r

00
USMVLABS TECHNICAL REPORT 65-60
00
I
ft)
DYNAMIC ELASTIC, DAMPING.
■. AND FATIGUE CHARACTERISTICS
OF SANDWICH STRUCTURE

1 By
w«n« M. NordlypoKPBivedAi. ci'iENUPifANO
W. C. Critman WCHWCAL WTOMlA'notl
ChorU» W. B«r "^"W [ «lororuhe.
*.3.$<): * 0 73"' ?¥* *4:
October 1965

I. I. All! AVUTdN IATEIIEL LAIIMTIIIES

PUT EISTIS. VllfiINU
CONTRACT DA 44-177.AMC-164(T)
UNIVERSITY OF OKLAHOMA RESEARCH INSTITUTE
.y
... J

^»
BLANK PAGES
IN THIS
DOCUMENT
WERE NOT
FILMED
 DEPARTMENT OF THE ARMY
U • ARM AVIATION MATERIEL LABORATORIES
PORT EUST«. VIROI NA tSCOA

This research effort was carried out under Contract DA-4A-

I77-AMC-164(1) by the University of Oklahoma Research Insti-

tute.

The report has been reviewed by the US Army Aviation Materiel

Laboratories and Is considered to be technically sound.

The conclusions made by the contractor are considered by

this Command to be valid.

 TASK 1P121401A14176
CONTRACT DA 44-177-AMC-164(T)
USAAVLABS Technical Report 65-60
October 1965

DYNAMIC ELASTIC, DAMPING,

AND FATIGUE CHARACTERISTICS
OF FIBERGLASS-REINFORCED SANDWICH STRUCTURE

Final Report

by
Gene M. Nordby
W. C. Crisman
Charles W. Bert

Prepared by
University of Oklahoma Research Institute
Norman, Oklahoma

For
U. S. ARMY AVIATION MATERIEL LABORATORIES
FORT FUSTIS, VIRGINIA
 ABSTRACT

Research was conducced to determine the basic dynamic properties of fiber-

glass-reinforced plastic (FRP) sandwich structure suitable for use as a
primary airframe structural material. The research program was carried out
in two separate parts: (A) determining dynamic moduli and damping, and (B)
determining fatigue behavior. In each part, two types of hexagonal-cell
honeycomb core materials were investigated: 5052 aluminum foil and HRP
(heat resistant phenolic) fiberglass. The selection of the sandwich con-
figuration and fabrication procedures was based upon previous research
carried out at the University of Oklahoma Research Institute for USAAVLABS
(USATRECOM TR 64-37 and USAAVLABS TK 65-66).

ill
 PREFACE

This report was prepared by Che University of Oklahoma Research Institute

under Phase II of U. S. Army Transportation Research Conmand (USATRECOM)
Contract DA 44-177-AMC-I64(T). Phase I, which pertains to fabrlcitlon
Is reported separately. The research effort is a continuation of the work
done in two other USATRECOM contracts which resulted in the following re-
ports: Research in the Field of Fiberglass-Reinforced Sandwich Structure
for Alrframe Use and Strength Properties and Relationships Associated With
Various Types of Fiberglass Reinforced Facing Sandwich Structure. The pre-
sent report contains the test results, conclusions, and recomnendatlons
for research conducted on dynamic properties of fiberglass-reinforced sand-
wich structure, Including dynamic elastic moduli, damping characteristics,
and flexural fatigue propetties, during the perio of April 9, 1964, to
November 30, 1964.

This report was written by Dr. Gene M. Nordby, project director and Dean of
the College of Engineering at the University of Oklahoma; Mr. W. C. Crisman,
project engineer; and Dr. Charles U. Bert, research engineer and Associate
Professor of Aerospace and Mechanical Engineering. Mr. Donald Hanson work-
ed as test engineer and statistician.
 CONTEHTS

ABSTRACT ill

PREFACE v

ILLUSTRATIONS lx

TABLES xl

SYMBOLS xill

SIMMRY I

DISCUSSION

A. DYNAMIC MODULI AND DAMPING 2

1. Introduction 2

2. Description of Experiment 2

a. Test SpecinenR 2

b. Test Apparatus 5

c. Experimental Procedure 8

3. Dynamic Moduli 11

a. Literatur« Review: Ssndvich•Beam Vibration. ... 14

b. Literature Review: Timoshenko Beam Theory .... 16

c. New Analysis 17

d. Sample Calculation 20

e. Results and Evaluation 24

4. Damping 24

a. Measures of Vibration Damping 28

b. Literature Review: Vibrating Damping of Sandwich

Structure 31

c. Sample Calculation 34

vii
 ./■tHHtf*

d. Results «nd Evaluation 36

5. Conclusions and Reconmendations 40

B. FATIGUE TESTS 42

1. Introduction 42

2. Dascriptlon of Experiment 43

a. Test Specimens 43

b. Fatigue Testing Equipment 45

c. Experimental Procedure SO

3. Test Rer Its and Evaluation 56

a. S-N Curve and Discussion of FaCigue Data 56

b. Description of Types of Fatigue Failure 59

c. Comparison of Results with Past Research 62

4. Conclusions and Recomnendstions 65

REFERENCES 67

DISTRIBUTION 73

APPENDIX Frequency and Mode-Shape Analysis by Timoshenko Beam

Theory 75

vlii

■
.

