Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Torsional Buckling Analysis and Damage Tolerance

of 6raphite/Epoxy Shafts

O. A. Bauchau*, P. S. Bryan~. and A. M. Peck««

Rensselaer Polytechnic Institute


Department of Mechanical Engineering,
Aeronautical Engineering d mechanics

«Assistant Professor
««6raduate Assistant

Abstract Alternatively. a longer unsupported length of


shaft can be used for a given rotational speed.
The torsional buckling load of Graphite/
Epoxy shafts with various lay-ups was measured For a realistic drive shaft system, improv-
by means of a static torsional test, and found ed lateral stability characteristics must be a-
in good agreement with analytical predictions. chieved together with improved torque carryinq
In a first attempt at characterizing the damage capabilities. The dominant failure mode, tor-
tolerance of these shafts. the residual strength sional buckling. is strongly dependent on fiber
of shafts with drilled holes was measured, and orientation angles and ply stacking sequence.
the dominant failure mode resmined torsional This paper will first present experimental mea-
buckling. The buckling mode patterns are seem- surements of the torsional buckling loads of
ingly unaffected by dasmge. but buckling loads circular cylindrical shells made of laminated
are lower compared to undamaged specimens. In composites with various lay-ups. and compare the
a second phase. actual ballistic tests were con- experimental measurements with theoretical pre-
ducted using 4 30 caliber weapon to damage tor- dictions.
sionally preloaded shafts. A direct 30 caliber
hit was found to decrease the allowable torque Design studies [1,2] have shown that signi-
by about IOB only. For a 0.60 inch ballistic ficant weight savings can be achieved with com-
dasmge it was reduced by about 30%. Furthermore. posite shafts. however their dasmge tolerance
the residual strength of ballistically impacted characteristics resmin largely unknown. Foreign
shafts were found to be similar to that of shaft object dasmge during maintenance of normal op-
with drilled damage. Torsional buckling was eration, as well as ballistic impact damage pre-
the dmninant failure mode for both undamaged and sent major threats to the drive system. A full
damaged shafts. hence stiffness characteristics. assessment of the damage tolerance of drive
rather than strength characteristics seem to systems should include ballistic impact tests on
dominate the behavior of thin-walled Graphite/ loaded. high speed rotating shafts. However,
Epoxy shafts. such a test is difficult and expensive to Per-
form, hence. this assessment was done in two
Introduction successive phases.
Advanced composite materials seem ideally In the first phase, the residual torsional
suited for long, power driveshaft applications. strength of shafts with drilled holes was measu-
Their elastic properties can be tailored to in- red in a static torsional test. Circu1ar holes
crease the torque they can carry as well as the of various diameter (as would result from a
rotational speed at which they operate [1.2]. high velocity impact), and imbedded delaminations
For thin-walled shafts. the failure mode under (as would result from a nonpenetrating impact)
an applied torque is likely to be torsional were investigated experismntally. Torsional
buckling rather than material failure. On the buckling remains the dominant failure mode. and
other hand, the rotational speed is limited by the buckling patterns were seemingly unaffected
lateral stability considerations: most designs by the damage, even though the buckling loads
are subcritical. i.
e. , rotational speed must be were lower than that of undamaged specimens.
lower than the first natural bending frequency
of the shaft. This frequency is proportional to In the second phase, actual ballistic tests
/(E/p ) where E is the longitudinal stiffness were conducted using the N-I rifle (a 30 caliber
modulus of the shaft and p the material density. weapon) to inflict damage in torsionally pre-
For lay-ups containing a significant proportion loaded, nonrotating shafts. Extensive damage
of fibers running along the shaft's axis this was observed at the exit side of the laminate,
ratio can be made larger than for metal shafts, involving delamination of the outermost ply.
resulting in higher natural frequencies. Taper- The preload level was increased until impact
ing the wall thickness distribution along the fracture was obtained. The residual strength of
span can further increase this frequency [3,4]. those specimens that did not fail upon impact

