You are on page 1of 54

Cod 1 2d

hd TEN

Ce ase
dol LY 52a

EXPERIMENTS ON THE BUCKLING OF SKEWED

HYPERBOLIC PARABOLOID SHELLS

Robert Elangwe Efimba

SUBMITTED IN PARTIAL FULFILLMENT

OF THE REQUIREMENTS FOR THE

DEGREE OF BACHELOR OF SCIENCE

3

Fe

MASSACHUSETTS INSTITUTE OF TECHNCLOGY

May 1963

Signature of Author
Signature redacted
Department of Civil Enginerring,
May 17, 1963

Certified by .
Signature redacted
; oo
nd

Thi 34s Supervisor


i.

EXPERIMENTS ON

THE BUCKLING OF SKEWED HYPERBOLIC PARABOLOID SHELLS

Robert Elangye Efimba

Submitted to the Department c¢T Civil Engineering on May 17,

1963, in partial fulfillment of the requirements for the


degree of Bachelor of Science

ABSTRACT

Two shells were designed and tested to check the

validity of Heger's equation for the uniform gravity


buckling load for a skewed hyperbolic paraboloid shell,

The first shell, loaded by air pressure, could not be

buckled owing to tensile membrane action made possible

by cantilever plate action of the walls of the air

chamber. The second shell was loaded with suspended

weights, and the buckling load indicated was 53 per

cent higher than the theoretical value.


111.

ACKNOWLEDGEMENTS

The author would like to express his gratitude

to Professor Frark J. Heger, Jr., for his advice, en-

couragement and untiring interest while supervising


this thesis.

The author's thanks are also due to Mr. William

A. Litle for his advice, and to the staff of the Model

Center for Structural Design, and of the Materials

Research Laboratory at M. I. T., for their guidance.

Finally, the author would like to express his ap-

preciation to the Owens-Corning Fiberglas Corporation


for providing the necessary funds.
TABLE OF CONTENTS

Chapter . INTRODUCTION Page


loL Scope and Objectives
1.2 survey of Previous Work on Hyperbolic

Paraboloids

Chapter 2 REVIEW OF THEORY

2.1 Definition of the Surface

2.2 Structural Surface

243 Structural Properties of the Skewed

Hyperbolic Paraboloid
D 0 Buckling Defined

Chapter 3 EXPERIMENTAL PROGRAM

3.1 Design of the Experiment 11

3,2 Shell Design 11

3.3 Forming the Surface 13


Job Moulding the Shell 15
3.5 Loading Systems 16

1.6 Measurement of Deflections 'Q

Ya?
3 .

Test Procedure so

Chapter u TEST RESULTS

+] First Shell DO? 1


9D 31
Second Shell
ny \
Observations during testing Be
t
Sources of Data 0

Chapter 5 DISCUSSION OF RESULTS

5.1 First Shell: Lateral Thrust "3


5.2 Effect of Leaks in Air Chamber 3
5¢3 Buckling of Second Shell 1h
5.0L
a

Post -buckling Behaviour

Chapter ¢ CONCLUSIONS

5.1 Recommendations for Further Research


uN
Rl
I~

™)
BIBLIOGRAPHY Ny 0

APPENDIX A COMPUTATIONS ~

APPENDIX B MANOMETER ANALYSIS a


LIST OF FIGURES AND TABLES

Table

' el
Shell Design

Low] -

Shell No. 2 : Summary of Results 5


=D Shell No. 2 : Test Results

Figure
2=1 Definition of a Skewed Hypar Surface

2 =2 Structural Surface

2-3 Edge Conditions


Dl Buckling Defined

3-1 Form for the Mould Rn

3-2 The Mould 14


3-3 Shell and Air Chamber 17

3M Shell Support 18

3.5 Details of Support at a Low Corner 18

3-6 Load Distribution Discs 20

3=7 Measuring Device for Deflecticns 21

}=8 Loading with Suspended Weights 23

Sample Load-~Deflection Curve oF


Ra

Air Pressure Loading


J

1 v
Load-Position-Deflection Curves

Second Shell
14 wr 3
Sample Load-Deflection Curves : Second Shell
halt
Observations During Testing 3
A=] Shell Design Computations +0

A=2 Buckling Mode of an Edge Beam ‘3


A-3 Manometer Analysis 14
viii.

NOMENCLATURE

340 length of edge beams

shell rise

displacement
oy
|
modulus of elasticity

shell thickness

Mat{ vr ony bending moments

Nou y Nr * No tangential stress resultants

uniform gravity load

Per critical buckling load

4, 9 4., J q., components of load in the U, V, W


directions

shear alon# edge beams

total axial thrust in an edge beam

skew angle

By horizontal projection of skew angle

U,V,W coordinate axes

UV ,W coordinates

lee, WV, horizontal projections of u,v


Poisson's ratio = 0.35

unit welght of air

unit weight of manometer liquid


shear stress
CHAPTER 1

INTRODUCTION

l.1 Scope and Objectives

Ihe experiments reported herein were designed to

test the validity of Heger's equation for the critical

buckling load of a skewed hyperbolic paraboloid shell

loaded by a uniform gravity load.

[wo models were designed to fail by instability,

and tested, using two different loading systems. Loading


by air pressure failed to buckle the first shells the

second was loaded with suspended weights, and it buckled.