ILLUSTRATIONS

Figur« Page

1 Test Apparatus for Measurement of Dynamic Properties. . . 6

2 Electronic Equipment Employed in Measurement of Dynamic

Properties

Wiring Diagram and Instnmentation Used in the Measure-

ment of Dynamic Properties of FRP Facing-Honeycomb Core
Sandwich

4 Detail View of the Test Set-Up for Measurement of Dynamic

Properties

5 The Effect of Beam Length on Lowest Natural Frequency

for Sandwich Beams with FRP Facings and Honeycomb Cores
of Aluminum and HRP Fiberglass 26

The Effect of Beam Length on Location of Node at Lowest

Natural Frequency for Sandwich Beams with FRP Facings
and Honeycomb Cores 27

Decay of a Damped Free Vibration Illustrating Quantities

Related to Definition and Practical Calculation of the
Logarithmic Decrement 29

8 Typical Polaroid Photo Used for Damping Measurements. . . 35

9 Typical Plot of the Decay of Maximum Facing Strain. . . . 35

10 The Relation Between Logarithnic Decrement and Frequency

of Vibration for FRP Facing Sandwich Employing Honeycomb
Cores 37

11 The Variation of Logsrithmic Decrement with Maximum

Facing Stress at Various Frequencies for FRP Facing Sand-
wich Employing Honeycomb Cores 38

12 Fatigue Specimens 44

13 General View of Fatigue Testing Apparatus 46

14 Detail View of Fatigue Testing Device Mounted on MB C-10E

Exciter 48

15 Equipment for Statically Loading Fatigue Specimens. . . . 51

ix
Fltuf Pagt

16 Typical Variation of Specimen Deflection with Time

During Fatigue Tasting 55

17 Pitlgue S-N Curve for FRP Facing and Honeycomb Cor«

Sandwich Structure Subjected to Pure Flexure Loading. . . 58

18 Typical Fatigue Failaret Illuetrated by the Family of

Specimens Tested at 14,000 pal 60

i
 TABLES

TabU ?age

1 Properties of Sandwich Comtitutntt, 4

2 Logarithmic Decrement and Node Locations, Aluminum

Honeycomb Core Sandwich

Logarithmic Decrement and Node Locations, Flberglaee

Honeycomb Core Sandwich

4 Comparison of Measured and Calculated Values of Lowest

Natural Frequency and Node Location 5

5 Shear Damping in Honeycomb-Core Materials, 39

6 Fatigue Test Data - Aluminum Core 7
7 Fatigue Test Data - Fiberglass (HRP) Core. 7

xi
 SYMBOLS

anpllCude

•h««r area (i.e., crosa-aectlonal area of core)

bending flexibility parameter (m&l2 /RTET)

beam cross-sectlonal width

core thickness

CD materUl-damping coefficient defined in Equation (40)

Cp ..., C^ constants of integration in expression for Y

C' ,,,, C' constants of integration in expression for S

specific daaping energy (i.e., internal energy dissipated

per unit volume per cycle)

la/2 effective centroidal distance of core-to-facing adhesive

from centroid of the composite

dynamic elastic modulus of facings

complex modulus defined in Equation (42)

damping modulus

natural frequency of composite beam, cps (MAt/zW}

gravitational constant, 386 In/sec2

dynamic shear modulus of core

total depth of beam (H c •»- 2 tf)

area moment of inertia of the facings about the beam

neutral axis

composite-beam mass moment of inertia per unit length of

beam

transverse-shear coefficient (i.e., shape factor)

beam length

xili
■ aaterlal-danping exponent defined In Equation (40)
M coapoalt«-b«ain mats per unit length
n number of cycles
r rotatory inertia parameter ( S /j/A /L)
■ shear flexibility parameter ( m /EI/GAK /L)
S normal mode form of C^
t time
ta effective thickness of core-to-facing adhesive
t£ thickness of one facing
wt specific weight of core-to-facing adhesive, psi
wc specific weight of core, psi
Wf specific weight of one icing, psi
x position along the bean length
«H position of a node
y total lateral deflection
Y normal mode shape (i.e., normal mode form of y)
*f dimenaionless parameter defined by Equation (A-14)
A dimenaionleaa parameter defined by Equation (A-15)
y (coah bmr • cos bA) I ( Asinh ba( - f sin hß)
S logarithmic decrement
£ position parameter (9 x/L)
€ node poaltion parameter ( m ^/D

^ ifi1 - t2) / K2 * a2)

X *<l/9
A effective denaity of composite ( g M/Bc)

xiv
 CT ■cress amplitude
p slope of beam deflection due to flexure only
6) natural frequency of composite beam, radians/sec
&>f natural frequency for pure shear vibration according
to three-dimensional elasticity theory
CO^ natural frequency for pure thickness-shear vibratioii
according to Timoshenko beam theory

xv
 SUMMARY

The research covered In this report consists of the determination of the

basic dynamic properties of fiberglass-reinforced plastic (FRP) sandwich
structure suitable for use as a primary alrframe structural material. The
research program was carried out in two separate parts: (A) determining
dynamic moduli and damping, and (B) determining fatigue behavior. In each
part, two types of hexagonal-cell honeycomb core materials were investiga-
ed: 5052 aluminum foil and HRP (heat resistant phenolic) fiberglass. The
selection of the sandwich configuration and fabrication procedures was
based upon previous research carried out at the University of Oklahoma Re-
search Institute for USAAVLABS (references 29 and 30).

In the dynamic moduli and damping experiment, beam strips were suspended
at the nodes for the lowest symnetrical mode and excited acoustically at
the frequency corresponding to this mode. Then th«? power was cut off and
the decay in facing strain at the beam center line was recorded versus
time. By means of a new refined vibration analysis of sandwich beams, it
was possible to use the static moduli (the modulus of elasticity of the
facings and the shear modulus of the core) to predict the lowest natural
frequency and corresponding node location. Over the frequency range cover-
ed (300 to 700 cps), the agreement between predicted and measured values
was good, so that it was concluded that the dynamic moduli were the same
as the corresponding static moduli. The damping was determined directly
in terms of the logarithmic decrement for free vibration of the composite.
The log decrement values obtained ranged from 0.023 to 0.049, with the
values for the fiberglass-core beams generally slightly higher than those
for the aluminum-core beams. There was no significant effect of stress
level on damping (up to the maximum stress level covered, 1270 psl), but
there was an effect of frequency, the peak occurring at approximately 500
cps.

