BEAMS ON A TWO-DIMENSIONAL SUBJECTED TO LOADS THAT
PASTERNAK BASE CAUSE LIFT-OFF
ARSOLD D. KERR1 and DOUGLAS 1%‘. COFFINS +Departmcnt
.\ktract-Fir’;t.
of Civil Engineering and ZDepartment of hlcchamcal University of Dclawarc. Nw;Lrk. DE IV716 U.S.A.
Engineering.
1~0 clowl~ rclatcd prohlcms. shoun in Fig. I. published by Chcrnigovskuya
In Ivslrdwun~~uperDm~nrik~~ Soord~mii
i Rawhem
(1961.
Konsrrukr.sii
ncr Cprugom O.mownii (Edited Moscow) and Ting (1973. J. Frunklin Insf. 296(Z).
by B. G. Korenev. pp. Il3-IJI. Gosstroiizdat. 77-89) are discussed. It is shown that neither of their formulations is correct. The aim of this paper is to show how to correctly formulntc and solve problems of this type. Utilizing the variational appro~lch for \;trl,lblc matching points dcribcd by Kerr ( 1976. Irrr. J. S~liclr Sw~c.~wc~.v12. I I I ). ;L formulation for the problem ;1naly7cd hy Tine is prcscnlcd. that is mcchanlcally rcwonahle and mathematically ucll posed. The ;unalytwul solutwn ohtaincd is cvaluarcd numerically and then compared with rclalcd test results by Durclli (‘I trl. (IVW. ./. Sfnrcr. Dr. :tS<‘E 95. 1713-1725). This paper concludes with ;I discuGm of the results ohtaincd.
The analytical aspects of unhondcti contact problems for continuously
supported structures
have hccn discussed by Kerr
( 1976. 1979). when the b;~sc rcsponsc is rcprcscntcd by ;I I’astcrn;lk ti~undation model. For prohlcms of this type the contact region is not known (I /~iori. C’onscqucntly. the matching conditions at the point 01’ separation must include ;III adtlition;il cqu;Ition li)r the clctcrniination ot’ this unknown. As an cuniplc, Kerr ( 1976) considcrcd ;I linitc beam resting on ;I two-dimensional Pastcrnak foundation and subjcctcd to vsrtical loads such that lil’t-oil’ 01’ the beam is possible, ;IS shown in Fig. I. Ikvcloping variational c;~lculus for varinblc matching points. hc sho\vcd that this problem was incorrectly formulated by Chcrnigovskaya
the GISC when q(.~) = const and i’ = 0, she heuristic;illy conditions at the separation point. .V = I: -E/w”(l)
= --q(L-lyp
- Eh”‘(I)
= q( L - 1)
h(l)
- Gw”(l)
prcscribcd
1
= 0
(1961). Solving
the following
matching
(1)
whcrc 1tj.v) is the vertical dctlcction. 11.’= d~/d.\-, G is the paramctcr of the shear layer. k is the paramctcr of the spring Inycr. q is the uniform
load. If is the length ofcontact of base
and beam and 2L is the Icngth of the beam. For the GISC of ;I platr strip bending with II’ = W(S). El is rcplaccd by D = E/I’/[ I?( I -IO’)]. For the two-dimensional
Pastcrnak
foundation
in cylindrical
model the contact prcssurc is (Kerr,
1964) p(s)
= h(.r)
- Gd’(s).
(3
Thus. the third condition in (I) prcscribcs that the prcssurc is zero at the point of separation. This is not correct. The proper condition. which results from the vari;ltional formulation, is that at s = I the slope of the shear layer is continuous. Thcrcforc. the contact pressure at the separation point may have :I non-zero value. Kerr (1979) pointed out that because of the reduced order of the differential equation that governs the Pasternak foundation response (as compared to the elastic continuum) only one of the two anticipated conditions 413
A. D. KERR and D. W. COFFIN
Fig.
I. Bum problem under consideration.
may be satislicd. For the problem under consideration hc rctaincd.
and not the 0x0-prcssurc
the tangency
condition
is the one to
listed in (I).
condition
A similar prohlcm \V;IS also incorrectly formulated Iatcr by Ting (1972). t Ic solvctl the pmhlcm shown in Fig.
I for the cast
of q(.v) = 0, by prescribing
Ihc following conditions
aL .Y = I:
,I$/) = 0 w”(l) = 0 w”‘(I) = 0 This
formtkltion
incorrectly
stipulates
Whcrcas this is true for the Winklcr model. As ;I wnsqtluu
of the first
1 .
