!sCA)-sLdt-

E-d

E

i-.

FOR AERONAUTICS

TECHNICAL NOTE 4011

SOME ASPECTS OF FAIL-SAFE

By Paul Kuhn

DESIGNOF PRESSURIZED

and Roger W. Peters

Langley Aeronautical LaboratoryLmgley Field, Va.

Washington

June 1957

FUSELAGES

https://ntrs.nasa.gov/search.jsp?R=19930084958 2020-02-13T20:24:39+00:00Z

ID

b

d

.

w

NATIONAL ADVISORY COMMITI!EE

TECBNICALI’KYm4011

SOME ASPECTS OF FAIL-SAI?EDESIGN OF PRESSURIZED FUSELAGES

By Paul Ki&n and Roger W. Peters

Separate investigateions have dealt with the critical crack lengthof flat sheets or of unstiffened cylinders and with the type of ruptureexperienced by stiffened cylinders. These investigations are correlated,supplemented by new tests, and cozibinedinto a uniform scheme for pre-dieting criticalized cylinders.

‘NET

AR

KU

Ku,CYL

z

P

r

ts

crack length and type of

INTRODUCTION

rupture in stiffened pressur-

The fail-safe design of pressuriz~ fuselages is a problem that hasattracted much attention in the past few years. This paper is a progressreport on work in this field and an attempt to correlate several lines ofinvestigation.

SYM20Ls

net area, sq in.

ring area (cross sectional), sq in.

stress-concentrationfactor at ultimate load for flat sheet

stress-concentration factor at ultimate load for cylinder

ring spacing, in.

load, kips

radius of cylinder, in.

skin thickness, in.

2 NACA TN till

5

a

%x%

ler@h of slit or crack, in.

stress, ksi

maximum stress, ksi

ulthnate tensile stress, ksi

DEFINITIONS

The problem is defined in a general way by two questions: If ini-tial damage, such as a fatigue crack, is inflicted on a pressurizedstiffened shell, how large can the damage be before the pressure causesa rupture of the shell, and what is the nature of the ru@ure? Theonly type of initial damage considered here is a longitudinal crackor slit of length 5 as shown in figure 1. (For aluminum alloys,to which this discussion is confined the difference between a fatiguecrack and a fine slit is negligible.j For any specified pressure orhoop tension, a critical length of crack exists at which the pressurewill rupture the skin. The term “confined rupture” in this paper willrefer to a rupture which stops at the nearest rings, as indicated onthe sketch at the left. The term “unconfined rupture” will refer to arupture which extends into adJacent bays, as indicated on the sketchat the right. An unconfined rupture of the skin often results in fail-ure of the rings and sometimes of the stringers.

SURVEY OF PREVIOUS INVESTIGATIONS

Critical C!rackLength

Figure 2 shows schematically the first two steps in the investiga-tion of critical crack length. At the top is shown a flat sheet undertension with a central crack of length 5. When the maximum stress

‘win the sheet at the two ends of the crack becomes equal to the tensilestrengbh of the material, the sheet will tear apart. At this instant,the maximum stress is equal to the product of the net-section stressP/A~T and a stress-concentrationfactor ~.

A method for calculating the factor ~ has been published inreference 1. The calculation involves the stress-strati curve and asize-effect constant which is determined from a tension test on a speci-men with a sharp notch of known radius. The presence of this size effect

i

●

m=

—

●

NAM TN km. 3

●

u

in problems involving cracks or sharp notches invalidates the mechanicallaw of similarity that geometrically similar structures fail at the samestress and mskes it impossible to draw generalized curves based on dimen-sionless Psm3meters.

The lower sketch in figure 2 shows a pressurized unstiffened cylinderwith a longitudinal crack. The stress-concentration factor for such acylinder is calculated by the formula shown. The quahtity Ku is the

stress-concentrationfactor calculated for the configuration obtatied byunwrapping the cylinder into a plane. The term in parentheses is the“curvature correction,” which was found empirically and applies to 2024-T3ahnntium aI.I.oyas well as 7075A6 aluminum alloy. The experimental basisfor the formula may be found in reference 2.

