You are on page 1of 5

Beams on random elastic supports

A. A. Mahmoud

Department of Engineering Mathematics and Physics, Cairo University, Cairo, Egypt

Magdy A. El Tawil

Department of Engineering Mathematics, Cairo University, Giza, Egypt

The fourth-order differential equation of a beam resting on elastic supports is highly dependent upon
the modulus of subgrade reaction, k. In general, the value of k is random; consequently, a beam
resting on random supports is represented by a random differential equation of random coefficient.
This dtfferential equation has no exact solution; accordingly, approximate solution is recommended.
The statistical moments of the solution process are expanded in terms of the stochastic scale parameter
4’and evaluated by using successive approximations.

Keywords: stochastic differential equations, stochastic beams, stochastic subgrade reactions

In this paper the fourth-order differential equation


Introduction
with a stochastic coefficient of a beam resting on ran-
Some mathematical models are influenced by the ran- dom elastic supports (Winkler’s model) is analyzed.
dom uncertainties included in the applied excitation or The stochastic uncertainty in the subgrade reaction k
in the coefficient of the represented differential equa- is scaled by a deterministic parameter 5. The estimation
tion. Stochastic analysis is of great interest for engi- of this parameter depends on the amount of the un-
neers, especially structural and geotechnical ones. certainty in the Winkler’s subgrade model. The suc-
Casciati et al.’ analyzed the three-dimensional frames cessive approximation method proved that the random
under dynamic stochastic excitation, using the sto- deflection can be obtained as a power series in the
chastic equivalent linearization. Li and Ibrahim2 in- parameter 5. The first and the second statistical mo-
vestigated the autoparametric interaction of two nor- ments of the random deflection are obtained by as-
mal modes in a 3 degrees of freedom (3-DOF) structural suming the first-order approximation.
model under a wideband random excitation. Muscolino3
evaluated the mean square response of a linear system Random deflection of beams resting on random
subjected to a multicorrelated stationary or nonsta- elastic supports
tionary process. Branstetter and Thomas4 estimated Beams resting on the ground are often modelled as
the first- and second-order statistical response mo- beams supported by elastic springs (Winkler sup-
ments for linear multi-degree of freedom dynamical ports). In general, the properties of the ground differ
systems having random loads and random structural from one point to another, which involves uncertainty
characteristics. Vanmarcke and Grigorit? discussed the in the characteristics of equivalent elastic springs. In
analysis of shear beams of random rigidity. Spanos and other words, the elastic supports are of random nature.
Ghanem discussed the solution of problems involving Consequently, the fourth-order differential equation of
material variability. The authors’ assumed cantilev- a beam resting on random elastic supports, shown in
ered beams with continuous stochastic rigidity and de- Figure I, is random and is given as
veloped the stochastic second-order differential equa-
tion with its random uncertainty in the excitation Ez d4y(x; w)
+ k(x; w)y(x; WI = q(x)
function. dx4
where E is the Young modulus of elasticity, I is the
moment of inertia of the cross section, y(x; w) is the
Address reprint requests to Dr. El Tawil at Department of Engi- random deflection, w is the random outcome which
neering Mathematics, Faculty of Engineering, Cairo University, Giza,
belongs to (Cn, 93, P) in which IR is a sample space, 93
Egypt.
is a a-algebra associated with 0, and P is a probability
Received 8 July 1991; revised 7 January 1992; accepted IS January measure, q(x) is the applied load, and k(x; w) is the
1992 random modulus of subgrade reaction. In the case of

330 Appl. Math. Modelling, 1992, Vol. 16, June 0 1992 Butterworth-Heinemann
Beams on random elastic supports: A. A. Mahmoud and M. A. El Tawil
. b.
random outcomes. This term is the solution of the de-
x
III d terministic system (i.e., 5 = 0). By using ye(x) as the
first approximation the second approximation is ob-
tained as

Figure 1. A beam on elastic supports ~Y(‘)(X. 2 w) = 4(x) - 5yoc477(x; WI (7)


The ith approximation is computed as
a simply supported beam the following boundary con-
ditions for all random outcomes are applied: LEy”‘(X;w) = q(x) - 57&V;w),(j- “(X; w)
i>o (8)
y(0) = y”(0) = 0 (24
which leads to the existence of the assumed series
y(L) = y”(L) = 0 (2b) solution. The first term in the series in equation (6) is
where L is the beam length. The modulus of subgrade obtained when solving the deterministic system and is
reaction k can be represented as given by
k = kO + l&x; W) (3) %Yo(X) = 4(x) (9)
where kO is the mean value of k or it is the subgrade When equation (6) is substituted into equation (I), the
reaction for the deterministic system. Here, 5 is a de- ith term is computed as
terministic scale parameter which represents the de-
gree of uncertainty in k, and 77(x; w) is any zero mean ~yj(X;W) = -yi_l(X;W)77(X;W) i>O (10)
random process. Substituting equation (3) into (1) yields Using the deterministic boundary conditions given in
equation (2), we apply the following conditions for each
ZYk WI = q(x) - [7/(x; w)y(x; w) (4)
term of equations (9) and (10):
where .JZis a deterministic differential operator given
as yJ0) = y;(o) = 0 (1 la)