(3)
that the dcllcction at the separation point is xro.
found:ltion,
it does not apply to the Pastcrnak base
two conditions
in (3). Ting’s
formulation
implies ;I
zero contact prcssurc at the scpatxtion point .Y = 1. This
is not correct for fhc reason given
in discussing the Chcrnigovskaya
major error in Ting’s
is his omission of the dilfcrcntial In rhc following The closed-form
formulation.
the correct formulation
solulion
Another
formulaCon
ccluation for the base layer beyond the point of separation. of Ting’s
problem is skltcd and then solved.
obtained is then numerically
eval~~;~tu.J and fhc rcsulls arc com-
pared bvith those of the photoclnstic results rcportcd by Durclli
I-OKMULATION
1’1u/. (1969).
OF PROBLEM
The problem IO bc anulyzcd is shown in Fig. I with q(x) = 0. It consists of an elastic bc;lm resting on ;I two-dimcnsiorxll Pastcrnak foundation that is subjcctcd at the ccntcr to ;I vertical conccntratcd load. I’. According to the derivation by Kerr (1976). utilizing symwstry. assuming the validity of beam bending theory, and denoting Hi, as the vertical dckction
of the barn axis in 0 c .Y < 1. Nap
I < .Y < L. and II.,
as the vertical
and G = const. the following
dcllcction
as the vertical dcflcction of the beam mis of the shear layer in I < s c
z,
for EI
in
= const
difTcrcntial equations rcsull:
EIw’~ -Gw'; +hc~, = 0
my =o
0 < s < I
I
Gw:'-X-w, = 0
I < .r <
(3) 00
The corresponding boundary and matching conditions are :
P
II
-J.
5
416
A. D. KERRand 0. W.
d-1 j = 0; .d7 =
.-I
(,
=
COFFIN
0
-.- --,‘---...Xf(p-+ri$C/’
(1’) !pcos(#jt)sinh(k.l)--ssin(,~(fcosh(k.l)t
‘_I, = __._--~;p--_‘Ef(p- f/P)‘JI
i&p
cos (ply cash (elf-(x2-p“)
sin ($1 sinh (~2)
H.hCK
The corresponding the prcssurc distribution
csprcssions
Ibr the bcncling nwmcnts.
-(K’-p2)sin
and
bcani shearing Ibrccs and
in the contact rcgiori arc :
(p.u)sinh (KY)-7~pcos
(p.~)cosh (KY)
(lb)
Beamson tuo-dlmensionai p(x)
Pasternak base
417
= kw,(.Y)-GG,v;(.v)
PG
=-
1 -
z, sinh’ (~f)-_=~ sin’ (y/)-i-:! (~~~.)I
i + ([
4,‘liEr
+
sin’ (p/l+
i:,
-II sinh’ (K/)+ z__
Ih-pp
h.z+-;T
-“$
2
sin (pr) sinh (X.X)
1
cos (p-u) cash (KX)
sin (p.r) cash (KX)-
:
)
cos (us) sinh (KY)
.
(17)
where
Next. the case G > Z-is
analyzed. The solutions of the three differential equations
in (4) are
whcrc
The integration constants 11,. . , . . II,,, dctcnuincd are
B,
=
__~_._
_.
I’
____
-
2E//l’(a- -/F)(J B s=O;
sinh (II) sinh (/l/)-a’] (21)
= ,E,,~r~f, .__iT__
&
P = _ _,-,_ .i_.._ ({I’cosh XIp-I-r#.J
B ,” =o.
cash (\U)-r/l
B,,=O
B,
B,
where
[/I? cash Irl)
[/I sinh (zl)-
z sinh (/U)j
(I/)-r’cosh
cash (xl)-z’cosh
A. D. KERR and D. W. COFFIN
418
# = flcosh (xf)sinh
and the condition for the determination
r/l[x sinh (fll)-p
The corresponding distribution
w\-) =
(xl)
cash (/U)--/I’
cash (xl)] = 0.
(23)
bending moments. shearin g forces in the beam and contact pressure
are
-P ‘($
z sinh (xx)-Psinh
_p’,
I
+[/I’
+
(/Iv)
([I/)--r/i
!~(I[b’cosh (d)cosh
9
L{[(p/r)’ -
sinh (xl) sinh (/il)-/1’]cosh
cash (xl) cash (/I!)-
l][r’
r/l sinh (xl) sinh (PI)--r”]
(/I!)-$sinh
+/l[/~2cosh (rl)cosh
4
cash (/#)sinh
ofl is
sinh (zl)]+/l[r’
+ - ([r’ cash (rl)cosh #
f
f/Q)--z
(r/)sinh
(/I/)--sc/isinh
cash (xl) cash (/l/)--r/j
([M--/l’]
(r/)sinh
(xv)
cash ([Lv))
(73)
sinh (xv)
(/il)-u~Jsillh(ll.r)j
sinh (zl) sinh (/U)-P’]
(25)
cash (xv)
(xl) sinh (li()-5’jcos11(I~x))
.
(26)
I
This completes the solution to the prohlcm under consideration.