The importance of the curvature correction is shown in figure 3.For the two alumdnum alloys, the critical crack len@h is shown as func-tion of the tensile stress for flat sheet and for cylinders with a radiusof 15 inches, which is a widely used size representing roughly a l/4-scale model of a fuselage. For all but very sh~rt cracks, the drop instrength due to curvature effect is obviously substantial. Curves ofcritical crack lengths of unstiffened cylinders corresponding to thecurves shown in figure 3 will. be used later as a yardstick or referencebasis for stiffened cylinders.

Type of Rupture

The initial investigation of the problem of type of rupture was con-ducted on cylinders with mdii of 15 or 24 bches, stiffened by stringersand riveted-on rings. Subjecting these cylinders to a constant internalpressure and to repeated torsion loads resulted in fatigue cracks at 45°to the cylinder axis.

Figure 4 shows the results of the initial investigation on 2024-T3cylinders. Hoop stress is plotted as the ordinate and ring-reinforcementratio, the ratio of the cross-sectional area of a ring AR to the area

of the associated skin Zts, is plotted as the abscissa. The dashed

line, Mbeled “theoretical criterion,” is based-on elemetary considera-tions and gives the area which the rings must have to carry the hoopload if the skin itself cannot carry it because it is cracked or cut.The circles denote confined ruptures, the x-marks denote unconfined rup-tures in the tests. The solid line, labeled “empirical criterion,” isapproximately the upper boundary of the confined ruptures. The ring-size criterion cam thus be used as a basis for predicting the type ofrupture. The data are taken from reference 3, except that the curveshowing the empirical criterion is drawn here in a somewhat more con-servative fashion than in the reference.

4 NACA TN 403.3.

A few tests of the ssme nature have been made on cylinders of 7075-T6 ●

material. The nuniberof tests was too small to establi;h an ercmirical’criterion for ring size,with cyclic torsion, has

and the method of testing, pressure c&bined--

been replacedby pressure cycling. b-

NEW INVESTIGATIONS

Presented herewith are evaluations of more recent data. The pres-entation is aimed at answering two questions:

(1) What correlation exists between the critical crack length forunstiffened cylinders and that for stiffened cylinders?

(2) How reliable is the ring-size criterion for predicting the natureof the rupture of stiffened cylinders when the cracks ere longitudinalrather than at 45°, as in the original investigation’2

The majority of the tests discussed were actually made to obtainsome prelimhsry information on the effect of parameters not consideredpreviously, for instance, type of rings used: As a result, the testsavailable at this time are inadequate for giving definite answers to thetwo questions posed, but they do indicate trends.

The tests fall into three groups according to size of cyltider:small cylinders with a radius of 3.6 inches; medium-siz~ ones with aradius of 15 inches; and full-scale ones with a radius of about 70 inches.All were stiffenedby stringers and rings or hoops.

*

Lr

The sma3J-cylinders had precut slits -were brought to rupture byincreasing the pressure steadily. The crack-length results are shown infigure 5. The curve, labeled %, gives t-k critical crack length of

unstiffened cylinders having the same radius.—

The circles denote againconfined ruptures; the x-marks, unconfined ruptures. Regardless of thetype of rupture, all the critical crack len@hs plot fairly close to thereference curve, which means that the critical crack lengths of the stiff-ened cylinders in this &roup of tests are equal to those of correspondingunstiffened cylinders.

Figure 6 showsthe ring-size criterion plot for the ssme group oftests. Each test point represents the same specimen as the point at theseinestress level in figure ~. The empirical criterion line is takenfrom figure 4. All “rupturesobserved are in agreement with the criterion:unconfined if above the curve, confined “H below the curve. The cylindersin this test group had either rather light rings or else rather heavy

●

rings; as a result this test group does not give a sensitive check on the*

NACA TN 401J 5

● accuracy of the ring-size criterion. On the other hand, the fact thatrather extreme ring sizes were used tends to increase the weight of theevidence regarding crack length shown in figure 5.

wThe next group of tests is on medium-size cylinders (la-inch radius)

of 2024-~ material. On these cylinders, the internal pressure was cycledto grow a fatigue crack, starting from an initial slit. Figure 7 showsthe crack-length plot. The critical crack lengths for this group of testsare consistently longer than indic@ed by the reference curve, the excesslength varying roughly from 40 percent to lCKlpercent. The result thusdiffers markedly tiom that obtained on the sndl cylinders.