In the fatigue experiment, a special specimen and an associated loading

fixture were devised to subject a test length of the specimen to a cyclic
constant bending moment along the test length. Exitation was provided by
an electromagnetic exciter with force control achieved by use of a piezo-
electric force g£ge. No significant difference was found between the
fatigue behavior of the specimens with aluminum cores and those with
fiberglass cores. There was no apparent endurance limit within 10 million
cycles of load as often occurs in metals. However, design for various
cyclic lives can be accomplished in terms of corresponding fatigue strength
values. Thus, the life corresponding to a stress amplitude of 38 per
cent of the ultimate static strength is approximately 105 cycles, while
for a stress of 29 per cent, the life is approximately 107 cycles.
 DISCUSSION

PART A - DYNAMIC MODULI AND DAgglHC

1. Introduction

The objective of the dynamic aoduli and damping phase of the research
was to obtain experimentally the two properties for fiberglass-
reinforced plastic (FRP) sandwich suitable for use in primary air-
craft structure. The dynamic moduli sought were the dynamic modulus
of elasticity of the facings and the dynamic shear modulus of the core.
Logarithmic decrement (defined in section A4a) was selected as the
measure of damping.

In particular, the sandwich construction evaluated consisted of

fiberglass-epoxy facings and honeycomb cores of either aluminum or HRP
fiberglass. Prior to the present work, very little information had
been obtained regarding the dynamic properties of this very promising
structural material. As discussed in the literature reviews that
follow, apparently there are no tests directly comparable to those
reported here.

In these tests recently conducted at the University of Oklahoma

Research Institute (OURI), the dynamic properties were determined at
five first-mode frequencies ranging from 300 to 700 cps, and, in
addition, the logarithmic decrement was determined at five facing
strers-amplitude levels ranging from 270 to 1270 psi at each frequency.
A new, refined vibration analysis for sandwich-type beams was developed
to facilitate the calculation of the dynamic moduli from the precise
frequency and node measurements made in the laboratory. The logarith-
mic decrement was determined by measurement of the decay of free
vibrations with time. A significant aspect of the program is that all
of the required data for a given type sandwich construction and i
given frequency were measured on a single beam specimen.

2. Description of Experiment

The experiment consisted of taking measurements on a series of beam

strips sized in length to resonate as closely as possible to the
following frequencies: 300, 400, 500, 600, and 700 cps. While
suspended in the free-free condition, the exact frequencies and flex-
ural node locations were recorded for sinusoidal exitation by acous-
tical means. Then the exitation was cut off and photographs were
taken of the decaying facing-strain oscillations. The details of
the work are presented on the following pages.

a. Test Specimens

The structural sandwich tested consisted of fiberglass-epoxy

ficing laminates separately bonded to hexagonal-cell honeycomb
cores. The laminates were fabricated In the following manner
using E-glass, 181-weave, fiberglass fabric and EPON 828 epoxy
resin activated by curing agent Z.

The fabric was Impregnated with heated resin In a multi-ply

coating nachlne designed and constructed for USAAVLABS in a previous
program (reference 30) and modified *» a part of Phase I of the
present contract (reference 29). Three layers of fabric were
simultaneously impregnated by the vat and roller system to form
a pre-preg which was allowed to partially cure (B-stage) for 10
hours at room temperature. The material was then cut to facing
size (22 x 28 inches for the aluminum-core panels or 19 x 28
inches for the fiberglass-core panels) and placed in cold storage
until used.

The facing cure was accomplished in a hydraulic press equipped

with electrically heated platens. The pre-preg was thawed, the
thin polyethylene protective films were removed, and the raw lamin-
ate was placed between thin caul sheets using 6-mil polyvinyl
alcohol (PVA) film for a parting agent. The caul aheets and lamin-
ate were placed in the press set at 160 degrees Fahrenheit, pre-
cured without pressure for 8 minutes, and then cured for the re-
mainder of 90 minutes under 30-psi pressure. In Phase I of this
research program (reference 29), these conditions of cure were
observed to produce a good balance between surface smoothness and
strength properties. After cure in the press, the laminates were
then post-cured in a recirculating, hot-air oven for 2 hours at 300
degrees Fahrenheit.

One facing laminate was taken from the sandwich program, cut Into
specimens, and tested for strength properties. The details of
specimen preparation and test procedure are discussed in ref-
erence 29. The test data are presented in Table 1 on the follow-
ing page. The average thickness of the facings was 0.0339 inch
at an average resin content of 35.7 per cent. It was noted that
these laminates possessed excellent surface smoothness, but were
slightly lower In ultimate strength than those previously pro-
duced by OURI (reference 30). These conditions are attributed to
the shorter pre-cure phase employed in the cure cycle--only 50
per cent of the resin gel time.

Two core materials manufactured by Hexcel Products, Inc. were

used in this study of dynamic properties of sandwich. Thesi ware
3/4-inch-thick, 3/16-inch hexagonal cell honeycomb constructed
of (1) 5052 aluminum, I-mil perforated foil (density ■ 3.1 pounds
per cubic foot) and (2) fiberglass (HRP-heat resistant phenolic,
density ■ 4.4 pounds per cubic foot). The shear properties of
these cores \aerm determined by test according to paragraph
5.1.5 of reference 38. The data are tabulated in Table 1.
 *f
1
i » A
* A 9> > & \J> m
< I • • ^* O ■ N> C-
ZI ft »* >» H- ►• O •* t 3 • 3 I
* m ft
» 1 9 •• 8-3 ü?S
A Hi I s
«1 »» R i: A "0 1
►*- A 3
• - *-
5 .
•3. < ooo
A H>n
ag. 3 1 •1 boo a r
K» W F
25 *• w ►-
I 1 A w> « at
n • X A • I i
m i
►- i» 9
►*• |j
3
00 A A **•
3 3
>/> ****** ►-
I Zln a1 0D QB <© w y* w 2« <• W)
0 o — m A — <c <« N) *■ -J
w o ^* W IJ< M
3 O fi t» A
i M M M H M M HUM >» 0» »»
a >^ 7 •«
1 ►- I1 A
n o o o O O O o o o •n
n • j I I I • • i I I I
0 n f» o
1 • c
1
X (A
U
»~ it
gr »»
•8A
5
??
a i
& r>
J
3
•1 « 5 i
1^
X^ A
■
"I
3 ? I
5s o
8 s I
•1
4
3 f» I
» 1
A If
I 5 I I
A
A
1
w *■ ♦- w w *•
: — ►- N) IS» N» N» o o » «O >-< w
A ^- Q W « • « • • • • A
^28 <»- O W
o §8§ i§8
•1 3"
■ »
1 ?
K»
j : i
■
|
1
?
:
I1
I W W C- Kl fsl W
•1 ^ W -J
• • • fc» «J W» *■>•■«
M M M H M M
1 I
A
ii 88§ •i
1
< ■ c
1
x
3
■ i
 The sandwich assembly was accomplished in the press at 3S0 degrees
Fahrenheit under 10-psi pressure for 40 minutes. Minnesota Mining
and Manufacturing Company's film-supported adhesive, AF-110B, was
used to effect the core-to-facing bond. For the aluminum core,
the ribbon was oriented parallel to the facing fabr.c warp direc-
tion; for the HRP-fiberglass core, the ribbon was oriented perpe.i
dicular to the warp direction. This was necessary in the latter
case because the HRP core could be purchased only with a 19-inch
dimension in the ribbon direction.

