rhe reserved.
Dovloded rom asslibrary og by Colorado School of Mines on O/1/18. Copyright ASCE, For pero se ony
Technical Note
Analytical and Numerical Observations on the Hetényi Solution
for Buckling of Beams on Elastic Foundations
D. V. Griffiths, FASCE"; and G. Bee, S.M.ASCE?
‘Abstract: This paper considers the minimum buckling load and mode shape ofa simply supported beam on an else foundation, Solutions awe
‘obtained by solving the eigenvalue problem delivered by a inte-element formulation and by sing the analytical sluons involving (1 tals and
(2) rounding of real numbers to integers as described by Hetényi (Hetéoy,
'M. (1546). Beams on Elastic Foundaions, Univenity of Michigan
Press, Ann Arbor, MI). The comparison shows thatthe solution by rounding can lead to overestimation ofthe buckling load close tothe transition
‘zone between mode shapes. The paper explains the reason forthe overestimation an offers simple direct algorithm that always leads tothe coreet
‘mode shape and minimam buekling lead, DOK: 10.0614ASCE)EM.1943-7889,.0000827. © 2014 American Society of Cl Engineers,
Author keywords: Beam Elastic foundation; Buckling; Finite elements; Analyticl solution,
Introduction
Daring a recent research study involving aumerical prediction of
thermal backling of pipes embedded in variable seaioor soils.
Li and Batra 2007; Bee 2013), a program was developed for
modeling buckling of a beam on a spatially random elastic foun.
dation. Code validation was performed for uniform foundations by
‘comparing solutions ofthe eigenvalue problem delivered bya finite
clement (FE) formulation, with analytical solutions described in
Hetényi (1946). The comparison highlighted an ambiguity in the
tical soludon for calculating the buekling load close to the
region in which one mode shape rarsitions othe next. The problem
lies in the ealoulation of an integer representing the made shape
‘number used in the buckling load formula It willbe shown thatthe
rounding strategy recommended in the analytical solution can lead to
an overestimation of the bocklng load. This paper investigates the
reason for this anomaly and suggests an improved algorithm for
always finding the correct mode number.
Review of FE Formulation
‘The governing differential equation for a beam of length J and.
stiffness £7, resting on a foundation of stiffness k under the setion of
‘transverse distribute load g and un axial compressive force,
siven by
ay a)
HQ +ew aq ©
"Professor, Dep. of Civ and Environmental Engineering, Colorado
School of Mines, Golden, CO 80401; and Australian Research Counel
Cente of Exeslence for Geowchnical Science and Engineering, Uni. of
Neweatle, Callaghan, NSW 2308, Auswalia (coresponding author,
E-mail: dv.grfths@ mines eda
“Graduate Student, Dep. of Civil and Environmental Engineering,
CColonido Schoo! af Mines, Gaiden, CO 80401,
‘Note This manuscript was submited on May 2013; approved on May
22, 2014; published online on July 1, 2014, Discussion peiod open unt
December 1, 2014; spare dscusions most be submited for individual
‘ipers. This technical noes pat ofthe Journal af Engineering Mechan
fes, © ASCE, ISSN 0733-939976014014(5)525 00
@asce
osor4014-1
‘Atypical configuration i shown in Fig. 1 with discretization by five
beam elements,
‘Atypical wo-node beam element of length Lis shown in Fig. 2,
in which risa spatial coordinate along the axis ofthe beam, 9 is the
transverse deflection, and @ = dy/dr i the (small) rotation
In the absence of transverse loading (q= 0) and afer dis-
‘retizaion in space see Smith and Griffiths 2004), the various terms,
‘of Eq, (1) lead tothe matrix form shown in Eo (2)
(hn) + [rin] Cw} = Mla} @
where} =n 65 92 8)7sand 3 and = (yi), are the
translation an rotation, respectively at node f= 1.2 The waiou
matrix tenn rom Eg (2) and thei counterpart fone feet
a. (1) ae given in the Append
“The beam and foundation stifines maties canbe combined