1.2 Survey ofPrevious Work on Hyperbolic Parabolecids

The theory of the structural action of rectangular

hyperbolic parabolecid (hypar) shells is now well estab-

lished. Reissner (8) has derived an expression (equa-

tion 2.3) for the critical buckling load of a shell

loaded by its own weight. Parme (5), Candela (1), the


Portland Cement Association (6), and Ralston (7) have

glso worked on the hyperbolic paraboleid.


The onlv known experimental work on the buckling

of rectangular hypars was done by Lee in 1962. He in-

vestigated the validity of Reissner's equation (equation


2.7) for the critical load for a rectangular hypar shell

subjected to a uniform gravity load. Lee found that ex-

perimental buckling loads were consistently higher than

theoretical loads, and attributed the increase in the

loads to tensile membrane action supported by lateral

resistance from the edge beams ard the edge regions of


the shell.

The skewed hypar has not received as much attention

as the rectangular hypar. However, Heger, Chambers, and


Dietz (2) have proposed a buckling expression (equation
2.1) for the skewed hypar similar to that derived by

Reissner for the rectangular hypar.


CHAPTER 2

REVIEW OF THEORY

theoretical skewed hyperbolic paraboloid surface


is defined, and modified to obtain a structural surface

for the experiments. Buckling behaviour is then dis-

cussed.

s
-
{ Definition of the Surface

A hypar surface 1s completely defined by its rise,


the length of its edges, and their orientation.

Consider a right-handed system of co-ordinate axes

u, V, Wy as shown in Fig. 2-1. U and V contain an acute

angle, 8, and W is perpendicular to their plane. Con-

struct a parallelogram OMQN, with one corner, O,

situated at the origin, and with two adjacent sides OM,

ON, lying along the axes U, V, respectively. The point


P is directly below Q. Lines are drawn parallel to the

plane PQM to intersect OM and NP. PQM is known as a

director plane, and the set of lines drawn parallel to

it, as the first system of generators. A similar set

of lines, known as the second system of generators, is

drawn to intersect ON and MP, using PQN as a director

plane.
The points of intersection of the two systems of

generators define a hyperbolic paraboloid surface. The


hypar is skewed if 6 is less than 90°, or rectangular

if 6 is equal to 20°,
The edge=-lengths are denoted by a and b. respec-

tively, and the rise by c, as in Fig. 2-1, These para-

meters are used to define the surface as follows

aT C_ uv (2.1)
ah

where u, v, w, are measured along the axes U.V, a

respectively.
The surface defined in Fig. 2-1 is such that the

curve joining M and N is parabolic. Any other curve on

the surface which lies in a plane parallel to curve

MN is also a parabola, and is concave upwards. Similarly,

OP and curves parallel to plane OPQ are parabolic, and


concave downwards.

The hypar can be generated by sliding one parabola

along another =- for example, MN along OP in Fig. 2-1.

From a practical standpoint, however, it is much easier

to generate the surface by means of straight lines.

The latter method was therefore adopted in making the

shells for the experiments.


242 Structural Surface

The theoretical surface defined in Article 2.1

and Fig, 2-1 is modified to get a structural surface,

shown in Fig. 2-2, by rotating it about an axis pas-

sing through M and N, such that M and N become high

corners with respect to 0 and P. Compression and

tension diagonals are as labelled in the figure.

The structural surface approximates the theoretical

surface if the angle of rotation, 6,» of the W-axis is


small.

2.3 Structural Properties of the Skewed Hyperbolic


Darabholnst”

The critical uniform gravity load, Pop? that will


buckle a skewed hypar shell, whose surface is defined
*
by equation (2.1), is given by Heger's equation :

p= — 2B hoc |
er rere
(0.2)
J3(1v2) ev uinPe

"Proceedings of World Conference on Shell Structures,


October 1962.
Notice that, for 6 = 90°, this equation gives Reissner's

equation :

o = BE _h%c® (2.3)
er _—
S31) a2p°
for the rectangular hypar.
The derivations of equations (2.2) and (2.3) are

based on a shallow shell assumption, and utilize the

(linear) Small Deflection Theory of Shells.

Further, the following boundary conditions (see


Fig. 2-3) are imposed :

{
at a = 0. hb}
¥
Se

0<v<b N
111)
i” 0

and (2.4%)
at 7 0. iy
xr ~
-

0< u-~a &



o 9 N yp

Equations (2.4) require that there should be no lateral

thrust at the edges of the shell. However, there is a

boundary shear given by

ab sin ©
S »

D (2.5)
Dn

‘neferencea 6
~
}

Dl Buckling Defined

A hyperbolic paraboloid shell buckles intc waves

along the compression diagonal. These waves disappear

upon reducing the load at buckling by even an infinitesimal

amount. A typical load-deflection curve for a point

along the compression diagonal is shown in Fig. 2-4

As soon as a hypar shell buckles, it loses its load-

carrying capacity in the compression direction all load


is now carried in the tension direction. There is, there-

fore, a resultant lateral thrust acting on the edge of

the shell.

le. If, as required by the boundary conditions


(equations 2.4), no support is provided to resist this

lateral thrust, then, the shell actually buckles at the


load predicted by equations (2.2).
2. If, there is support for lateral thrust, how-
ever, then buckling occurs at a load higher than that

predicted by equation (2.2), since the required boundary


conditions are no longer satisfied.
CHAPTER 3

EXPERIMENTAL PROGRAM

3.1 Design of the Experiment

Iwo shells were designed so as to have identical

surfaces, but different thicknesses. The surface was

formed on a mould of "plaster of Paris". The mould was

used to form the shells, out of sheets of polyvinyl


chloride, in a * vacuum forming machine. Each shell was

glued to a supporting frame, and loaded, the first by

air-pressure, and the second by suspended weights.