Prior to sandwich assembly, precise determinations were made of

the core and facing specific weights. These values are also pre-
sented in Table 1. The specimens weighed for these determinations
were all of sandwich panel sit«.

After fabrication, the sandwich panels were cut into 3-inch-wide

strips with their length directions parallc*. to the facing fabric
warp direction and earmarked for either the damping or the fatigue
program by use of a random number table. The strips assigned to
the damping program were further trlnmed to length on a table saw
equipped with an abrasive-wheel blade, the exact lengths of the
beam strips being dictated by the test frequency (see the test
procedure for the details of cutting). The dimensions of the
specimens were measured with the following accuracy: length--
within 1/64 inch; width--within 0.01 inch; thickness--within
0.001 inch. On the average, the specimens were 3.0 inches wide
and 0.815 inch thick. The remaining data are tabulated in Tables
2 and 3 on pages 12 and 13.

b. Test Apparatus

The apparatus which was used to measure the material damping pro-
perties of the FRF facing-honeycomb core structural sandwich con-
sists of an acoustical driving system and a system to detect and
record the damped facing strain oscillations of a beam «trip speci-
men. Photographs of the test set-up and the test equipment are
presented in Figure 1 and 2, respectively. A block diagram of the
entire system is given in Figure 3, along with a list and brief
description of the major components of the electronic equipment.

The test apparatus is in essence a vertical, rectangular fn

with three plywood cross members (Figure 1). The vertical members
of the frame «re steel all-thread screws -hlch permit accurate lo-
cation of the cross members. The two lower cross members support
the speakers between which the beam specimen is suspended. The
upper cross member supports the specimen by way of nylon lines
(0.014-inch diameter) looped around suspension pins pushed into
the core in the center of the sandwich thickness dimension and at
the flexutal nodes. In this configuration the dual speaker systev
acoustically drives the specimen at its mid-span in free-free
flexural vibration.
Figur« 1. T«st Apparatus for Meaturevent of Dynam-
ic Propartias. A, Suapanaion Linas; B, Test Speci-
men; C, Driving Speakers; D, Test Stand; E, Trigger
Switch.

Figure 2. Electronic Equipment Employed In Meas-

urement of Dynamic Properties. A, Audio Oscilla-
tor; B, Frequency Counter; C, Power Amplifier; F,
Oscilloscope; G, Oscilloscope Camera and Camera
Mount; H, Carrier Amplifier; I, Power Supply; J,
Bridge Circuit; K, Carrier Oscillator; L, Battery
for Trigger Circuit.
Tl ^ «»J
• an.
n oo
H» ft c
00 (» (•
X X w
o T •
9 Pi
9 00
PS
1 O H-
oe
|| B
«'s
o o a.
3*
KH (A
>1 1
u I« It
?3 J£ t :s - •
A 3 • M *
13 f
B I »• a
ill ■ *
11
•1
i
O r» o P2 rja il a «^
1 Es
c | — •
§S 2 i N>
i o - ►« ►" <
iia Ml I*
»1 i
s ? 9 751
i» »
a n
IP
o
i £
M- a •wir .oe c
" •- I 1 M ■ to
n •
(A v
II
■ n if
I Iff i t
a
H K n n K
1 A o B a m
ä «I SB "-
! A A
m | •• A
k — •
i 1?
•• <
fl »- w —
m m
ni <* —■ a o
I oS if I Si
%
» o
" Ml O
•1 A n 2
I ■ r
a- ►- ~
wi ^
•!a
^5
! 3 1
I
 The following factors were considered in the choice of the
driving speakers: size, power handling capacity, and frequency
response. The sise of the specimen (3 inches wide) dictated a
small speaker to insure maximum concentration of acoustical
energy on the specimen. A horn-type (siren) speaker was selected
rather than a cone-type because of greater power handling capacity
and acoustical efficiency. The energy transfer to the specimen
was optimised, not only by isin- the dual arrangement (speakers
operating 180 degrees out of phase), but also by placing an
insert (plug) in each horn, which was of optimum depth and diam-
eter of opening (diameter equal to specimen width, see Figure 4).
Ideally, the frequency response of the speakers should be flat
over the operating range, but in view of the time and cost neces-
sary to develop such a speaker, those comnerdaily available were
favored. The particular speaker chosen had the maximum frequency
response in approximately the middle of the operating spectrum
(3C0 to 700 cps).

The response of the specimen wss detected by strain gages located

on the top and bottom facings at the mM-span. The recording
system included an AC bridge circuit, an AC amplifier, and an
oscilloscope with camera attachment. The AC bridge was selected
in order tt ise the reliable, low-cost AC amplifier. The partic-
ular amplifier used was chosen for its low noise characteristics.
A carrier frequency of 6,000 cps was determined to be optimum
for this particular assemblage of equipment. Thr strain gages
were placed in adjacent arms of the bridge relstive to the
impressed voltage (reference 12) for a larger output signal and
for temperature compensation.

The bridge circuit was operated in the unbalanced condition and

the signal was not demodulated fci the following reasons: (1) all
the desired Information was displayed in the modulated signal,
(2) the signal is usuall Jamaged by the demodulation circ itry,
snd (3) calibration .nd hotograph measurement are simplified
(see data reduction procedure).