[Ba] = Ub] + (mn ®
and, after assembly and introduction of boundary conditions, the
lobal generalized eigenvalue equation (where global matrices and
Vectors are denoted by uppercase symbols) can be written after
rearrangement as
[Gul } = 5 [Kn] {W) a
Solution of Eg. (4) by vector iteration (se Griftiths and Smith 2006)
will lead at convergence to the largest eigenvalue 1/Ner and ei
genvector{W}, where N,, isthe critical lowest) buckling load ofthe
beam onan clastic foundation and {W } isthe eorespending buckled
rode shape,
Analytical Solution
‘The analytical solution forthe critical buckling load for beams on
clastic foundations with hinged (simply supported) ends is now
summarized from Chapter VII of Hetényi (1946), Its first shown
‘hat the ertical buckling Ioad ofan infinitely long beam of flexural
siffness AY on an clastic foundation of modulus kis given by
J.Eng. Mach,
J.B. Mach, 2015.18, i
i
|
Dolan roms tcelibvay or by Colorade Sehoo of Mins on 0
Fig. 1. Simply supported beam on an elastic foundation discretized
with ive elements
*
P ye i
SAA ae
1 4
Fig. 2. Typical beam element used for buckling analysis
Norse = 2VEET o
and it is further noted that because of the additional restraint, the
critical buckling load ofa simply supported beam of finite length
‘with all other conditions the same, must be greater than or equal (9
this valve (See De Angelis and Cancellara 2012)
Ney = 2VRET ©
“Thus for the ease ofa simply supported beam, Hetényi gives Eg, (7)
for the critical load N,,
n= (tefers (Ly °
wher the integer shouldbe chosen by trials so ato make the value
of amiimum(¢g.,Brushand Almroth 175). Laterinthe paper,
two examples will compare the method using tals with a direct
algostam proposed by the authors,
‘By differentiation and leing. dN /d
pression i obtained
the following ex-
®
Cleary the value of x from Eq. (8) is a real number; however, it
‘cannot be subsitued directly into Eq. (7) because { would cancel,
land the expression would simply return the infinitely long beam
Solution given in Eq, (5). As explained by Hetényi in relation to his
Fig, 118, “Taking form an integer number whichis the nearest othe
‘ale determined rom the equation above (Eq, (8)] and substituting
that number in (126) (Bq. CD], we get the vale ofthe cial loa.”
(equations in brackets refer to the current paper)
Tan also be noted that odd and even values of integer m imply,
respectively, symmetrical and antisyrmetrcal modes of buckling
where is the number of waves (or maxima and minima) inthe
buckled shape as shown in Fig. 3 for the frst thre mode shapes
siven by n= 1,2,and3.
‘Comparison of FE and Analytical Solutions
In the following it can be assumed that dimensional quantities are
provided ina consistent system of units. Fig. 4 shows thee pots for
@AscE
osoraore.2
n= 2,Antisymmetrie
3 Symmetric
Fig, 9. Fist tee mode shapes of a buckled beam on an elastic
foundation
4 simply supported beam with stiffness 7 = 100 on an elastic
oundation with sifiness £= 30. The fist cwo plots are of bear
length (D versus buckling load (W,) a8 computed by the Hetényi
‘analytical solution with rounding as explained previously, and by
FFEs with beam element lengths of L = 0.01 (Le, 200 beam elements
would be used to model a total bear length of = 2, and so on). The
tind plot is of beam length (D versus number of iterations for
‘convergence of the vector iteration method used to solve the e
genvalue problem from Ea. (4). A beam length range of 2=1= 6
was chosen in Fig. 4 because it captures the transition between the
first (n= 1) and second (n= 2) mode shape as shown in Fig. 3.
is clear from the FE solutions thatthe number of iterations
required for convergence grows rapidly asthe transition point be-
tween the first (n= 1) and second (n= 2) mode shapes i
proached. Indeed, the algorithm would fail to converge entirely if
fan analysis was aitempted atthe exact transition point.