"2
J shell Design

The shells were designed to have the dimensions

shown in Table 3-1.

The following pol»t: were considered in the design


of the shells :

l. The material of the shells should be easy to mouldy


it should buckle before it failed due to direct or shear

‘Structures Laboratory, M. I. T.
12,

Table 3-1 : Shell Design

Shell Number

Parameter

Nn (ins) C20 > 3

8, (deg) +5 Hy 8
4

8 (deg)
a=b (ins) 12.9 12.9
c (ins) 5.9 5.9

Pop (psi) 0.813 1. 3

stressesy it should be easy to tell, preferably from the

look or touch, when the material buckleds and it should

be inexpensive. Clear polyvinyl chloride = a thermo-

plastic with a glass transition point between 90° and


100° C, a Poisson's ratio of 0.35, a modulus of elasticity

of 500,C00 pounds per square inch, and a shear strength

of 3,000 pounds per square inch = was such a material,

and was selected for making the shells.

2. The dimensions of the shell (and its mould) must

be such that it could fit into a 16" x 16" rectangle,

the size of the platform in the vacuum forming machine.

lee Appendix A Lor computations


h)

L 15" x 15" rectangle was used for the model shells

for this project.

3. The shell material must be thick enough so that

variations in thickness (about 10 per cent ) caused by

stretching during fabrication cculd be neglected. And

yet, 1t must be thin enough to facilitate softening be-

fore moulding. Sheets of PVC that were twenty-thousandths

and thirty-thousandths of an inch thick, respectively,

were considered suitable, and were avallable.

4. The shell must not be too deep, or it would be

difficult to buckle. This is obvious from equation (2.2).

-
\

Forming the Surface

The surfacewlds generated by means of straight lines,

on a wet paste of plaster poured into a form built for

the purpose.

The form (Fig. 3-1) was built so that its topmost

edges had the same dimensions and rise as the shell

surface designed in Article 3.2, table 3-1. Thin steel

plates were used to provide better defined edges (see the

figure).
A paste of quick-set plaster of Paris was mixed

and poured into the form. The paste was covered with
A.

Fig, %3=1 Form for the Mould

Fig. 5-2 The Mould


—r

1 one-inch layer of "hydrocal" paste.


A 1/16-inch steel bar was then used to generate

the hypar surface by sliding it along two opposite


edges of the form, and parallel to the director planes
of the other two edges. Markings (Fig. 3-1) at one-inch

Intervals along the edges, were used as guides.

The final surface was rot very smooth because the

bar had a finite thickness. A thinner bar would have

produced better results.

After twenty-four hours, the mould was removed

from its form, ready for use (Fig. 3-2).

-
L
Moulding the Shell

The mould was set on the platform of the vacuum forming

machine, and a 0.020" thick sheet of PVC was firmly

clamped to the upper frame of the machine. The sheet was

heated until it became rubbery. The platform was raised,

and the vacuum was turned on. The sheet was sucked on

to the mould surface whose shape it assumed (including

k
"hydrocal” is a greyish material similar to plaster of
Paris. It is also used to make moulds, and it has the
advantage of a longer workability time (about 35 mins;
c.f. 5 minutes for quickset plaster), but it is more
expensive.
1 6.

the imperfections), and retained it upon cooling and

hardening again. The shell was then cut out of the

mould. The second shell was made from a 0.030"-thick

sheet of PVC.

V5 Loading Systems

The two shells were loaded, respectively, by two

of the loading systems in current use for structural

models.

(a) Loading by air pressure


Shell No. 1 was loaded by air pressure. The shell

was used to close the top of an air chamber (Fig.

3-3), made of 1/8-inch thick plexiglas walls.


The shell was glued to the plexiglas. The chamber

was connected to an air-suction pump, and to a

standard mercury manometer. It is shown in Ap-

pendix B that the pressure registered by the mano=

meter is approximately equal to the pressure acting


on the shell.

(b) Suspended weights


Shell No. 2 was loaded with suspended weights.

The shell was glued to a support frame (see Fig.

3-4) ,consisting of edge beams designed as 1n Ap-


pendix A (3), and was erected on the testing frame
7.

fig. 3-32 Shell and Air Chamber


3

0
pA

|.

HPT

Fig. 5-4 Shell Support

Fig. 3-5 Details of Support at a Low Corner


17)

as shown in the figure. At each low corner, two

steel angles were glued for 3/% inch along the

sides of the edge heams. Details are shown in Fig

3-5. The high corners were left free. An 8 x 8

grid (spacing 1.5 inch ) of generators was drawn

on the shell, and nylon strings were passed

through holes drilled at each of the 64 grid-

points. Loads suspended from the strings were

transferred to the shell by means of plastic discs.

(see Fig. 3-6). Each disc had 3 rubber "legs"


which acted as a cushion between it and the shell.

