used in reference [13], and that a discretization of three modal DOF and 100 time
steps was utilized by the author. Unfortunately, verification results for the dynamic
magnification factor Dm, an important factor in structural design, are not given in
references [ 1l] and [ 131.
TABLE 1
Dd RI
\
a Exact [I61 r111 iI31 Exact [I61
0,125 1.121 1.122 1.112 - 1.027 1.031
0.250 1.258 1.259 1.251 1.258 1.089 1.082
0.500 1.705 1.706 1.700 1.707 1.389 1.390
1XtOO 1.548 1.550 1.540 1.547 1.273 1.286
306 M. OLSSON
For more general moving load problems one can easily realize. the advantage of the
finite element method. Equation (11) may in fact symbolize a linear analysis of any
structure under an arbitrary number of moving forces. The complications appear when
structure-vehicle interaction is to be described. When this often important interaction is
included, the contact forces between structure and vehicle are no longer assumed to be
known in advance but depend on structure and vehicle deformations. To model the
interaction phenomenon, the system of equations of motion has to include both structure
and vehicle equations. More importantly, when (permanent) contact constraints are
imposed all system matrices generally become time dependent (see equation (11) and
references [ll-181). In addition, the matrices C and K are often non-symmefric because
there are convective derivatives in the constraint relations.
The complications mentioned above cause computational difficulties in solving moving
load problems by standard FEM packages as shown, for instance, in reference [30]. One
way to simplify the interaction analysis is to use special building blocks in terms of
structure-vehicle finite elements, as suggested by the author [15, 161. Non-linear moving
load problems also call for special considerations. The non-linearities may be due to
non-linear structure or vehicle components, but may also be due to coupling between
vehicle speed and structural as well as vehicle deformations (see references [19-221).
5. CLOSURE
A discussion of the fundamental moving load problem of a simply supported beam
subjected to a constant force moving at a constant speed has been presented. Different
finite element solutions to the problem have been compared with the analytical solution.
The assumptions inherent in the present problem imply possible generalizations to more
complex moving load problems such as problems involving structure-vehicle interaction.
The important role and some computational aspects of the finite element method in such
advanced problems have been discussed.
ACKNOWLEDGMENTS
My co-worker and friend Professor Lot Vu-Quoc is gratefully acknowledged for his
assistance and never-ending encouragement. The support and encouragement from Pro-
fessor Hans Petersson, Sven Thelandersson and Robert L. Taylor are also greatly appreci-
ated. I also wish to thank Mr Chong Yoon for proposing improvements on the manuscript.
The work was partly financed by the Swedish Council for Building Research (BFR).
REFERENCES
1. L. FR?BA 1972 Vibration of Solids and Structures under Moving Loads. Groningen: Noordhoff
International Publishing.
2. T. HUANG 1976 Shock and Vibration Digest 8(3), 61-76. Vibration of bridges.
3. E. C. TING and J. GENIN 1980 Structural Mechanics Archives S(3), 217-252. Dynamics of
bridge structures.
4. E. C. TING and M. YENER 1983 Shock and Vibration Digest 15(12), 3-9. Vehicle-structure
interactions in bridge dynamics.
5. W. KORTOM and D. N. WORMLEY 1981 Vehicle System Dynamics 10, 285-317. Dynamic
interaction between travelling vehicles and guideway systems.
6. W. KORTCJM 1984 Proceedings of the International Conference on MAGLEV Transport Now
and in the Future, Solihull, England. Vehicle response on flexible track.
7. L. FR+BA 1987 Vehicle System Dynamics 16, 129-138. Dynamic interaction of vehicles with
tracks and roads.
ON THE FUNDAMENTAL MOVING LOAD PROBLEM 307