Figure 8 is the ring-size plot for the medium-size cylinders. Thet~es of failure are in agreement with the prediction, if the predictionfor the border-line case on the line is made conservatively. This plotshows only 3 points, and thus no counterpsz%s are shown for many ofthe points shown in figure 7j the missing points are discussed in thefollowing paragraph.

On the small cylinders, and on the three medium-size cylinders dis-cussed in the previous paragraph, the rings were riveted continuouslyto the skin. On the other medium-size cylinders, the rings were eitherfloating, that is, not touching the skin at all, or they were touchingthe skin, but riveted to it only at the intersections with the stringers.

●

Rings of either t~e are believed to have little if any power to confinerupture; therefore, the assumption is made that the presence of either

‘- type of ring justifies a prediction of unconfined rupture, and conse-quently, it is unnecessary to plot the point on the ring-criterion plotin order to arrive at a prediction of the type of rupture. Sa far, therea~ears to be no evidence that the assumption is unduly conservative.

Figure 9 shows the crack-length results for medium-size cylindersof 7075-%. All the points plot close to the reference curve. Sincethe corresponding plot for 2024-~ material (fig. 7) showed excess cracklengths ranging &cm 40 percent to 100 percent, it may be said that2024-T3 has a “hidden mrgin of safety,” under some conditions, whichappears to be lacking in 7075-T6 material.

Figure 10 is the ring-size plot for the 7075-T6 cylinders. Only thetheoretical criterion is shown, since no empirical criterion is estab-lished, as mentioned previously. The type of rupture is as expected insJJ cases, but the location of the points is such that the empirical cri-terion stfi remains undefin~. -

Figure 11 with data taken from references 4. made by two aircraft manufacturers on full-scale

‘a radius of about 70 inches. The critical crack

+

—

and 5 shows two testsmodels of fuselages withlength is just above the

6 NACA TN 4011

4

reference value for specimen A and about 60 percent larger for specimen 33.This scatter suggests a scatter band somewhat shiil.arto that obtained onthe medium-size.cylinders. u

In specimen A, a saw slit had been made through the skin as well asthrough the ring underneath it. With one ring thus out of action, thering-reinforcement ratio was marginal, which would lead to the predictionthat the rupture will probably be unconfined. For specimen B, an uncon-fined rupture would be predicted because the rings were floating. Thesynibolsindicate that both spectiens had unconfined ruptures. It should

—

be noted that both specimens served as startingof final designs.

. points in the development —

CONCLUDING REMARKS

The crack-length criterion and the ring-size criterion appear tooffer some promise as tools for putting some aspects of fail-safe designon a quantitative basis. The crack-length criterion in its present formstates that the critical crack length of a stiffenwl cylinder is at leastequal to that of an unstiffened cylinder of the same radius. The ring-size criterion, which is reasonably well established for 2024-T3 aluminumalloy but not for 7075-T6 aluminum alloy, appesms to permit a prediction

*

whether the rupture will be confined or unconfined when the rings are—

continuously riveted to the skin. When the rings are not so riveted (or &-otherwise fastened)} the rupture will probably be unconfined.

Much work remains to be done, however. The conditions under whichthe critical crack lengbh of the stiffened cylinder canbe greater thanthat of the unstiffened cylinder should tieestablished more fully. Onlystray bits of information are available at present on a nmber of factors,such as effect of stringers, of load-carrying members bridging a crack,and of producing the initial damage very rapidly.

—

Langley Aeronautical Laboratory,National Advisory Committee for Aeronautics,

Langley Field, Vs., March 6, 1957.

.

v

?

● REFERENCES

* 1. hfcE’vily, Arthur J., Jr., Illg, Walter, and IW’drathj Herbert F.:Static Strength of Aluminum-Allqy Specimens Containing FatigueCracks. I?ACATT?3816, 1956.