9 Measurement of Deflections

The set-up shown in Fig. 3-7 was devised to measure

the deflections at seven selected points on the compres-

sion diagonal of the shell. A vertical measuring rod

was erected above each point. The rod had a sharp point

at its lower end, and its upper end bore on the plunger

of an Ames dial. The dials were graduated to read to

0.001 inch. The rods were clamped to a horizontal rod,

and the dials were supported on another (threaded) rod.

To measure a deflection, the appropriate rod was

lowered until its sharp point barely made contact with

the shell. The rod was clamped, and its Ames dial was
20.

Fig. 3-6 Load Distribution Discs


21.

Seman mE a ———r tT Ae Ain + SA4=AR

re nan vmmer

ms terryipte
rE A——

ni ry SR Ae2DBA im Pnann prirtoned) El

TO MANOMETER AND VACUUM PUM’

LEGEND -
.2.,3 ~ SUPPORTING RODS
4 ~ AMES DIAL
5 ~ MEASURING ROD
Fle. 3-7
6 ~ LAMP
MEASURING DEVICE 7 -~ COMPRESSION DIAGONAL
FOR DEFLECTIONS OF SHELL
AIR CHAMBER
D2

read. The rod was immediately raised well above the shell,

and clamped in position.


This measuring device utilized the accuracy and

sensitivity of Ames dials, but avoided the significant

pressure that they would exert on the shell if they were

placed in direct contact with it.

3.7 Test Procedure

1 Shell No. 1 was installed for loading by air pressure

(see Article 3.5 (a)), and the deflection measuring


device (Article 3.6) was set up. The air chamber

was checked and found to contain leaks, which

were so minute that they could not be located.

Pressures in the chamber were maintained by

keeping the motor (constant speed) of the suction

pump running throughout the test. The pressure on

the shell was gradually increased, and the shell

deflections were read at each load increment.

2, Shell No. 2 was installed for loading with suspended

weights (Article 3.5 (b)), and the deflection


measuring device was set up. A 1 pound rod was

hung from each strinz and the deflections were

measured, The loading operation was continued


\

(Fig. 3-8) according to the schedule in table 4-2,

- no -

Fig. 3-8 Loading with Suspended Weights


CHAPTER 4

TEST RESULTS

v1 First Shell

Shell No. 1, loaded by air pressure did not buckle.

This result was obvious from the look of the shell and

from the load-deflection curves (a sample is given in

Fig. 4-1) which did not resemble what was expected

(Fig. 3-4). See Chapter 5 for discussion.

Second Shell

Tests with shell No. 2, loaded with suspended weights,

indicated a buckling load of 2.50 psi. The results are

summarized in table 4-1, and details are given in table

4~2 and Figs. 4-2, L-7 ,


-

> —-—

TABLE 4-1

Shell No.2: Summary of Results

Poisson's Ratio ¥v = 0.35


Modulus of Elasticity, E = 500,000 psi

Skew angle, 38.5


FEdge-lengths, = 13.0 ins
Shell rise . 5.7 ins

Shell thickness, h ~~.


0.030 ins

Theoretical P A
er

ri
Experimental Pp 2 =n psi

Percentage difference 53 per cent


=f
—y
ply

CET
Jr o ber
pg HE Poli
bet hii Ege [ran
Lan En
mins LA BRT eed St

SHELLNo
HELL fe:

SHELL. THICKNESS
vo .
crf
oad
Zw? ee

ret fle ER —s

rr

Bi Cy

. hilt
Cie ob

ot
madd

Veen sae fren


Can).

pi “uy Seg me

liolo2S
: ‘ TEL, LA Saag
ol . col
bn nbd 7
Fr Wd .

wo safer ER
iP

pei fess
va .

aden
oox

Uo h
h
ween To

www gp 0tryeaiil To
he wae a.

a — eb
¢ odsen
Qo “a
Lol oT

bial
» ia

Sipps he SN “
“.
bp eve
Lhe
Ed

Hi ‘0.620
S100
Lie
cafe

err diIiE ai
iL

Te
‘a

mtr dt pnb
Soo To re :

eed

f1iyes Cas
SE
wage

Hak od cnet
vosinf
ern

vopplivnadoa
bo me {on
eleadeas

Faredieritesa
bine} bg ing det .
astilbin
a Ca oe . .

—t— — —— d—— p——

7 Ee

pla
piensaii]
fave.

seiabiies
tiiils

Cee.
/ baht
be bad dbs
ron
edo.
nT Tn

ye Ah
oh.
Re
cab
Tae
Cee a
.

4 Tol

Ea
ten

Fido nctpregvygwasn
opcbinhn
1t: is it
th oeete}lere
sep
Tatas

rw peedlela] tena

treere

tis :

i 4.010
1

1.
|

selppted
1

[0
|
|
$5 i

+
, Pa

doeel
Ai.
lars 40aell a

SERRE ERE EEE RY BE one


iadll
bee
alo 05 BEI
Fig
or 1
i,
*

| tilieres
ed he “bd bb oaae

anbassion
ced.
«dv
bt
po SEEuf.
webisiilioiig
ESE EE

EE
0-000 ___
PETE Ln
di 2g Sameae
in wan wll om Peasy
Fu mds ase foieolay.
st :
54
Vee}
2
gail |... 0.0 he)
Loi
sug > mri

