Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
FOUNDATIONS
FRED W. BEAUFAIT?
and
PETER W. HOADLEY~
University of Texas, Austin, TX 78712, U.S.A.
Ahstraet--The midpoint difference method is organized for solving the basic differential equation for the elastic
deformation of a beam supported on an elastic, nonlinear foundation with rigid or elastic descrete supports. An
iterative approach is used in solving the nonlinear problem; a weighted averaging scheme is employed to speed up
the convergence to a solution. The method of analysis, as programmed for a computer solution, considers the
continuous elastic, nonlinear foundation to be active only when the beam is pressing against the foundation: the
foundation is assumed to be inactive in the regions where the beam has displaced away from the foundation. Two
examples are presented to illustrate the application of the method of analysis.
and
B) INTERNAL
kTIONS
Fig. 3. Sign convention.
or
[‘%I{Y% = VA (7)
Continuity equations
Should anyone of the four variables of eqns (3) have a
where
discontinuity or constraint imposed on it at some location
Y’ = y (deflection) (3a) within the span of the beam, special continuity equations
must be written to specify the continuities that do exist
at that point and account for the discontinuity. At the
point of a discontinuity, two boundary stations and a
d2y (fictitious) segment of zero length are identified, and a set
Y’ = El-&bending moment) (3c) of four equations must be written for the fictitious seg-
Analysis of elastic beams on no~inear foundations 671
ment describing the continuity in terms of the four The nonlinear nature of the problem requires that an
variables,eqns (3), at each of the two boundary stations. iterative approach be taken. Using the bilinear curve to
For example, at the point of application of a concen- model the load-displacementresponse of the foundation,
trated point load, F, there would be a discontinuityin the Fig. 2, all foundation springs are assumed to be active
shear. The conditionsof continuitythat exist at this point and to have a stiffnessequal to the slope, & of the first
could be expressed as segmentof the load-displacementrelationshipin the first
iteration of the analysis. Once the beam has been
analyzed, the displacementof the midpointof each real
segment attached to a foundation spring is checked to
determine (a) if the spring is in tension (positive
deflection) or (b) if the displacementof the spring has
exceeded the elastic displacement A, (Fig. 2). If the
spring is in tension, the beam has tried to lift off of the
foundation: hence, the spring stiffness is set to zero for
the next iteration. If the spring displacement has
exceeded the elastic displacement A,, a new spring
stiffness is assumed.Using the new stiffnessesfor those
foundation springs that had to be modified, a second
analysisis made. Once againthe midpointdisplacements
In general, the four continuity equations that must be are checked to determine (a) if an active spring is in
written for any fictitious segmentcan be expressed as
tension, (b) if an inactive spring, i.e. a spring of zero
stiffness, is in compressionand should be a active once
[Cjl{Y”lj+ [Ql{Ydli+l = {Llj+l/* (9) again, (c) if the displacementof a spring has exceed A,
‘TLsegmentidentification ‘L._station label and (d) if the assumedspringstiffnessis compatiblewith
the actual load-displacementresponse of the foundations
for the computed displacement. New or improved
stiffnessesare assumed,were appropriate, and the beam
Writingthe midpointdifference equation for each real is reanalyzed. This process of checking the displace-
segment, the continuity equations for each fictitious ments, assumingnew spring stiffnesses,and reanalyzing
segment and the boundary equations for ends of the the structure is repeated until the actual foundation
beam, a system of 4n t 4 algebraicequations, response and the assumed spring stiffnesses are in
agreementor within some acceptable range of error.
I@{Y) = VI (IO) Because of the computationalefforts involved in the
analysis it is desirable to keep the number of iterations
(where n is’ the number of segments, both real and required to achieve a solution to a minimum.This is
fictitious) is established, replacing eqn (11, and can be controlled to a large degree by selection of each new
solved for the 4(n t 1) variablesidentifiedat the stations. springstiffness;the better the estimate of the more rapid
The reader’s attention is called to Ref.[4] for a dis- the convergence to a solution. In order to enhance the
cussion of the characteristics peculiar to this set of convergence to the solution, a weightedaverage scheme
equations and their solution. is employed.
When pa~itioning the beam into segments, both real
and fictitious, stations should be located at all points of WJMXTEDAVRRAGEs
discontinuityof any one of the four variablesof eqns (3) Before the weighted averages scheme can be utilized
as well as at points of discontinuityof the loadingor the two iterations of analysis must be carried out. For the
geometry of the structure. Segments can be of any first iteration all foundation springsare assumedto have
convenient length. The segments are identifiedby num- a springstiffnessof
bering, sequentially, from left to right; the stations are
identifiedin a similarmanner.