The case G = I,/‘kK/
is not of interest for this study. Nest. the solution obtained is compnrcd with cspcrimcntal results.
COMI’ARhTlVE
STt’D-t
The only results located related to the problem under consideration are those of photoelastic tests conducted by Durelli c*f a/. (1969), which were used by Ting (1973). Durelli PI (11.tested beams ofdifTerent lengths resting on an elastic foundation and subjected
Bums
Fig. 2. Test
on tao-dimensional
set-up b) Durclli CI ul. (IY69)
hard transparent plastic (CR-39)
ruhhcr (HYSDL
load.
(I
419
base
in = 2.54 cm)
7 It consisted of a beam set-up is shown in Fig. _.
to various loadings. Their experimental madcofa
Pwtemak
resting on ;I slabofsoft
transparent polyurcthanc
44X5) of the same width as the beam. The beam was sub.jccted to ;I vertical
f’ = I40 N (3 I .J6 lb). placed
in the center.
The elastic propcrtics for thcsc materials are
CR-39
r1 = 2.22 x IO” N cm
’ (2.X x IO” psi)
I’ = 0.4’ IIYSOL-44X5
E=
W.ONcm
‘(525psi)
(‘7)
\’ = 0.47
Lvhcrc E is Young’s modulus and v is Poisson’s ratio. The bending moments M(X) were calculated directly from the rccordcd fringe numbers. Then, noting that
the shearing forces in the beam I’(.r) and contact prcssurc distribution by successive numerical diKcrenti;ltion Durelli
of M(s)
Ed trl. prcssntcd the bending moment.
tributions in the non-dimcnsionnl
normalized
ill*(r)
beam shear and contact prcssurc dis-
form
= 4i.M(.r),‘P
c’*(s) = 2C’(s),‘P /I*
p(s) wcrc clctcrmined
with respect to s.
= Zp(.r),!PL
(28) 1
where i. = ,‘/ki4EI. They determined the k value needed by using the expression derived by Biot (1937). who matched the results for an infinite beam attached to a two-dimensional elastic continuum. According to their calculations. k = 27.58 N cm ’ (40 lb in _‘) and i. = 0. IO83 cm ’ (0.275 in ‘). This non-dimcnsionalization
was found necessary by Durclli EI al. since they compared
the photoclastic results to a Winkler
model. Because i. is a parameter that appears in the
A. D. KERR and D. W. CLIFFIS
IO
5 x
tiiJ
15
b-4
J
---_ I
15
x b-4
Winklcr foundarion analysis and not in the one which utilizes ;I Pastcrnak hasc. as considered in the prcscnt paper. the non-dimcnsionolized graphs wcrc converted back to their original form. The results are shown in Figs 3, 4 and 5, as dashed lines. In a subsequent discussion Pandit (1970) pointed out that the contact prcssurc distribution, as determined by Dureili PI nl. (shown in Fig. 5). contains the inaccuracies inhcrcnt in the process of double dilferentiation
of experimental results. and that this is
&ams on two-dimensional
Pasternak
421
base
most likel) to be the cause of the sharp increase of the contact pressure distribution the load. This
possibility
was acknowledged and accepted by Durelli
LVhen considering the contact pressure distribution vertical equilibrium
under
et uf. (1970).
in Fig. 5 it should be noted that
must be satisfied. Thus,
(29)
must hold true. IVhereas in the test P 2 = 70 N (31.46 lb). the integration of the area enclosed by the experimental contact pressure results in 92 N ; this is 3 I Y6 larger than P 2. Another
problem ivith
the experimental
pressure distribution
is that the point of
separation (beyond which the contact pressures are zero) corresponds beam. IS.24 cm (6”). This
to the end of the
contradicts their statement that when subjected to P = 140 N.
the beam separated from the foundation
at both ends. Furthermore.
moment diagram shown in Fig. 3. the point of lift-off from the center of the beam. Therefore.
according to the test
appears to be at .Y = I I.7 cm (4.5”)
the contact pressure distribution
calculated from
the bending moments recorded appears to be inaccurate near the load and should not exist past s = I I .2 cm.
In order to compnrc the analytic results based on the Pasternak prcscntcd previously
with the photoclastic
results. the I’astcrnak
must hc dctcrmincd tirst. The methods for dctcrmining rcccntly discussed by Kerr
(1085). The criterion
f~~undation model
foundation
ths foundation
parameters
paramctcrs wcrc
suggcstcd for the determination
of thcsc
paramctcrs is that the analytical results agree as closely IIS possihlc with the actual situation. for the range of loads under consideration.
It was utilized by Kncifati
(198.5).
‘I‘hcrcli~rc. in the follo\ving the two I’astcrnak paramctcrs arc dctcrmincd by collocating the test results with the corresponding points to hc usctl is t’qt1:11 ti\o thta
analytical cxprsssions.