2%$+k,,
y;(L) = y::‘(L) = 0 (llb)
(5)
The impulse response function associated with the de-
The random deflection of the beam can be expanded terministic operator 5!?is known to be*
in power series expansion as
Cc h(x) = cash (7x). sin (yx) - sinh (yx) . cos (yx)
Y(K w) = 2 C’Y;(Gw) (6) (12)
i=o Assuming uniformly applied load and solving equations
It should be noted that the first term of the power (9) and (10) together with boundary conditions (lla)
series, ye(x), is deterministic and independent of the and (11 b), the first and second terms of the series are

1
cash (yx) . cos y(L - x) + cash y(L - x) . cos (yx)

cash (yL) + cos (yL) (13)

and L

yi(x; w) = Q(CZ+ P/m;) cash (yx) . sin (yx) P(W) = PI Ih(L - dv(s; w)yo(s)ds
0
+ Q((Y- Plmo) sinh (yx) * cos (yx) L

S)T(S; w)yo(s)ds (17)


- Q/W&>
+ P2 +(L -
_f
i h(x - sh(s; w)yob) ds 0
0 in which
(14)
(Y~= y2/(Elmo). (sinh (yL) . cos (yL)
where
+ cash (yL) . sin (yL)) (18)
Q = ll(fimo) (15) a2 = l.l(2EZmg r) * (sinh (yL) * cos (yL)
L
- cash (-&). sin (yL)) (19)
a(w) = aI h(L - s)v(s; w)y,,(s)ds
I PI = y2/(EZmg r). (cash (yL) . sin (yL)
0
L - sinh (yL) . cos (-yL)) (20)

+ a2 4(L - sh(s; w)Yo(s)~~ (16)


p2 = l.l(2EZmg) * (cash (yL) . sin (yL)
J
0 + sinh (yL) . cos (yL)) (21)

Appl. Math. Modelling, 1992, Vol. 16, June 331


Beams on random elastic supports: A. A. Mahmoud and M. A. El Tawil

ma = (kolEZ)1’4 (22) The variance of the random deflection is given as


y = molti (23)
f = (2y2/mo) * (cash* (7L) * sin* (@) Vary@; w) = 5 [*j vary& w)
i=l
+ sinh* ($_). co9 (-yL)) (24)
<b(x) = 2y2 (cash (yx) . sin (yx) + 22 2 ~(R+l)C~~{yn(~;~),yj(~;~)} (27)
j=l n=l
+ sinh (yx) * cos (yx)) (25)
where the covariance term, Cov (y,, yj), is the expec-
Variance of random deflection
tation of y,, *yi. The first approximation of the variance
Since 77(x; w) is a zero mean process, then the expec- is
tation of y;(x; w), i > 0, is also zero. Consequently,
the mean function of the random deflection is Var y(x; w) = t* . Var y,(x; w) (28)

EYCG WI = yaw (26) where

Vary,(x; w) = yll(x) Var (Y+ y12(x) Varp

k; x
+ h(x - sJ.h(x - s*)*Y0(~1)~Y0(~2)
2E2Z2mg I
0

(2%

in which
X
Zn = h(x - s)yo(sMs; WI ds (30)
I
0
yll(x) = (l/~mo)~(cosh(yx)~sin(yx) + sinh(yx).cos(yx)) (31)
y,*(x) = (l/V%zo)~ (cash (7x). sin (yx) - sinh (7x). cos (yx)) (32)
2 4k2
COV(%P) = & -21--94....,
ma
+ $*(iWR3
0
- WR3) - &WIv.R,>
0
(33)

1 k;
Cov ((.u,In) = ti Ezm8 .Pti mar (2y*M. & - N. Rd (34)

ki
Cov (P, InI = ti kzrn8. ti moT Cb*M. R4 + M. Rd (35)

L
where
M = ~movldU (36) R4 = h(L - s)y;(s)h(x - s) ds (41)
0
N = Yhm2y12(L) (37)
L

R, = h2(L - s)y;(s)ds (38) Results and discussion


I
0
The above formulation is applied to the case of a simply
L supported beam of rectangular cross section resting on
R2 = +*(I, - s)y;(s)ds (39) a random elastic foundation. The modulus of subgrade
I reaction is taken as a deterministic value, ko, plus a
0
modulated white noise. The general mathematical
L
expression for the modulated white noise9 is
R3 = c#Q. - s)ya(s)h(l - s)ds (40)
0 77(x;w) = e(x) . n(x; w) (42)