THENONLJNEARPROBLEM
If the load-displacementresponse of the foundation is For those springs where the displacement of the mid-
nonlinear then the governing differential equation, eqn point of the segment exceed; the elastic limit A,, a new
(1) and the approximati~ set of algebraicequations,eqn springstiffness is arbitrarily defined as
(IO),became nonlinearequations.The defiected shape of
the beam is a function of the load actingon the beam and ,,=2&t&,
the load is a nonlinearfunction of the deflectionof beam ’ 2AictAi" (12)
and the supporting foundation. A solution to this non-
linear problem is sought through a systematic as shown in Fig. 5. Whenever a spring displacement
modificationof the foundation modulus,using the famil- exceeds A, and a spring stiffness of kc has been used in
iar cosecant approach, until both compatibility and the analysis, that stiffness is modiied for the next itera-
equilibrium are satisfied. In utilizing the mid~int tion according to eqn (12). For those springs where the
difference method to analyze the beam, the supporting displacementis less than A,, the initialstiffnessk. is still
foundation can be modelled as a spring of stiffness k valid.
when in compression and zero when in tension, located Once a second iteration has been completed the
at the midpoint of the segment. The stiffness of the weightedaverages scheme can be employed to obtain a
spring is a function of the displacementof the midpoint better estimate of a springstiffnessfor the next iteration.
of the segment. Definingthe displacement of the foundation spring at
612 F. W. BEAUFAIT and P. W. HOADLEY
I k"
F.
F,
F,
Force
Force
Area
Area
•i f I
4 AU
Deflection
Deflection
Fig. S. Assumed spring stiffness where foundation first strained Fig. 6. Unbalanced forces used in determining weighing factors.
beyond elastic limit.
Now, the weight assigned to the upper limit spring
the midpoint of segment i evaluated in the second itera- stiffness is
tion as A?, a spring force E’ associated with the
assumed spring stiffness k’ can be computed according
(21)
to
E’ = ki’Ai and the weight assigned to the lower limit spring stiffness
(13)
is
and a spring force FF associated with the assumed spring
stiffness k* used in the second iteration can be AL=I-AU. (22)
established according to
Using the weighing factors, a new spring stiffness can be
E2 = k:AF. (14) defined for the next iteration as
Now the actual reactive force of the foundation cor- kl’” = A ukiU $ A l.kiL, (23)
responding to the computed displacement Ai is
Should the upper and lower limits of a spring stiffness
Es = Fe t ki” (A: - A,). (15) not bracket the actual reactive force of the foundation
for the computed displacement, A:, then a new upper or
Using the information obtained from the first and second lower limit will have to be established. Referring to Fig.
iterations, provided by eqns (II)-(H), weighing factors 7, if both the previous and the present stiffnesses yield
can be established for determining the spring stiffness to forces greater than the actual reactive force then a new
be used in the next iteration for the foundation spring of lower limit spring stiffness must be established before
segment i. For subsequent iterations the information eqns (16)-(23) can be used. This is accomplished, as
from the previous and the present iterations is used to shown in Fig. 7, by defining a new lower limit stiffness
establish the new spring stiffness for the next iteration. that underestimates the actual force by the same amount
At the conclusion of an iteration, say the pth iteration, e, that the present lower limit overestimates the actual
it is assumed that the spring stiffness used in the pre- force. If both the previous and the present spring
vious iteration, kP_‘, and the spring stiffness used in the stiffnesses yield forces that are less than the actual
present iteration, kp, define upper and lower bounds, ki” foundation force, Fig. 8, then a new upper limit stiffness
and kiL, for the stiffness, as shown in Fig. 6 and that for must be identified. This is done by defining a stiffness
the displacement of the midpoint of segment i, Ai’, these that will overestimate the actual force by the same
stiffnesses bracket the actual reactive force exerted by amount, e2 that the current upper limit underestimates
the foundation. Now an unbalariced force for the lower the actual force.
limit stiffness is defined as the The iteration process is completed when all of the
foundation springs that are compressed beyond the elas-
tic limit, Aie, have been identified, no additional springs
need to be released due to a positive displacement of the
where beam, and the difference between the new spring stiffness
and
F.l*
I = k.LA.p
II (18)
where
Deflection
F.” = k.UA?1.
I , (20) Fig. 7. Establishing appropriate lower limit.
Analysis of elastic beams on nonlinear foundations 673
20
Force (kips)
Area (fe) l6
Force
Area
4 4 I
004 01 02 03
Deflection
DeflectIon ( ft)
Fig. 8. Establishing appropriate upper limit.
Fig. IO. Examples I and 2: load-displacement relationship for
soil (I ft = 304.8 mm; I kip = 4.448 N: I kiplft’ = 47.88 kN/m’).
20 klps/ft Example I
The two span continuous beam of Fig. 9 was analyzed
by the midpoint difference for the indicated loading. The
beam has a constant cross-section with a moment of
inertia I equal to 0.6175ft“ (0.0053m4); the modulus of
elasticity E is taken as 432000ksf (20680MPa).
Fig. 9. Example I: Two span continuous beam (I ft = 0.3048 m;
In partitioning the beam, the left span is divided into
I kip/ft = 14.59 kN/m). IO segments of l.5ft (0.457m) and the right span is
divided into I7 segments with the first and last segments
1.25ft (0.381 m) in length and the other 15 segments 1.5ft
and the present spring stiffness for each active spring is (0.457 m) in length. A fictitious segment is placed at the
within a specified tolerance. interior support.