EEaI [ee

pte bn
1
LOADONSHELLINPSL: celui

Jou
——

: iii

: ——
: ond
; Ho be pn
I=; . . Cra Ala hee :
Uy oe Affa md
aastep mtn
of fas ua
Git]
thi
5 yd : . . ———- web
wT . Pransdss cm gatas HY; panto LL. HL
= hk finial | 2
i ] i y Ba ihe ley Teall presbsnecmdTone ved gs = TE
hh
TET
| =
Ra
Ha gs
cl
nh
rea fean
: HE
“yi:
th
* EE
BERRdrownoe
dad
’ HR ¥
+. i
2 —
drt
1a
deh
Cra .
ie hes
“a
Lens ny
bogey x2 re
: {iCee
retail wb o Hwaxel oefiinibnis vinnnrng REE ch : ee a AE
. etre . PE ° CL . wowokid kk tm ~ a wala . By tr wip Baha Hn i FHI iI,
> § 81 Spon fmt ats : . Bf omer gion. wh - -
.. | - = MAB A any at cs Cipro | EEE !
: Cf 4 opp] pod t HN TeiyLIiagIiiialIl eran won reeset nl rit . —_— 1
TT
: guest ho | at I ire etesie fn "af ake Is
: : io piping deg a
; ES oA A py srerperrrfeses
pie . t : : . oe nema
+ a es ees I . «. Te
[EE h
wn boob ior aaa Sli Lill rote Hagin HU : nt |
Cee + in nib oman mimi},wo Hmm ztteprecn serheean tyne] od ifaieitans . id Lt ee CITT
awpeak EE cer epee Coober en a pigs Tlilret intial . {lien es erie = oy

Sdn
Freepeaiadis
ClaanalREIEL INS edi
ebCad Jpg Uo
hE ELE) {rit . TI tr rrrtiacses npeadiaaad vy ae tea inn Chenier
ipE verdeinitpirpon
Hirsi bobs
LINN iat ain bose
adteechi
Rumpus 3 sean eee tle Thr oa eon Tulle J103000003 00 {1513 2s : : - 143 ay ily py 2 ;
Pa L TTT ETT ysiidite Lath int lt np cme bil lptlLeirebe nn
ve tanner viailwrYatasleieigeszi
i, See sfruend Tnigeen
bo iii
Iiifeere refrac,
siiedenes I ihanaaa
mtr tsmary diwrens
iis yer
deve2d3 om ibree i douiiddi
donne+ rn; tyes Waid
SER ici
Hi der
Tn iabrarper SIT
nnat
& bee * a HE: = 4 . . . dititenar 4 pbs A pint rrihte Suro ve ST oe TI : 1s : iH i i Ht = i

Be
|
tint
|
Ri Ei ip JH
ee Taian TIT
2h
iis i: inetd saiitiiag!Laailils
herr veepertitiilac le reaytiresl
}
3 3 iE
5
Lod
rari trL
thine
he rt Ua
———t— Cede
¢
ai Iii artiinabntes rEELrrL icin
Toe yet
Trrefrrrtferripiess
rin
«isi Aprils
L980.
fotense
i PTTRaab abt a
iimatilgtliy 1s iitfagtd
riiiri fie ey
isin,
Sigs
oo
TETee
rity
3g
wig
| : | iii, BRIE REEL: Se SHIH crisis Pte be TT Pieagcan Pid

7 . . . criioiiniiape
oni, !
Sino nonin airfare
SIT
Crprrrifense
aon toed,
oy fvterhein
ght ng
IIT Tn cotillion
mp
SiLLRNyRI Sipe
pHa Jil
nT nnn
Hg
SEER
ese fi:
Ff
fo ps |
0 ti pon TTT rates cotinine rbrirhiaiihiinidapii, ot bafageadoitige ro
ariah .- ' . eee boned. cic havea bean dea voor nefbicabetsni AAA a8 Ee — -

2 : : a a Eh et . Te Seibipin, did HHH iit iat


wenssasiils . ie ! ‘ olan onan nbn einige Hail rE iil Lifiignndiny hi
Ce : — tee Poudre be ee ihr ht sri ts 5 pn:+5bokeh [ergo basboyn pif epee

jrecglesan -ve J "ual Lil Cee ey


Cd | i i HL nr dalek tT Pitiligbitpehid 2iailaaiit alge.
cE rid TTS
: Hippie Ti es y Yili
0 slp Ira chloong
Tottori ut; t pha Pashalecbo —
LL
LL rot4{ent0 0 ire
: . .
Lh
oe dsm43
wil ¢ oe sedges
Poe]
Er
Spied
v Pes
ae]
Senin
¢ Irs cm YL.
deen ITY toy
l “jeferda
«5
CTT SeltnInEp
Tet LL Epi, alrlir J
Borat.
Coy tiopoitopn 2315s stbvrob rnin oid rls prt er ia: 2g 3 e ng Lo
TABLE 4-2 :¢ Shell No. 2 : Test Results

DEFLECTIONS , 10 ins.