2. Peters, Roger W., and KUhn, Paul: Bursting Strength of UhstiffenedPressure Cylinders With Slits. NACA TN 3993, 1%7.

3. Peters, Roger W., and Dow, Norris F.: Failure Characteristics ofPressurized Stiffened Cylinders. NACA TN 3851, 1956.

4. Sorensen, Arne: Some Design Considerations for Tear-Resistmt Air-plane Structures. Preprint No. 618, S.M.F. Fund Preprint, Inst.Aero. Sci., J=. 1956.

5. @afiding, E. H.: Observations on the Design of Fatigue-Resistantand ‘Fail Safe’ Aircraft Structures. session 8. paper 2. Hesentedat International Conference on Fatigue of Metals sponsored by BritishInst. Mech. Eng. and A.S.M.E., Sept. 10-14 (London) & Nov. 28-30(New York), 1956.

NACA TN ~11

*

TYPES OF RUPTURE

4-ii–-y-+,–-–’.,- 4—.—-——-‘\–\ ; ‘\ ~\\ y- ;—*—-—.- +—___! .4 )

-. —-— .

\—.—-——-—+—-—T-/’//!! -F*T//-—I——— .—. — +___

/ / / ‘+J--+-–+-—’/1CON=INEORUPTURE UNCGWINEO RUPTURE

Figure 1

STRENGTH OF CRACKED SPECIMENS

‘MA)( = +T KU= au

Figure 2

*

r

NACA TN hell

.

9

HOOPS_RESS,KSI

60

40

20

1

CRITICAL CRACK LENGTHSUNSTIFFENEDCYUNDERS

r=ls

o 2 4 6 8CRACK LENGTH,8,IN.

Figure 3

RING-SIZE CRITERION FOR 2024-T3r=15AND 24 IN.;PRESSURE AND TORSION

40r x

I ‘Mp’R’cALcR’TER’ON7+’

r0/ o CONFINEDRUPTURE

/x UNCONFINEDRUPTURE

, 1 v I , Io .2 .4 .6

RINGREINFORCEMENT RATlO, AR\ltS

Figure 4

CRACK LENGTHS RR 2024-T3r=3~ IN.; FRE.SW+ Rl~~

30

Ii

x

208~ o

HOOPUNSTIFFENED

STK&S,CYLINOER

o

,~,,4

o8, ltiCHES

Figure 5

RUPTK’E

o CONFINEO

x u4c(X4FINED

4

RING SIZE CRITERION FOR 2024-T3r E 3.6 IN.: PRESSURE RISING

33

vx

200

x

+!f!:s,

r o 00

10 :MPimml

RUPTURE

o CONFINEO

x lR4~FW0

figure 6

I

* 9

k!

II I

#b 4. * r

CRACK LENGTHS FOR 2024-T3r=15 IN; PREWRE CYCLING

30

[ ‘\RUPTURE

o 0 C(XWINEOX UNCONHW

tI I 1 1 I I

o 2 8 10

8, lM(lNCLtDING lNITli DAMAGE)

R1.gure7

I

!4

RING-SIZE CRITERION FOR 2024 -T3r=15 IN.; PRESSURE CYCLING

30-/

/

/ 0

20 - 0

HOOPS’m&s,

/

/

/1 I

0I

2 .4 .s

AR/its

Figure 8

PP

30 -

20 -

IimPSTKRf@

10 -

CRACK LENGTHS FOR7075-T6r=151N.; PRE2SURE CYCLING

\

RuPTuRf

o CONWED

x UNCXINFINED

o

Lx

o8~

UNSTWENEDCYLlfOER

L 1 I I r I

o 2 4 6

8,1 N.(INCLUDING INITIAL DAMAGE)8

10

Figure 9

● ’1

RING-SIZE CRITERION FOR 7075-T6r=15 IN.; PRESSURE CYCUNG

30

r /’

I /20 “/’

HOOPSTRESS,

I

xxKSl /~

/

10

r /“/

/F 1 1

0I

.2 .4 .6

AR\lfs

Figure 10

3D

.

b

NACA TN kOll 13

CRACK LENGTHS FOR 2024-T3r=701N.;PRESSURE CONSTANT

30=

20-

I-IOOPSTRESS,

KSI

10-

I 1 I I Io 4 8 12 16 20

8REF

UNSTIFFENEDCYLINDER

S,INCHES

Figure l-l

.

NACA -Langley Field, Va.