332 Appl. Math. Modelling, 1992, Vol. 16, June


Beams on random elastic supports: A. A. Mahmoud and M. A. El Tawil

where e(x) is a deterministic envelope and n(x; w) is ulus of elasticity 210 t/cm2, width of 25 cm, and depth,
white noise with the following statistical properties: d, 60 cm, 80 cm, and 100 cm. The beam rests upon a
soil of deterministic portion of the modulus of subgrade
En(x; w) = 0 (43) reaction, which is made to vary from 2 kg/cm2 to 18
and kg/cm2. The load intensity is assumed to be 1 t/m.
Figures 2-4 show the expected deflection along the
Eqbl; wM2; w) = (44)
&A. 4.~2) . %I - x2)
beam length for different beam depths and soil subgrade
where S( -) is the Dirac delta function. The Romberg reactions. In general, the expected deflection is de-
technique is used to calculate the resultant determin- creased as the beam rigidity and the modulus of subgrade
istic integrals. The solution obtained above will be ap- reaction increased. Figures 5-7 show the change of
plied for simply supported beams of length 10 m, mod-
S(x)
- * - too.0

X X
-0 loo 200 300 400 500 5ao 700 800 900 moo 0 100 200 300 400 500 500 700 800 so0 ll 50

Figure 2. The expected deflection along the beam length for Figure 5. The change of the variance of the deflection along
k0 = 2.0 the beam length for k0 = 2.0

E(x)
0.121
I

.-_--

X Y
0 100 200 300 400 500 000 700 BOO so0 lG0 -0 100 200 300 400 500 500 700 800 900 moo -

Figure 3. The expected deflection along the beam length for Figure 6. The change of the variance of the deflection along
ko = 10.0 the beam length fork,, = 10.0

Figure 4. The expected deflection along the beam length for Figure 7. The change of the variance of the deflection along
ko = 18.0 the beam length for k,, = 18.0

Appl. Math. Modelling, 1992, Vol. 16, June 333


Beams on random elastic supports: A. A. Mahmoud and M. A. El Tawil

S(x) (lE-04)
or

-I>

-IL

X - x
2 lo 111

Figure 8. The relationship between the variance of the deflec- Figure IO. The relationship between the variance of the de-
tion and the deterministic portion of the modulus of subgrade flection and the deterministic portion of the modulus of subgrade
reaction for depth = 60.0 reaction for depth = 100.0

S(x) (1E -04)


7

6
Conclusions
6
The following conclusions are derived:
I. For soft soils the variance of the deflection in-
creases as the stiffness of the beam increases while
the expected value decreases.
2. For stiff soils the expected value of the deflection
together with its variance decreases as the stiffness
of the beam increases.

Figure 9. The relationship between the variance of the deflec-


tion and the deterministic portion of the modulus of subgrade
reaction for depth = 80.0 References
Casciati, Fabio, Faravelli, and Lucia. Hysteretic 3-dimensional
frames under stochastic excitation. Int. J. Struct. Mech. Mater.
the variance of the deflection along the beam length
Sci. 1989, 26, 193-213
for the above mentioned beam and soil properties. It Li, W. and Ibrahim, R. A. Principal internal resonances in 3-
can be noticed that the variance increases as the ri- DOF systems subjected to wide-band random excitation. J. Sound
gidity of the beam increases for soft soils (up to 12 Vib. 1989, 131, 305-321
kg/cm*) and decreases as the rigidity of the beam de- Muscolino, G. Stochastic analysis of linear structures subjected
to multicorrelated filtered noises. Eng. Struct. 1986, 8, 119-126
creases for stiff soils (k, 2 I2 kg/cm*). It is also noticed Branstetter, L. and Thomas, L. Dynamic response moments of
that the variance of the deflection increases along the random parametered structures with random excitation. Pro-
beam up to 60-80% of its length and decreases again ceedings of the Third Conference in Dynamic Response of Struc-
to zero at the end support. Figures 8-10 show the fures, March 31-April 2, 1986, pp. 668-675
relationship between the modulus of subgrade reaction Vanmarcke, E. and Grigoriu, M. Stochastic finite analysis of
simple beams. J. Eng. Mech. 1983, 109, 1203-1214
and the variance of the deflection at different points Spanos, P. D. and Ghanem, R. Stochastic finite element ex-
along the beam length. It is clear that the variance pansion for random media. J. Eng. Mech. 1989,115, 1035-1053
becomes constant when the modulus of subgrade re- Mahmoud, A. and El Tawil, M. On stochastic analysis of beams.
action exceeds the value of 10 kg/cm2. It is noticed Model. Simulation Control 1990, 33, 1-8
Pipes, L. and Harvill, L. Applied Mafhematicsfor Engineering
that the minimum variance is obtained for soils of mod- Physicists, 3rd ed. McGraw-Hill, New York, 1970
ulus of subgrade reaction between 4 kg/cm* and 12 Morensen, R. E. Random Signals and Systems. John Wiley,
kg/cm2. New York, 1987

334 Appl. Math. Modelling, 1992, Vol. 16, June

You might also like