1091
was also measured. Significant damage tolerance Torsional Bucklin
was observed for ballistic damage of up to 0. 60
Behavior
inch, and the residual strength of shafts with
ballistic damage was found to be closely related Table 2 lists the experimentally nmasured
to that of shafts with drilled circular holes of buckling torques Tcr. their experimenta'I coeffi-
comparable size. cients of variation, and the critical shearing
stresses vcr » Tcr/2nR2h (h is the wall thick-
5 cimen Fabrication ness) for the various specimens; each critical
load is an average of 4 to 6 tests. For each
Circular cylindrical specimens were fabri- experiment, the torque-twist curve was recorded
cated from 6, 8, or 10 layers of prepreg wrapped (load-cell torque versus cross-head twist) and
around an aluminum mandrel. The prepreg con- buckling was detected by a sharp decrease in
sists of Union Carbide's T-300 graphite fibers this curve. The bucklinq pattern consists of
in Fiberite's 948AI low temperature curing epoxy. helical waves winding around the cylinder.
These helical waves were equally spaced over the
Speciemns were vacuum bagged and cured in an
autoclave under a 94psi pressure. The curing
circumference of the shaft and the number of
circumferential waves n fs also reported fn
cycle consists of a 30 minutes hold at the resin
flow temperature ( 130'F), followed by a two hour
Table 2. No fiber or matrix failure was appar-
sure at 250'F. Table I sumnarizes the lay-up
ent at buckling; in fact, some specimens were
(starting from the innermost ply, 0' is along tested repeatedly up to buckling and nearly
the axis of the shaft and positive angles are identical buckling loads were obtained each time.
defined in Fig. I), the nman radius R. the
Since the tests were run under stroke control it
was possible to trace the behavior of the samples
length L. and the number of specinmns tested in well past the buckling load. however. the applied
each group. Static torsional tests were conduc-
ted in a MTS servohydraulic testing machine at a load continuously decreased after buckling. In
constant twist rate of one degree per minute. other words. no post-buckling strength was ob-
The resulting shear strain rate was typically 40
served. and had the experiment been run under
torque control. catastrophic collapse of the
microstrain/sec. End attachments consist of tube would have occurred at buckling. Under
aluminum plugs which fit tightly inside both
ends of each tube. The specimen is bolted onto
stroke control as in the present experiment,
material failure and a total collapse of the
these plugs by means of eight equally spaced tube only occurred for very large twists. equal
bolts and washers are used to distribute the to several times the twist at buckling.
clamping pressure. To avoid stress concentra-
tion and prenmture failing at the bolting connec- Table 2 also lists the theoretical predic-
tion the tubes were reinforced with s45' layers tions for the torsional buckling of cylinders
over a 3cm zone at both ends. In these reinfor- with clamped and simply supported edges, accord-
ced zones the wall thickness was doubled.
ing to the model described in Reference 6.
The stiffness properties of this material Theoretical predictions are found in good agree-
ment with experimental measurenmnts ~ except for
were obtained for tensile
coupons in an indepen-
lay-up L65 where theory overpredicts the buckling
dent effort [5] as EL 134 GPa, ET 8. 5 6Pa. load by about 405. In all cases the experismn-
GLT GI3 = 4. 6 6Pa ~ 623 4 GPa. vLT ~ 0.29.
tally observed number of circumferential waves
and the ply thickness is 133.4 ID 6 m. was equal to the predicted number. In the ex-
perimental set-up the boundary condition was
neither clanmed nor simply supported. however
theoretical prediction for simply supported edges
are in slightly better agreement with experiment-
al results (about 45 average error versus 6% for
clamped edges). Theoretical predictions were in
general slightly higher than measurmaents.

It is interesting to
note that lay-ups Lgl
and L62 have the same number of plies and ply
orientations. but different stacking sequences.
In fact Lgl and L62 have identical inplane stiff-
ness matrices A [7], identical bending stiffness
amtrices D, and their coupling stiffness matrices
8 are of identical magnitude but opposite signs.
The 8 matrix appears to have a destabilizing ef-
fect in the first case whereas in the latter case
the change of sign has a stabilizing effect.
This results in a 30% increase in buckling stress
from 56 to 75 Npa. The sane observation is true
for L67 and LBB where the stacking sequence is
about 805 from 56 to 105 Npa (in this case the
matrices 8 and D are different for th'e two lay-
Ups
Consider now L62 and L63, the results for
these two lay-ups were actually in opposite di-
mv. I myl con»»t\w ~ w» w»»try w u» u»u. rections. For consistency, all the results in