The Tektronix 503 oscilloscope with camera attachment (Model

C-12) was used to photograph the decaying modulated carrier signal.
A two-pole, double-throw switch in a simple battery circuit, as
shown in Figure 3, was med to simultaneously interrupt the driv-
ing circuit and trigger the oscilloscope.

c. Experimental Procedure

The sandwich strips were first cut to length according to the

desired natural frequencies. This was accomplished by trimming
the length of the beams by small increments until their lowest
natural frequencies approached the desired values. The frequencies
were determined with the aid of powdered molybdenum disulphide.
Figure 4. Detail View of Test Set-Up for Heaaurenent of Dynamic
properties. A, Lower Driving Speaker Showing '-he Insert with the
3-Inch-Diameter Opening (Upper Speaker Has Been Removed); B, Alu-
minum Core Test Specimen; C, Small Sliding Clip Used in Locating
Flexural Node«; D, f rain Gage; E, Strain Gage Leads; F, Suspen-
sion Lines.
The bean» •trips were suspended In a staple cradle or loop In the
vicinity of the node points, and a •mall amount of the black pow-
der was dusted on the upper facing. The driving frequency was
then adjusted until the lowest frequency was found that would
cause the powder to collect into a thin line at the node locations.

After the specimens had been tuned (the lengths established),

the strain gages were bonded to the top and bottom facings.
Eastmcn 910 cement war nttd for this purpose. The gages were
located within 1/64 inch of the center of the facings and within
3 degrees of true alignment.

Number 37 copper wire was used for the strain-gage leads. The
vires were taped to the facings from the strain-gage connection
co the vicinity of the vibration nodes, with small fatigue loops
being provided at the gage connections. The lead wires were
brought from the specimens at the nodes to minimise energy losses
(see Figure 4).

Next, the precise locations of the vibration nodes were determin-

ed. For these and the remaining test special effort was made to
insure that the driving speakers were vertical and that the
specimens were suspended horlzo illy and in the center of the
speaker arrangement.

The specimens were suspended at four points by nylon lines

looped around the pins of the small clips shown in Figure 4.
Observing the peak-out of the strain-gage signal displayed on
the oscilloscope, the previous determination of the specimen
natural frequency was confirmed and the exact value was recorded
to the nearest cycle per second. With the specimen being driven
at this frequency, the black powder was dusted on the upper
facing and the suspencion clips were moved along the edge of the
specimen until the node lines became straight and the suspension
points were located at the ends of these lines. It should be
mentioned that the clips were made from 10-mil-thick brass shim
stock and, hence, were very light, yet stiff enough to maintain
any preset position.

The apparatus was then shut down, and the powder lines were re-
placed with pencil lines. After the pencil lines had been drawn,
three measurements were made of their locations from the ends of
the specimens. The average values are recorded in Tables 2 and 3
on pages 12 and 13, respectively.

Although the clips were convenient for the node determinations,

aluminum pins pushed into the core midway between the facings was
favored aa the means of suspension during the damping tests. It
was possible to insert the pens within 1/32 inch of the nodes.

10

I
 With the pint in place, each specimen was again installed in the
test apparatus. Both before and after each damping test, polaroid
photos were taken of the carrier signal and the carrier signal
modified by the shunting of a calibration resistor across an
active strain gage. Each specimen was then vibrated at maximum
amplitude. The oscilloscope was placed on external trigger and
the camera shutter opened in preparation for the photographing
of the decaying strain-gage signal. The throwing of the trigger
switch followed by the closing of the camera shutter then com-
pleted the operation. The film used for these recordings was
the Polaroid 3,000 speed. Type 47.

Initially, in the preliminary tests, the speakers were placed

as close to the specimens as 1/4 inch to induce the greatest
possible facing stresses. However, it appeared that the proximity
of the speakers was contributing to the damping of the specimens,
and for this reason, a distance of 1 inch was used during the test
program. The distance of 1 inch for speaker-specimen separation
was determined in an extensive side investigation in which separ-
ation distances up to 1.5 inches were compared. The 1-inch dis-
tance was the minumum distance at which no speaker damping effects
wer«* rot ed.

The photographs were measured by projecting them onto a large

screen where an engineering tape graduated in sixteenths could be
used. The average magnification was 12 to 1. The orthogonality
of the projection was assurei1 by comparing th" magnifications
of two accurately measured blocks that had been taped to the
left ani right sides of the oscilloscope screen am', thus, were
photographed with the wave forms. Each carrier and modified
carrier signal was measured six times, and the average value in
each case was used in the calculations. Only one measurement
was made of the peak magnitude of each strain oscillation.
Flg.ie 8, on page 35, is a typical photo of the decaying strain
signal.

The experimental data are presented in Tables 2 and 3, on the

following pages. The calculation procedures and the interpre-
tation of the results are presented in later sections.

3. Dynamic Moduli

Before presenting the new analysis used in this project to determine

the dynamic moduli, it is worthwhile to discuss previous analyses
of sandwich-beam vibration; thlf is done in section a that follows.
Furthermore, since the new analysis is an extension of the so-called
Timoshenko beam theory and since many analyses based on this theory