‘The analytical and FE solutions fo the bucking load are almost
indistinguishable for most beam lengths, falling f© a minimum of
Ner 74 given by Eq, (5) forthe infinitely long case;
however, the two solutions diverge in the range 5.28<1< 5.60.
In this range, the analytical solution continues to rise to about
Ner= 190.3 at 1= 5.60, before falling suddenly to rejoin the FE
curve at N= 165.3. Itean be noted that /= 5.28 and /~ 5.60 are,
respectively, the lengths at which the FE solution and the analytical
solution transition tothe = 2 mode, The reason for the divergence,
however, i thatthe analytical solution does not ound to n= 2 until,
the real Value of n given by Eq, (7) reaches n= 1.5, which does not
‘cur until = 5.60,
‘Although not presented inthis paper, similar discontinuities were
‘observed at all transition points corresponding to higher values ofn
but the one between n= 1 and n=2 highlighted here isthe most
pronounced, It can be concluded that in all transition zones, the
analytical solution based on simple rounding will always over
estimate the tre buckling load. Simises and Hedges (2006) showed
a similar result and noted thatthe epproximate solution based on an
infinite beam onan elastic foundation becomes increasingly accurate
as the mode number increases
Adjustment to Hetényi Solution
In the following, the authors have taken the liberty of modifying
Hetényi’s original notion to avoid confusion between integers and
‘eal numbers. The symbol ms strictly an integer, and represens the
4g, Moon
4L.Bag, Mech 2015.81, Downloaded fom acelin. org by Colorado Shot of Mines on 0/1215. Copyight ASCE. For personal we ony alright reserved.
pu | nite et
Saver haratons
0 Benet 100
Ha: Hat (ut sounding)
[Tension Zone
FE Solver Iterations
Fig. 4, Beam length {versus buckling lod Noy and FE solver iteration count
‘mode number previously called in Fig. 3. The symbol mis stitly
8 real number as computed by Ea, (8).
Table 1. Summary of Real) n Values forthe Fis Three Transition Poins
‘Atay aon oe bting od mse ie samefor Menon Comes ete sor etna
any two consecutive moe wih integer valves m and m= 1, 80
from Ea.) 2 VE sata 1s
23 Bis 2s
= p 3 vig 34641 35
Ne = WEE LIE fy PEE y 1H
7 ti at ®t Genie
‘After earangement, Ea. (8) can be writen as 2 Choose an integer such that
vaio = tft ao eee ©)
which equals the value of the real number n provided by Ea, (8)
Using Fg (10), Table | presents the actual vale of (eal) nat which
the changein mode occurs. Clearly, the made change happens before
‘reaches (2m-+ 1)/2, which is where anumber would normally be
rounded up,
Although the rounded value of m is always an overestimate
‘as shown in Table 1, the error tends to zero asm increases as shown
by
Vem 5(2m-41) as mae (Lt)
Direct Algorithm to Find the Critical Load and
Mode Number
‘A direct algorithm for finding the mode number (1) and de enitcat
(owes) bucking load (N, fora simply supported beam of length !
and stiffness EY resting on an elastic foundation of stifiness kis,
oa
easce
J. Bag Mesh, 201
osoraote-s
3. Select the correct mde number m
Ue jsnsViGF0, let maj (14a)
i ViGFsnsj+1, tet m
4. Compute the buckling load from
Ne= (82)'a1+ (
Das könnte Ihnen auch gefallen
- Méthode Des Déplacements PDFDokument12 SeitenMéthode Des Déplacements PDFDruss4Noch keine Bewertungen
- Méthode Des Déplacements PDFDokument12 SeitenMéthode Des Déplacements PDFDruss4Noch keine Bewertungen
- Cours Béton - Poutre Continue - Rev 0Dokument4 SeitenCours Béton - Poutre Continue - Rev 0Druss4Noch keine Bewertungen
- Méthode de Calcul Des Structures Élastiques - Jean CourbonDokument26 SeitenMéthode de Calcul Des Structures Élastiques - Jean CourbonDruss4Noch keine Bewertungen