1092
Table 2 are for positive torques, as defined in vfrg1n specfnmns. Tubes with two diametrically
Ffg. l.A negative torque applied on L¹2 was opposed holes of 12.7 sm were also tested.
is interesting to note that nearly identical
It
treated as a posftive torque on L¹3, where L¹3
has the sane orientation angles as L¹2 but oppo- bucklfng loads were obtained for one or two
sfte signs (Table I).
The dfrectfon of the holes: 485 versus 470 N. m (or 3. IS decrease) for
L¹l, and 550 vs. 530 N. m (3.7% decrease) for L¹2.
applied torque appears to be very important and
the buckling stresses show an 805 dffference Here again the presence of two holes did not seem
from 40 to 75 Npa. Here again th1s effect can to affect the overall buckling pattern and no
be related to the coupling stiffness matrix B. interaction between the holes was apparent.
Finally. the case of lay-up L¹5 calls for Specimens with a 25. 4 nm diameter exhibited
some addftfonal
cosments. In two references [8. a completely different failure smde: the pre-
9] Lagace proposed a Ply Buckling Nodal perti- vfously described buckling pattern was not ob-
nent to the buckling analysis of lmninated cy- . served, but instead. sudden massive fa11ure oc-
linders under compressive axial load. In this curred around the hole involving through-the-
model the outenmst ply of the tube is assumed thfckness cracks and extensive delamination.
to buckle prematurely because of fnftial imper- It fs important to note that this failure mode
fections in the manufacturfng (such as voids, dfd not take place untfl the daamge size reached
eccentricities, etc. ) and because it fs less approximately one third of the shaft's diameter.
constrafned than inner plies. Hence, the load Nearly identical failure loads were obtained for
carrying capability of th1s buckled outer ply L¹l and L¹2 as should be expected since both
can be neglected, and the final buckling load of configurations only differ in their stack1ng se-
the tube can be obtained by cons1dering the quence which has 11ttle effect on material
fnner plies only. If this nodal is applied to failure in this casa.
the present tors1onal buckling problem. the
cr1tical torque for L¹5 could be obtained con- Finally two additional tests were perforsmd
sidering the lay-up 30, -30, 30, -30, 30'. For on specimens L¹l and L¹2 with delamfnations.
this configuration the calculated buckling loads The purpose of these tests fs to assess the ef-
are 408 and 390 N. m for the clamped and sfmply fect of a blunt object impact on the shaft that
supported cases respectfvely, which appears to could create delam1nation without a through-the-
be fn close agreement with the experfmental value thfckness crack. A delamination was obta1ned by
of 390 N. m. However, a vfsual examinatfon of the fnsertfng a 25. 4 nm diameter Teflon disk between
L¹5 specimens after buckling d1d not reveal any piles durfng the manufacturing process. A first
evidence of outer ply bucklfng and the failure test was conducted with one delamination between
mode of these tubes dfd not appear d1fferent the outermost ply and the rest of the laminate,
from that of any other specimen type. Applica- whereas in a second test three delamfnations
tion of the Ply Bucking liodel to the other lay- were present between the three outermost plies.
ups of this study yielded very poor correlation Torsional buckling was the observed failure
with experimental results. Hence, this model mode and the buckling pattern was again unaffec-
does not seem to be appropriate for mode11ng the ted by damage. Once overall buckling occurred,
torsional buckling behavior of tubes. and the the outer ply buckled out at the location of the
good correlation obtained with L¹5 might be for- Teflon fnsert. No delamination growth was ob-
tuitous. served. and the buckling load was reduced by
about 7 or 85 as compared to virgin samples.
Residual Stren th of Shafts with Drilled Holes
Ballistic Tests on Loaded Shafts
Layups L¹l and L¹2 were selected for this
first phase of assesssmnt of the damage toler- In this second phase of testing, actual ba-
ance characteristics of Graphite/Epoxy shafts. llistfc tests were conducted in the firing range
These two configurations were selected as they of the Hatervliet Arsenal, Natervl1et, N. Y. The
appear to be realistic candidates for helicopter N-I rifle (a 30 caliber weapon) was selected for
dr1ve shafts applications [1,2]. Circular the test. A fixture was built to allow remote
holes were drilled in the specimens wfth a car- triggering of the weapon, and fts wefght ( 180
bide tip drill and a wooden plug was inserted lbs. ) was such that the recoil would not alter
inside the tube during the drilling operation. the afm, cons1derably easing the testing proce-
Holes of four different s1zes were considered: dure. A second test fixture was bu11t to tor-
6. 35. 12.7, 19.05, and 25. 4 mn in diameter. The sionally preload the shafts. The aluminum plug
residual strength of these damaged spec1mens was at one end of the shaft was fixed. whereas the
measured by means of a statfc torsional test other end plug was fitted through a bearing and
conducted at a constant twist rate. Table 3 connected to a cantilevered arm. Ne1ghts could
sumnarizes the failure loads for the various be adjusted at various positions along this arm
configurations (the average of our tests is re- to obtain the desired preload torque.
ported for each category. as well as the coeffi-
cient of variat1on). The testing proceeded as follows: for a
given danmge size. the preload level was pro-
For all hole sizes, except 25. 4 mn, the ob- gressively increased until the 1mpact fracture
served failure mode remains torsional buckling was obta1ned. Those specimens that did not fail
and was detected by a sharp decrease in the tor- upon impact were removed from the f1xture, and
que-twist curve. The typical buckling wave their residual torsional strength was measured
pattern remained seemingly unaffected by damage, in a subsequent static tors1onal test. Lay-up
however the bucklinq load was lower than for L¹2 was selected for all tests.