11
 31
» i — H- H- ►- IS) \y> i i
I ' i <r 1k> fA-
w
• •^J •« •M I»; 3
r> H vC ►— 00
^ 3 90
h*n
a- :r 3? h- ♦■- 00 r 00 • W ►^
a> <t « ft
9 •o r ^
a «t a> K
3
n < 5
P| rr n i
H- »1 r* f
1 -vl Ch [o JO
g2 < t) «> c> y
•ITO 1
Wi A A 90 0 NO « o k> e
►- A § oo <o s> t» A
•-
1 ^ 1 * l
K- a* »<
3 M • < P o3
O r» to i *<
3" « h-
• H C
H. A
W i 0 o o o o M
>r 1 • S*
0 li o b o o b o
K- i-h N) w Ul u> M
• a *» «0 vy» Ut h"
C ^J M ►- v/< ** H
£ Q. r
I 3* ! o O o O o £
n D> o b b b b
1
0 3 N> w w w N» s DO 1 C?
a 00 NJ o» *> *« 1 o
^J oo ^J
« h■3 \*1 ( p
IT i **
S1 m r*- I ^ 9
3-
» 0
ll
o ! o o Wl ■If
x* 1 • •o • • H- 13 31
3 O o o o
M
u> n IM 5
A w w w Ul 1
1 ■vj
2 '-'
B ^- »J Ul u ö
r» » a» K) oo <*« ■ A S0
3- ►«. n 1
(* H< 9
B
I x n
n 0
1 o o o «» |
•o ^ SH
B • o b b b o ^
3 1*^
r» M u, Ul Ul Ul 1 f» j Ic
N* c> o« 00 N* "O 10
>J ISJ 00
n
fT
(T i
H>
*" 1 s H M
1153
u
t 3* 1 S ^
A
0 ^-
"> o o o p 1
ft A o o <*
«
3 O 1 w M t Q ■ Ufl o
Pi
b ?
n H- Ni H- * > m
<r a «- M r
» 1 ■ IQ t^
■o 3 H-
« i <*< IM p
n
c * M
A o •h-
I »1 •o M Q
3 A o
i 1 1 t O
M i «
»> w
^ o 2
n ►*•
i N>
»
| as
r» 3*
3 A
i
a. i **
H. 3 u «« *" Ul P 1
H, o. ,w • m
►*> w Ö* w M 3 ^
■ o> » ? a fi 1
»1 •o *
A 1 I 1 > Mt
|
S*
n
1
H- 9 ^ 1
•w •u> •w •*- • i
■ B t | ►*•
*« >vj 4 9- rsj 1 3
00
■■
■*
! u> V^» V« o UJ 09 {*
1 >-' i|
n
•
n 5
H
■
i 8 1i
 Cl
>—►-»—►— r>j to
00 M ■^ A
H- 9 A
b 00 9 OC O
ST 5» 5-3 • ft
A 0 (» S^ 9*
oo 9"
i a
n o • 5 •»1
k C 1 A ►1 o
r» o- A H <~> A 01
I» i- a • rr
»1 A ■o OP o o o O ha c a
A A 8 a A *
I O ^ 9
£1 M
:r r> < O
0» rr H- a
9 A < ^-
1 A C
O H- ~ A
«< a N>
O o O
w O ha
C NJ a o
9 C
n a a a
i
ft o
o
o o o o 3-
u> N) s
I O v/» 00 V
vD a
o
A
o o
I 3 o
o
o o
1
a o ha
(0 3
rt
o a
rr
> W
CO
o rr
1
» to A
a
I 9-
A
r-
a A
■O <
A A
O
8 o
3
a a
C
z
o
a
re
UJ .P- ui
m
3
| o a 8
9
A t
a
o
a <uJ U> U> ^ Wi 3
3 M
I 00 v*> o a
NJ NJ
$ It $
w
•»1
1
equation for the natural frequency corresponding to any mode.
The primary objection to their analysis, as well as all sub-
sequent energy analyses In this area, Is that they computed the
energies based on a jiode shape assumed to be the same as that of
a simple, homogeneous (Bernoulll-Euler) beam. They should have
adjusted the mode shape to obtain a minimum value of the natural
frequency. Due to the considerable effects of transverse shear
snd rotatory Inertia In a sandwich beam, the adjusted mode shape
would be expected to be quite different than the shape they
assumed. Thus, the frequency value predicted by their theory
would be expected to be higher than the exact value. In their
paper, Klmel, Ravllle, at al published experimental results
only for the natural frequencies associated with the fifth and
higher modes. Agreement between natural frequencies calculated
by using static moduli and the natural frequencies measured
experimentally was fair. Theoretical values were high In two
tests and low In the other two.

The next paper, by Ravllle et al (reference 33), also was an

energy analysis. The assumptions made were similar to those of
Klmel, Ravllle, et al, except that here the facings were assumed
to be Identical and the boundary conditions were those for a
beam clamped at each end. This analysis did not result In an
explicit express on for natural frequency; Instead It resulted
In a frequency equation In the form of a series requiring 36
terms for proper convergence. In all the beams tested, the
theoretical frequency values based on static moduli were higher
than the experimentally measured values, as would be expected
when an energy method Is used and the mode shape Is not adjusted
as previously described.

The next analysis was one by Glaser (reference 13), In which the
rotatory Inertia was neglected. There was no good reason to
neglect this, since It would have resulted In merely a change In
the coefficient of the shear flexibility term In the final
differential equation. The natural frequencies of the first
five modes were measured for sandwich and monolithic beams with
various end conditions. In all cases, the calculated frequency
values were much higher than the experimental ones for the first
mode, quite close for the second and third modes, and -»wer than
the experimental ones for the fourth and fifth modes. Jince this
trend was observed even In the case of the monolithic beam, it
must be concluded that there was a serious error In either the
calculations or the measurements.

In 1961, James (reference 16) made an energy analysis which was

very similar to that of Klmel, Ravllle, et al except that James
considered a free-free beam with identical facings. He considered
only the lowest natural frequency and obtained an explicit equa-
tion for It. Although James obtained good agreement between his

15
 theoretic«! and experimental values, the theoretical values were
higher In eleven out of twelve tests. It should be mentioned
that good agreement in terms of natural frequency alone is not
sufficient, since node locations may »Uli be in considerable
error. Unfortunately, James did not measure the experimental
node locations.

The first direct application of Timoshenko beam vibration

theory to sandwich beams was made by Clary and Leadbetter
(reference 8) in 1963. However, they used the same boundary
conditions used by Kruszewski (reference 2-') in 1949, which
were later shown by others to be incorrect, as discussed in
detail in the next section.