’oint on Com-

pression Diagonal
Load per Equiv. Load
Load Foint on Shell
(1bs) (psi)

U 0 C 0 0 0 )

| > 0125 ous 042 0540 0385 l 0500 0030


0910 0885 1035 0135 |
C.71 0425 0950 0917
0.89 0455 1317 1217 1480 1185 | 1285 | 0300
1.C7 0570 167" 1392 1950 1475 11720 0465
1.43 0825 23. 1832 2800 2455 | 2585 0830
1.79 1400 3420 3087 3830 3135 3835 | 0830
2.1% 1865 Leh o14062 4850 3285 6045 | 0850
i Jo 2.5 003% 323514802 54.92 | 3435 5755 ' 0870

Downward deflections are taken as positive

Dead Load neglected: nominal value of load (iron rods)


-

J3 >

Sep

cides
‘ah
Lud ELT
phe EE
. Cra

eokids aad . iia


i La .
san deme)
Corbitt
i Hatha,
he

FIG...
rT TT,
4-2 OT eT

LoAD-posITION ~DEFLECTION
Cy
Lea
Ce


cogupet gece > sheLL
vegans cage
2 edad gn Saag
:

A legen.
heey pny show hk oa so

onl ony. Co
to ptr piri pnd dottrararate

frvdeiin oes .
~ Paral eaatom
ed J 4 sos laii

a
ETI

a
at

lin ht pihsir

rv
Soe ug rar ed edad

ao te 17:
provi
es
Ton
5) in
inh
dietenn
Toy TTT

Geperyeieny
\ ENN whew
AE 3 4 4 +
obo.
om

NG Thre
~ Cn Eien
feiliian
atdd
oo
WEDBEEp rari
i bad bee bea fen

£ -
ih ZF
HG tytn
3

fm
Vols
foi

forb
i

=i.
BS
i
Ld soap \ PC
- spiininihy
sole 9 Y ivi;
bE pidotBebb
weber

ily, hn
Fare
wisniLeaamw
ea
. ui G bea
bated be.
ed oe
vera ss.

KO
.l hay
; a anes
atin I Lv PE Fe

Pa [ERE Taan foes


U:
Cl
u
«sl
val
g> vr
pe
phi
wow [EI
anata niia
bangs des bes

ed vena db .

rT
o>
+
drapskiows

bb ipetti bi eb
io. ‘ + be FE

| La
EY

Fr rth fanaa,
HH efi
Y IE he eis Lh

SO > vba Lie


aa
PE
ul
w veda dae habe chet
ce dren bone ead

3 Lif oasr beens d in caga


Cres i
i
w daadibeabiefeeLh
I
oO Selita
Ea eb tai “a pods

SE pen a! Code
vb Ve ih, gr med id
. a . cae ve

pb d es a rm bor mdm pn

sud 5= yoda

eons
Toadies

.
eefeesiriaanole
dec

orb

ih
ei bbe

ae
Sebel

Ww
ee

wwii
bean

oe
~hpriagriilyits
labiaphe vib ens
tl,
" .
iia pp
— ds = Sl

- Srila Jods
pefeles apd shiloh
ciate
Spiel
dbea braneeanaabe
cia
Pht
Taaliea
spelen bb reaps .
Piatti iva teehee
‘ ek dw cee 4s x wnaee %
; s be retet
wees} iment oT PESTLE pot oped
chavs brea beri peas MII PE
C Phesab pba
doting
tlie
Pega
JIITT Ira FI
fib
an IR
traabeees
fy
3 besiege bea abi vie debe
fants rR
TT TT Ti —
fin
re
Glib ET
iii a) ——— -atb

2 rite
brefeldin greTn
fia ent
dp this
» “oe bE. ie TE Fa bes {oi porte
- sd. ——— le brit

elt
SEE he DLN Lorriiitapi
oe as teteoie fede eet dgfbb tpmredeenfa
calf nan
suf
La
rrr ed
ol vo
ar
Ce
magic? Hii!
ae Pee Ca “oa
PE a He
i — —— — pieh

LULL.
ai pi I
edadre prriifippes
bibs sree emia. A J .
pan vim.ey

'y Sef cobb Ts


1 .
1h Sh

cline
. “
wt

ndii
we
an pp
PE
wei:

prop
Fa
some
awe

ohnanpnonein on
ede . .
eli,
[a
Hi,
ra

SUE
: re

Lob twa FH EEE Sledag cep


wh [ooh eet ITT
SEH £1 Lo cole)
reat agbanh tha ee eee
AY is boo
LL Coq TIRE TIME NTI .
— NS ook a a. bebettf stpthttm np ete —-- de
sm— —

.
———————

, x for
. -

Cay ; CL
eben Ir ued
Taig ee [ER

Lg ie |
Ce

—t ereetbrraliates
brberer>frmyy Gs -

Lilly 3
ee eee _ 03.gh
dy Cle
rad ol Soto
] Jia Ql
al. i yaadiit
Lei
AHIHF Seagal igs d 3 geodedeafornils 300 .
cee mre orb bere ey CB nh HQ
afi
Colne g nnfi 4: . Bul oaLL LeSpite
“erndnbindin Li DAWNIWAR CDEFLECTIONS IN
phar

IMEHES] iin
.
feet ti
. !
eek ee
eda

chap
rtf een br wp
baer

bibyeae dae
hi 5

shi:mogssied ngs -
Ce

- pe ne wn eee TTT TTT TITTT TY] ¢ agen » wwL

EE EE REE
od :
cob

po

Seth5A

Aide

3.4
Lo cep
#7 be
SiH
So a HE
fin end
bio Soy Jie, efvaan
Rit of Hs a HE
= pen om epee lee
0.40 —_— a

HRN Fie eiligaiiifoeofen


pe

an pid Co

Ce \ ; Loy an

Nii
Tae

bf cides
no vege ls
} A
ts HRs

Ta ta

LeeeNp oll
ko
Poon
PotNT
L-f 1
HEaTds
3 li fol ‘0||ponT #4 I
-

I pot com PRESsIaN]


yt ~
DIAGONAL:|il
——y 1

a.25
redid dns
Cob
. t

Jian
Po am
,
~
Fh?
TE ovINTTC
SHELL THICKNESS.