1093
Figure 3 sumaarizes the results in a nor- F1nally alT the specfamns that did not frac-
malized preload versus damage size plot. All ture impact were subjected to a static tor-
upon
results were normalized by the buckling load of s1onal test to measure their residual torsional
undanmged specfsmns. For the 0.30 inch dasmge strength. This data is su)maarfxed fn Fig. 4.
size, no impact fracture was observed up to 90% Once aga1n the res1doal strength of shaft with
preload. Both entrance surfaces of the bullet drilled holes is plotted for cumparfson. It is
were relatively clean. whereas the exit surfaces important to note that the drilled and ballist1c
showed extensive delamination of the cute)most damage curves exhibit very similar trends.
ply. The nomfnal on the abscissa of
damage size
Fig. 2 corresponds to the
nominal bullet dfame-
ter. A 0.45 1nch damage was obtained w1th two
overlapping shots. At 806 preload one specismn
fractured upon 1mpact, another one did not.
eee
esaaa
~ ewp
0.60 inch damage was obtained wfth three over-
lapping shots. and at &8% preload one spec1men
fractured upon 104)act. another one d1d not. The
sol1d 11ne fn F10. 3 joins the points of maximum
allowable preload to avoid impact fracture. The 00
residual strength of shafts with dr111ed holes es
is plotted on the same f1gure. for comparison. ill
Even though the number of tests performed 1n
t&1s study 1s not sufffcient to cla1m statisti- 0.0 o.~ 0.~ I.a 1.~
aeaeae sees« ehe 0«l
cal significance, the trends exhfbited 1n Ffg. 3
clearly show that 6raphite/Epoxy shafts possess
significant damage tolerance. fit. ~ &«asllael sastoast ttaeeth 00.
&&leal saaast stae (la)

Conclusions

The fo1lowfng conclusions can be drawn from


this study:
0
I) 6ood agreement is found between measured tor-
s1onal buckling loads and the theoret1cal pre-
d1ctfons.
2) Torque d1rectfon and stacking sequence can
drastically affect buck11ng loads (up to 80% was

e—
e L Nl
observed experimentally).
L N2 3) The dominant failure mechanism for shaft with
L Nl alth tie le)ts drilled or ballistic damage is torsional buck-
~ L 'N2 slth alai I«usa ling. Hatarfal failure did not occur until the
e Oelsmasmos damage s1te reached approx10mtely ono-third of
the shaft's d1asmter.
0.(m 0.00 10.(m 15.00 20.00 20.00
Onf)AGE SILE/HOLE OIAftETER (ss) 4) 6raph1te/Epoxy shafts exhibit significant
tolerance to ball1st1c damage. A direct hft
with a 30 caliber weapon reduces the allowable
fte. 1 seat&el atlaesta of asefta «Ith I&les ael torque by about 108 only. A 0.&0 1nch ballistic
eels«hell&&. daamge reduces it by about 306.

5) The res1dual strength of ball fst1cally fmpac-


ted shafts appears to be slightly lower than
0.0
that of shafts with drilled holes of comparable
el&0 ace« s1ze. however the d1fference is not drastfc.
~ I ~ «aae saa Hence. meaningful information about damage
tolerance can be obtained fr(ms measurements of
the resfdual sttmngth of shafts wfth drilled
holes, a simple fnexpensive test.
~ll 10 I
0.0 This research was sponsored by the Arsqf Re-
0 0.S I .0 search Office. under 6rant DAA6 29-82-k-0093.
Ieasel l&aeol 0&ass) Dr. Robert $1ngleton fs contract lm)niter.

ms. l ealsot taao &01000 m.