The most recent analysis is one by Dar (reference 10), who used
an approach similar to that of Raville et al and included the
effect of thickmja-normal deformation, which turned out to be
negligible in practical cases.

b. Literature Review; Timoshenko Beam Theory

The elementary Bernoulli-Euler theory for the vibration of a

macroscopically homogeneous beam considers only flexural flexi-
bility and lateral translational inertia. A beam incorporating
in addition the effects of transverse-shear flexibility and
rotatory inertia is usually referred to as a Timoshenko beam,
since the first derivation of the differential equations which
govern the motion of such beams is generally attributed to
Timoshenko (reference 40). The incorporation of the effect of
rotatory inertia alone is generally attributed to Rayleigh
(reference 34) in the 1870's, while Timoshenko added the effect
of shear flexibility in 1921. However, according to Hindiin and
Deresiewics (reference 27), Bressc (reference 6) as early as 1859
wss the iirst to ad'' both rotatory-inertia and shear-flexibility
effects to the simple Bernoulli-Euler beam theory.

Although the so-called Timoshenko beam equations date back many,

many years, apparently the most exact boundary conditions for use
with them were not derived until 1951, by Cengler and Goland
(reference 11), in treating beam impact. The first use of these
boundary conditions in beam vibration analysis is due to Anderson
(reference 2).

The series solution of Anderson is cumbersome to apply. However,

Traill-Nash and Collar (reference 44) obtained a general closed-
form solution of the differential equations and gave the resulting
frequency equations for a variety of simple boundary conditions
(all combinations of fixed, simply supported, end free ends).
Recently, Huang (reference 15), apparently unaware of the work of
Traill-Nash and Collar, presented an equivalent analysis and also
gave equations for the normal mode shapes.

16
 The proper determination of the transverse sheer coefficient
K for even a beam of homogeneous, Isotropie material has also
been the srhject of considsrable controversy. In his original
analysis. Timoshenko (reference 40) used the static value of 2/3
for a rectangular section. In a later paper (reference 41),
Timoshenko used a static value of 0.889, which was based upon
some photoelastic measurements by Filon. In the latest edition
of his vibration book (reference 42, page 335), Timoshenko
recomnends the us» of 0.833 for K, which coincides with the
results of Rels.n^r's static analyses (reference 35) using a
variational principle. As early as 1951. Mindlin (reference 26
and 27) noted that a static value of K is not appropriate, since
the problem is a dynamic one. For a homogeneous, Isotropie beam
of rectangular cross section, he obtained a value of 0.822 for K.

Since ehe values of 0.833 and 0.822 are very cloae, there is
no significant difference between the static and dynamic values
of K for a homogeneous, Isotropie beam.

c. New Analysis

In previous work (reference 8) in which the Tlmoshenko-beam

approach was applied to sandwich beams, no mention was made
of the value used for the shear coefficient (shape factor) K.
Apparently a value in the vicinity of 0.8, which is applicable
only to rectangular-cross-section beams of homogeneous construc-
tion, was used.

In his energy solution for Che vibration of a sandwich beam,

James (reference 16) used an effective value of K given by

K - 1 + (tf/c) (1)

where t£ is the thickness of one facing and c is the core thick-

ness. For the sandwich constructIr-- with which the present
project is corcerned, this amounts to a value of 1.04 for K.

The approach used here to arrive at the correct dynamic value

of K for a sandwich beam is a imply the same approach used by
Mindlin (reference 26 and 27) to determine the dynamic value of
K for a rectangular-cross-section, monolithic beam.

Application of Equation (A-l) in the Appendix to pure thickness-

shear motion requires that y and Its derivatives in Equation (A-l)
be set equal to tero. Thus, Equation (A-l) reduces to simply

<¥/<* 0. (2)

17
Now putting y ■ 0 (no bending deflection) end Equation (2) into
Equation (A-2) in the Appendix results in the following differ-
ential equation in the shear slope:

0+ *¥ - 0. (3)
from Equation (3), the natural frequency U' (rsd/sec) of pure
thickness-shear vibration according to the Tlnoshenko-bean theory
is

4,^ ' (GAK/J)1/2. (4)

However, according to the three-dimensional theory of elasticity,

the natural frequency <Ja for a rectangular cross-sectional
element (of the core) of depth c in pure shear notion is (refer-
ence 43)
CJ% - (fT/c) {Gift )1/2 (5)
where f> is the effective density. Using total mass of the facings
and core gives
(9 - M/Bc (6)
where B is the beam cross-sectional width.
Equating CJ* n in Equation (4) to Cöh in Equation (5), Incorporating
Equation (6), and simplifying:
K - (J/M) (fy/c)2. (7)
It is to be noted that Equation (7) shows that for a sandwich
beam of rectangular cross section, K depends upon the combination
of rotatory mass moment of inertia, mass, and core depth. This
is in contrast to a monolithic beam of rectangular cross section,
for which K is equal to 1f2l\2 Cp»0.822) regardless of the beam
mass or cross-sectional dimensions. However, as a check of
Equation (7), if the specific relationship for J/M which holds
for a monolithic rectangular cross section, namely,
J/M - 1/3 {elf.)2 - c2/12,
is substituted into Equation (7), the result is K > 73r/12,
which checks.
The considerable difference between the shear coefficient of a
sandwich beam and that of a monolithic beam is shown in the next
section, which cov«rt s imple calculation.

18
In a previous research inveitlgation (reference 22), neatured
natural frequency has been used in conjunction with theoretical
analysis to calculate the dynamic shear modulus 0 of the core,
assuming that the dynamic elastic modulus E of the facing Is
equal to the measured value of the static elastic moduluc.
However, this violates the basic principles of good experimenta-
tion, since there Is no more basis for assuming that E has a
dynamic value equal to Its static value than for making the same
assumption for G.

In another study (reference 13), experimental results obtained

for one set of boundary conditions were used to determine the
dynamic value of G, and then this value of G was used to compute
the natural frequencies associated with other sets of boundary
conditions. These latter values were then compared with the
measured natural frequencies for the other conditions to assess
the accuracy of the frequency analysis. The agreements between
theory and experiment were poor perhaps because of two reasons:
(a) the theory was too crude, and (b) different boundary condi-
tions results In different natural frequencies and yet both E
and G may be frequency dependent. One way to overcome this
difficulty would be to use beams of different lengths with the
different boundary conditions and to select these lengths so that
there would always be at least one pair of sets of boundary
conditions lor each frequency. However, this approach still
requires a large number of beams with at least two different
kinds of support conditions.