The load-displacement response of the soil is given in
ILLUSTRATIVE EXAMPLE Fig. 10. This is somewhat characteristic of a medium
In order to illustrate the proposed method of analysis, dense sand.
two examples are offered. The first example shows the Six iterations were required to solve this problem. The
analysis of a beam where a portion of the structure lifts analysis was halted when the new stiffness of each active
off of the foundation; the second one demonstrates the foundation spring was within 1% of the current spring
versatility of the method analyzing a beam structure on stiffness. The deflection curve and the variation of the
elastic supports as well as a nonlinear soil foundation. foundation pressure are given in Fig. II; the shear
3
m)R;_
W--
z
kk
2
m9
,YY.00 6. 67 13.33 20.00 26.67 33.33
I 8 3. 00
DISTANCE. X IFT)
m
cn
Fig. I I. Example I: Deflection curve and soil foundation pressure.
614 F. W. BEAUFAITand P. W. HOADLEY
20ff _i 30tt
/8
/ I i / I I
I I = 16875 tt*
of a grid
1 2 3 4 5 6 7
foundation supported on a nonlinear soil (1 ft = 0.3048 m;
lterotions
1kiplft = 14.59 kNlm; 1ft-kip = 1.356
kNm; 1kip = 4.448k&J).
Fig. 13. Example 1: Convergence of foundarion stiffness
(1 kip/ft* = 47.88 kNlm*; 1ft = 0.3048 m).
support springs represent the restraint offered by trans-
verse beams framing into the girder to be analyzed; the
diagram and the bending moment diagram for the beam three rotational elastic springs are used to model the
are given in Fig. 12. To illustrate the convergence of the rotational constraint offered by the transverse beams and
method of analysis, the variation of the stiffness of the the columns framing into this particular beam. The
foundation spring at two locations in the right span is moment of inertia of the beam is taken as I.6875 ft4
plotted in Fii. 13. Note in Fig. 11that the left span of the (0.1456m4~ and a modulus of elasticity of 432000ksf
beam has lifted off the soil foundation. The final solution (20680MPa) is used. The rotational support springs have
is listed in Table 1. a stiffness of 25400ft-klrad. (34440kNm) and the trans-
lational support springs have a stiffness of 2500 klft
Example 2 (36910kN/m). The load-displacement response of foun-
The two span continuous beam described in Fig. 14 is dation material is that given earlier in Fig. 11. For the
part of a grid foundation. The three translational elastic analysis, the left span was divided into 10 segments of
F. W. BEAUFAIT and P. W.HOADLEY 615
TableI.Analysis
of two spancontinuous
beam-Example I
I
00 8.33 16.67 25.00 33.33 41.67 50.00
DISTRNCE. X IFTI
;;
.-
Em
+-us_
LN
\
Y
u-7
m&
W--
z
LL
L
Lo
m9
,% 00 8.33 16.67 25.00 33.33 41.67 ' b.00
DISTRNCE. X [FT)
Fig.15.Example 2:Deflection
curveand soilfoundation
pressure.
676 F. W. BE4UFAIT and P. W. HOADLEY
:
! ,
‘0. 00 6. 33 16.67 25.00 33.33 41.67 50.00
DISTRNCE. X (FTI
2ft (0.61 m) length; the right span was divided into 15 special operators, or equations, are needed to handle
segments of 2 ft (0.61 m) length. difference discontinuities. The obvious disadvantage is
The beam was analyzed for the indicated loading. Six the number of equations that must be solved. The tech-
iterations were again, required to achieve a solution. The nique proposed in this paper is very versatile and can be
deflection curve and the variation of the foundation organized to solve other types of problems that can be
pressure are shown in Fig. 15; the shear diagram and the described with the nth order differential equation.
bending moment diagram are given in Fig. 16.
CONCLUSIONS REFERENCES
The midpoint difference method coupled with the use I. M. Hetenyi, Beamson Elastic Foundation, The University of
of weighted averages and the cosecant approach offers a Michigan Press, Ann Arbor, Mich (Jan. 1946).
general numerical technique for analyzing the problem of 2. Fred W. Beaufait, Numerical analysis of beams on elastic
a linear elastic beam supported on a nonlinear foun- foundations.J. Engng Mech.DIV. ASCE, 203, No. EMI 205-
209(Feb. 1977).
dation. The method is well suited for programming on 3. R. L. Kondner, Hyperbolic stress-strain response: cohesive
the digital computer; it converges rather rapidly to a soils, ,I. Soil Mechanics and Foundations Div. ASCE, 89,
solution. The method has the advantage over other 115-143 (1%3).
difference methods that might be used to solve this type 4. Fred W. Beaufait and George W. Reddien. Midpoint
of problem in that segments of varying length can be difference method for analyzing beam structures. Corn@.
used without complicating the analysis and that few Structures 8,745-751 (1977).