NEE

‘em
. Ea Ion
'
sb

———
pg

es
0.20
EN arp . 8 pit

Ie
3 LIL.

ny
breed bee weber dee a faa

Ary,
Pa on

0 LL
foavmoen .

2.15 bs ream i eee


sabi ERTS
Leaf aang fim
Lo

ar JH
.
Wf
Em
Sh Saepbi

————r
0-10 Vt ae
a Spiny Hie
EEL ib
iisos

hg as

cad Pring oo Las:


fri “7 4 Ten ae eae

So aoe en
ead
es

ke TO TE i
0-08
poo i Ea or
——
ey
BUCKLINGLOA
0.07
Hs Zen 02S Th %e To
LOADunPst La

beRIGTASSAMPLELOAD P RE to
Cobre
rn eR
i i iE nie aes SEO
ernDISMELL
poe laiilaatdf
ETT PEE Efi RNR ETNIES egress yrenabeernd es enpnnnahipofergs played titlepren
de
0.

Observations during Testing

The following observations (Fig. 4-4) were made during

the testing of shell No. 2.

During load change from 2.14% to 2.50 psi

1. minor shear failure across a shell edge, near -

low corner

2. an edge beam pulled away from angle support, and

was bent inwards.

At 2.50 psi, shell not yet rubbery.

3. Two "bumps" along compression diagonal


4. Shell buckled into a deep valley (along com-

pression diagonal) near a low corner.

5. Loads were almost touching floorj loading could

not be continued.

Sources of Data

The Data in table 4-1 was obtained as follows

Poisson's Ratio, Modulus of Elasticity, and Shell thick-


ness were obtained from the manufacturer.

The Skew angle, Edge Lengths, and shell rise were

measured after the shell was aligned for loading.

The Theoretical Buckling Load was obtained by sub-

stituting for the above parameters in equation (2.2).


1.

I )
a

Fig. 4-4 Observations during Testing of


Shell No. 2 : numbers in figure
correspond to observations in
Article 4.3.
DD

The Experimental buckling load was determined 1n the

light of observations made during testing (see Article Eel).


The dead load (about 0.003 psi) was neglected.
CHAPTER §

DISCUSSION OF RESULTS

In tris chapter, the non-buckling of shell No, 1 is

explained, and the experimental buckling load for shell

No. 2 is set at 2.50 psi.

5.1 First Shells T~t-~ral Thrust

Shell No. 1, which was loaded by air pressure, failed

to buckle because one of the edge conditions (see Article

2.3, equations 2.4) was seriously violated. This was the

condition that there should be no support for lateral

thrust along the shell edges.

At loads lower than the predicted buckling load


(0.765 psi)y the loads were carried in both the compres-

sion and tension directions of the shell. At the buckling

load however, all lozd could still be carried in the

tension direction. This was possible because = through

cantilever action = the walls could support a lateral

thrust from the tension direction.

. Computed for the actual dimensions prior Xo loading


5.7 Effect of Leaks in Air Chamber

[t 1s reported in Article 3-7 (a) that leaks in the

air chamber necessitated keeping the pump motor running

all the time. The ensuing air stream through the chamber

not only altered the desired static nature of the shell

loading, but also introduced buoyancy effects, which


might have affected the buckling of the shell. However,

since the leaks were very small, their effect could be

ienored.

5,3 Buckling ofSecond Shell

Observations made during testing of Shell No. 2

are reported in Article 4.3. Observation (3) was a sign

that buckling was near. The author felt that, under the

prevailing conditions, the shell would have buckled at

a little over 2.50 psi. Buckling was probably delayed

because of what happened in observation (2). With the

further aid of Fig. 4-2, therefore, the experimental


buckling load was estimated at 2.50 psi, 53 per cent

larger than the theoretical value (1.63 psi).

Kk
For an estimate, it was all right to use the nominal
value of the live loads and to neglect the dead loads.
wr

The above result is surprising because the imper-

fections (Article 3.4) in the shell should lower its

load-carrying capacity.

5. Post-Bucklinz Behaviour

The high buckling load is due to the fact that the

edge beams and the edge regions of the shell provide

lateral support (see Section 2.4) for post=-buckling


tension, thereby enabling tensile membrance action.
36.

CHAPTER 5%

CONCLUSIONS

The two simple tests performed 1ndicate, respectively,


the following conclusions :

l. The uniform gravity buckling load of a skewed

hyperbolic paraboloid shell is much hirher than the buck-

ling load predicted by Heger's equation.


2. Loading by air pressure cannot be used to buckle

a shell unless some way 1s found to prevent cantilever

plate action of side walls which would restrict inward

movement of the shell, and permit membrane tension.

5.1 RecommendationsforFurtherResearch

More models should be tested

1. to prove (or disprove) the validity of con-

2lusion (1) above,


2, to find how the buckling load for a skew

hypar shell varies with the skew angle, the rise, the

edge lengths, and the thickness, respectively, when all

other parameters are held constant.