&elm) eaaaea Slm I la)

1094
tmea (Il«I4» ata mmm« lll) m4 «I»eel toim Nm»w
References
tmm mtlaltl lal I
0 trna . » . mao«tow 1. Lfm, J.N. :
and Darlow, H. S. "Optimal Sizing
of Composite Power Transmission Shafting".
lm 4, N, .ts, ls, »,ls OAN 41.tl J. Amer.
1986,
Relic. Soc. , Vol. 31,
75-83.
No. I, Jan.
lm «.». .Is.-«, -is, ls e.»e «.4 pp

W AL N, N, «AI, N . 4.W «.Il 2. Lfm, J.N. : "The Optimal Design of Composite


. .»,N,
ll, ll, ll, -N 4.lll Il.ll Drive
stitute,
Shafts". Rensselaer Polytechnic
Raster Thesis, July 1984.
In-
al -el N -al al N 4.l\4 Il.tt
IN «, A4, 4, -«, IO, AS e.lse n, il 3. Bauchau, O. A. : "Optimal Designof High Speed
ill 4 Is ««A444 e.lse si.tl Rotatfng Graphite/Epoxy Shafts". J. Comp.
lee e,ts, l, -ie, e,«. -m flat. . Vol. 17. Narch 1983. pp 170-181.
4 44 le, ««4 «, Ae 4 W Is.'4 4. Bauchau, O. A.: "Design, Manufacturing, and
Testing of High Speed Rotating Graphite/
Epoxy Shafts". Nassachusetts Institute of
Technology, Ph. D. thesis, June 1981.
wte
Hu, G. S, : "Improvements
~ Oeesatw sl eeeeaee lactea»alas 0« ta. a), as«4am
5. fn the Strength Pre-
«me»Is O.e.ts), aal
NWN Ils IN« em saa«tmt W««W I diction of Notched Conposftes". Rensselaer
Polytechnic Institute, Ph. D. Thesis, June
1985.
a«easel »amati 6. Bauchau, O. A. , Krafchack ~ T. N. and Hayes.
em ta.e. ) Ie a e)aam allee J.F. ( 'Torsional Buckling Analysis and
Damage Tolerance of Graphite/Epoxy Shafts".
Journal of Composfte Naterfals, Vol. 22,
IN 0» II.I») N t«(IN) (I m)
tm No. 3. Narch 1988, pp 258-270.
SN W I ~ .«\ '» ~ tm IW) lit (~ .le)
Nl Na (L4« 44 ~ I» (~ 4) W (4 ll) 7. Tsaf, S, N. and Hahn, H. T. : "Introduction to
lel Nl N. m) Il ~ (l.m) W (l, m)
Composite tfaterfals". Technmsfc Publ. Co. ,
1980.
IN Ial (e, m) N I Sla (W) W (lis)
~N( ~ N) N W(te) 4» (N) 8. Lagace. P.A. : IStatfc Tensfle Fracture of
IN Sil (N.e) N 4 ml I~ Nl me (A.N) Graphite/Epoxye. massachusetts Institute of
lls ls» Ie.m) tes
Technology, Ph. g. Thesis. April 1982.
INI (AAI) lml ( ls)
lel tl44 (le.e) Il) I ««(4) $4ll ( ~ m) . 9. Vfzzfni .
A. J, and Lagace P. A. "The Role of
Ply Buckling in the Conpressive Failure of
:
Graphite/Epoxy Tubes". AIAA Journal, Vol.
23 ' No. 11, Nov. 1985. pp 1791-1797.
t»le 0 lease llaa. earn«aae et mate« O.e.ls). ml a«ac

~»4 «leetat ~elle» a»4 (4.~ )


'
. Ie«lae aw

~N Is,ml )elallel
W (4,ll) ~aaltal
~0\ laNs NN. ) «4 (S.N) ~melan
aat. ls 0» (Lm) I»a)les
4W.I w (~ .m) ~«Is(el lallw

~N (IAI) ~«alias
Nl (4.«) ~eallee
»OL I Nl (t,m) saeatm
tm Isles mes) Nl (4.«) ~el lies
a«sea IIAI)
w (i.m) tstlme
aa mtw am. t ~«IOLI«)' '«as)tel
laa calm. »4»A «a (4,ls) tall)»a

1095

You might also like