The number of collocation
to the number of unknown foundation
paramctcrs. Thcrcfore.
points arc ncctlctl. f:or the problem unclcr consideration
the separation length I is
obtained from the test. fsccause of the small size of the test beam, it is reasonable to assume th:lt the clrect of the weight ol’thc beam is ncgligiblc on the test response. Thisjustilies the USCofthc presented analysts in the following compar;ltivc study. Since the foundation paramctcrs arc not known (I priori. the analytical results for both cases G < 2,/1?Ei conhirlcrcd.
and G > 2,/kEI
From the test curves in Figs 3,4 and 5. the moment distribution it ~V;ISobtained directly from the test data. Thercforc,
must be
is most accurate. since
it will be used for collocation purposes.
From this curve two data points arc chosen. They arc : ilt
.v, = Ocm
M, = 272 N cm (24.1 Ibin)
at
.vl = II.2cm(4.4in)
iz/? = Olbin
(30)
Substitution ofthssc values into the moment expressions given in (15) and (24). noting that according to f:icv C’ 3 the separation distance I = s1 = I I.2 cm. results in two non-linear equations for the determination of k and G for each of the two casts G 5 2Jkl. Using the fMSL
routine NEQNF
the foundation
and a Fortran
program for the problem under consideration
parameters are found to satisfy G > 2&c. X-= 9.65Ncm~‘(l4psi).
They are :
G = 1864.5N(4IC)Ib).
(31)
It is of interc>J_to note that this solution program when used on the two non-lincarequations for G > 2,/kEI did not converge. It is also noteworthy that I = .v? = I I.2 cm obtained from the test curkc in Fig. 3 and the above two parameters k and G do satisfy eqn (23); the analytical condition for the determination of 1.
A. 0.
P__
KERR and D. W. COFFIN
The bending moments. shear forces and contact pressure distribution cally determined using eyns (ZJt(26).
are then numeri-
They are plotted as solid lines in Figs 2.1rtnd
5.
According to Fig. 3 the photoelastic results and the analytical results agree very closrly. throughout.
for the beam bendine moments. although only the two extreme points (.Y = 0
and .r = I) \s‘ercmatched for the determination
of k and G. The diqreement
of the I’(.\-)
cur\es und the pf.u) curves seem to bc caused mainly by the inaccuracies introduced by the numerical JiITerentiations
of the .\I-test curves. as discussed previously. However. it should
he noted that the slope dis~ontinlIit~
at the concentrated load Pof the ~~n~tI~tic~lf~-obt~lin~d
p(.v) curve (Fig. 5) is another basic shortcoming+
of the Pasternsk
model and is caused by
the term Gu,” in eqn (2). It is reasoniible to expect that in an actual situation pressure ~ti~tribLiti~~n I’(.\‘) will have a horizontal
the wnlxt
tangent at P; the point of s~nirnetr~.
It mily be shown that the vertical equilibrium
equation
is satisfied for the ;~nalytical {t(s) expression. iis anticipated.
Utilizing
the vari;ltional
:I f~~rt~i~~l;~ti~~~~ for thf rcasonahlc :lIld
i\ppro;tch for variable matching points dcrivcd by Kerr ( 1976).
problem under ~~~(l~i~~~r~lti(~n was prcscntud that is me~h;~ni~;illy
\VCll p0SCd.ThC ;tiialylicul solution
lll;lltl~~~~;~lically
~lhlili!lCd
W;IS
numcric;~lly itnd then comp;trctl with rclatcd photocl;tstic test results by Durclli It is rtotcwrwthy
that
mined hy colloc;ltin~ qywnrcnt
~~lt~~~~l~~l~the two
beam-shrtriring force distrihulion It
~~~ul~~~;iti~~~~pnrumctcrs were dctcr-
I’astcrmk
two cxtrcmc points on the test cur\c ft)r bending moments.
is very close tltrouphout
the lo;ttl I”).
this curve. 7‘11~agrccmcnt is only satisfactory
I ‘( v} met the conlact-prcssurc
~iis~ribuli~~n p (cxccpl
is pointed out lhal this tlisagrccmcnl seems to hc
inaccuracies ;tss&ttcr!
with thu numcrial
dill&unti;rtion
p~~sc~~t~tl
when formltl:~ting
solution of the uorrcct formulation probfcrms of structures
t In addition to the occurrcncc bnsc, as discussod by Kerr ( I’M-iL
nf concentrated
CilllSlXl
the
for the fiwr
mainly by the
01’the M-test curve.
The disa~ssion of the incorrect analyses by Chcrnigovskuyrt atld the
CV~llll~llCd
CI d. ( IW~).
(1961) and Ting
(1973).
show that caution has to bc exercised
which rest on ;L Pasternak-type found:ilion.
reactions at the fret beam
endsthat
do
not scparnlefrom