)

to study the effect of edge members on the

buckling cf the shell, and


de
r

- investigate post-buckling tenaviour of the

shell.

5. better method should be devised for loading

(and buckling) a shell by air pressure. One suggestion

would be to modify the equipment in Fig. 3-3(a) by cut-

ting the walls just below the shell, and sealing the cut

with a flexible tape, so as to remove the lateral re-

straint on the shell.


18.

BIBLIOGRAPHY

l Candela, F., "General Formulas for Membrane Stresses


in Hyperbolic Parabolcidical Shells", ACI Journal,
October 1960. p., 253.

>, Heger,y F.J., Chambers, R.E., Dietz, A.H., "On the Use
of Plastics and Cther Composite Materiels for Shell
Roof Structure",ProfessionalPaper P 62-10 ,
Dept. of CE.» TT T.

}. Heger, F. J., Dietz, A.G.H.,"Structural Feasibility


of Reinforced Plastic Shell Roof Construction
for the Atomedics Research Hospital", submitted
to Owens-Corning Fiberglass Corporaticn.

Lee, S.T., "The Buckling Behaviour of Hyperbolic


Parabolcid Shells", S.B. Thesis, Dept. of C.E.,
M. I. T., May 1962.

5, Parme, A.L., "Hyperbolic Paraboloids and Other Shells


of Double Curvature", Paper No. 1057, ASCE
Proceedings, September, 1956.
5. Fortland Cement Association, "Elementary Analysis of
Hyperbolic Parabolecid Shells", Concrete In-
formation, 1960.
/. Ralston, A., "On the Problem of Buckling of a Hyper-
bolic Paraboloidal Shell Loaded by Its Own Weight" 9

Journal of Mathematics and Physics, Vol. XXXV,


No. 1, April, 1956.
8. Reissner, E., "On Some Aspects of the Theory of Thin
Elastic Shells", Journal of the Boston Society
of Civil Engineers, Vol. 42, No. 2, April, 1959,
PDP. 100 -133 .

9. Timoshenko, S.P., Gere, J.M., Theory of Elastic


Stability, Mc Graw Hill, New York, 1961, Chapter 11.
39,

APPENDIX A

COMPUTATIONS

Design of Shell

Shell No. 1
- 1+=C
Geometry : Try S.. (Fig. A-1)
~~
|)
Tr

Then 1]. »
11,65"

12.5"

Buckling load

020%
-

Ta i b=

~
( 2%500,000 : J ((0:02)%(5.9)°
3(1-00.35)2) (12.5)2(12.5)2s1n%1°
6
0) 1.32 x 1076
1.62%

Pow = 0.813 psi = 4.2 cm mercury


1]

Shell No.

Geometry : Same as shell No. 1,

Buckling load
Trv h 0 ~~

Pop ( L030 2 0.813 = 1.83 psi = 9.45 cm Hg


0.020

c
AnalysisofShell

Shell No. 2

As finally aligned, shell No. 2 had these dimensions:

=
BB = 38,5°

a = b= 13,0"

Cc : 5,7", h = 0.030"

Theoretical Buckling Load -

Dope i: 1 33 psi using equation (2.2)

Boundary she.r

g _— As = 15.02 lbs/lin.inch.

Shear stress
rn = 2 = 12.03 _ 501 psi
0.030

'h was actually about 10 per cent less.


DD

Total axial thrust on an edge beam

T = Sa = 195.4 1bs

Area per load point =11/2 x1 1/2 sin ~0 5Y = 1,4 sq.ins.

Theoretical p_. = 1.63 psi = 2.28 1b,/load point

Experimental p,. = 2.5 1b/load point

Experimental p_ = 2.50 psi.

3. Design of Edge Beam

Plexiglas (1/%4" thick) o "0,00C ~

3,000 psi
i
B 500,000 psi

Buckling Mode
The edge beams are assumed to buckle in the manner

shown in Figure A-2, and the critical axial thrust in

the beam is given by

T (sm), = L232 gr (A.1)

Manufacturer's specification
x K
Reference (4), p. 29
For assumed buckling mode (Fig. A=?)

(SL ) = 2.837 EI (A.1)

Shell Nc

~
But *.0 ins
mT
snd (SL), = 195.4 lbs

Jsing (A.M
r
nN
8,42 10°F in

Nith factor of safety = 2.

[ = 1.68 1075 In

I'ry Tit
——
* /8" section

1
Tr = x1x (4)
=) = 1.63 - in
1079

Check o = 229% = 1,563 psi 1¢.,200


1x1/8

Therefore 1" x 1 /7n secivion {

but use 1" x 3/8" section to facilitate glueing operation


45,

APPENDIX B

MANOMETER ANg 21S

Consider the sketch shown in Fig. A-2, The pressure

at a point B just below the shell is given by

PR + Y Hy + YH, = Patm

Capillarity effects in the tube are neglected. The pres-

sure acting above the shell is atmosrheric. Therefore,

net pressure on shell is

y "4
>

D
Potm Pr Y Hy
-

Try Hg-manometer

aa, YH

r v.He
p =~ h - Ho
J = 0.813 psi
noe 0.8 x 76 cm Hg
14.7

h We.
ry 2
Cm Hy

Standard mercury manometer o.k.

You might also like