In order co reduce the number of tests and the number of kinds

of support fixtures required In the present project. It was
desired to arrive at a method of obtaining sufficient Information
from a vibration test of a given beam specimen vibrating at one
frequency (the lowest natural frequency) to be able to determine
the dynamic values of both E and G at tha frequency. The methrd
devised to achieve this objective Is the simple one of measuring
the node location In addition to the lowest natural frequency.
(Node locations are determined quite easily experimentally by
sprinkling fine powder on the specimen. The location to which
the powder moves Is the node location. This was the method used
many years ago to obtain the famous "Chadnal figures" associated
with plates vibrating in various modes.)

To establish the validity of the concept used, It must be shown

that:

(1) The theoretical expressions for natural t equency and

node location depend upon E and G.

(2) For a beam of given geometry and composed of core and

facing materials rf given density, there Is only one

19
 unique combination ol values for E and G which will
result in the measured combination of lowest natural
frequency and the associated node locations.

To meet the first requirement mentioned, it is noted that from

the second of Equations (A-8) in the Appendix, the natural
frequency depends upon E and the parameter b*. To calculate
the node locationa, the right-hand side of Equation (A-26) in
the Appendix must be set equal to sero to obtain the following
equation:

cosh b<tf - AVslnh b^€ +

(l/S) cos bf * - Vsin bfl€ - 0. (8)

The value of € which satisfies Equation (8) is the dimension less
position at which the node occurs, denoted by CM ORX^/L, where
xg is the actual position of the node). From Equation (8), 6^
depends upon the parameter b. Thus, both the natural frequency
and the node location depend upon the value of b. From the
frequency equation, Equation (A-24) in the Appendix, the value
of b depends upon the value of s^, which in turn depends upon
the ratio E/G, as is easily seen from the definition of s2 in
the last of Equations (A~8) in the Appendix. Thus, it has bean
shown that A> depends upon both E snd E/G and that > ; depends
upon only E/G.

From the foregoing, it is seen that although a given node location

xN does not uniquely determine a single combination of values of
E and G, it doea aatablish the necessary condition that whatever
be the values of E and G, they must have a certain ratio E/G.
"•tabllshment of the E/G ratio establishes a unique value for
s2, which for a given mode in turn establishes' a unique value
for b2 by virtue of Equation (A-24) in the Appendix, and thus,
from the definition of b2 in the second of Equations (A-8) in
the Appendix, establishes a unique value of E. Now there can be
only one unique set of values of E and G which meets the require-
ments of a unique value of E and a certain ratio E/G.

d. Sample Calculation

All of the calculations were carried out using the following

average properties and dimensions:

Beam Width: B ■ 3.00 Inches

20
 Facing;

Effective modulus (average of

tension and compression values): E - 3.33 x 106 psl

Specific weight per facing: iOf - 0.00195 lb/In2

Thickness per facing: tf - 0.0339 Inch

Core: Aluminum Fiberglass

Shear Modulus: G - 40,700 psl 6,200 psl

Specific weight: wc - 0.001345 psl 0.00189 psl

Thickness: c ■ 0.747 Inch 0.747 Inch

Adhesive:

Shear modulus: Included In the

measurement of core shear modulus.

Specific weight per layer: 0.000556 psl

Effective thickness per layer: t| - 0.010 Inch

Twice the effective centroldal

distance from centrold of composite: 0.737 Inch

Using the above values, the various parameters required In the

analysis were first calculated. An aluminum-core beam la used
to Illustrate the procedure.

M - (2wf + 2wa + wc) (B/g) - (2 x 0.00195 + 2 x 0.000556 +

0.001345) (3/386)
or

M - 4.94 x 10-5 lb-sec2/ln2

A - Be - 3.00 x 0.747 - 2.24 In2

Btg (C tf)2
I - ! + !ll»l/2 (3.00) (0.0339)
2 6
(0.781)2 - 0.030; In4

21
 J
-12.

001345 0.00195
12 x 386 10.0339 ^ ^ \ 0.747 0.0339
3
0-000556\ (o.747) - 0-00055b (o.737)3J
•)

- 5.70 x 10"6 lb-i«c2

K - (J/M) (tT/c)2 - (5.70 x l0"6/4.94 x 10"5)
(Tr/0.747)2 - 2.04
Th^n for the «lumlmm-rore beam with A length of 22.88 inches, the
diiueniionleas parameters b2, r*, and s2 are:

b2 . 4yr2ML4 f2 . 4(9.87) (4.94 x IP'5) (22.88)4 f2 .

EI
(3.33 x 106) (0.0309)
52.0 (f/100)2 (9)
r2 - J/ML2 - 5.70 x lO"6/ [(4.94 x lO"5) (22.88)2] -
2.20 x lO"4
,2 . _EI__ , (3.33 x 106) (0.0309) . 10.56 x l0-Ut
2 2
GAKL (40.700) (2.24) (2.04) (22.SB)
The transcendental equation to be aolved is Equation (A-24)
developed in the Appendix. It is repeated here for convenience.
2 2 r2
2 (1 - coah »W cos hfl ) + ^ r-r-t
t ' ( " •2)2 +
" (1 - b2r2s2)l/2
(3r2 - s2)J sinh b< sin b^ - 0, (10)
where 1/2

•T - (1/^2) (- (r2 + s^) ♦ fr2 - s2)2 + (4/b2)]l/2 j (11)

1/9
^ - (1//2) f(r2 ♦ s2) ♦ f r2 - s2)2 + (4/b2)] U2
} (12)
The procedure used to solve the above system of equations is as
follows:
(1) Select a trial value of f and calculate a corresponding
value of b2 using Equation (9).

22
 (2) Using this value for b and previously calculated values
for r and s, values are calculated for •< and if by means
of Equations (11) and (12).

(3) The above values of b, r, s,ef , and 4 are used to com-

pute the value of the left-hand side of Equation (10).

(4) The above procedure is repeated until, by iteration,

the magnitude of the left-hand side of Equation (10) is
as close to zero as possible. The nearest integer value
of frequency (cps) to the lowest one which makes the left-
hand side of Equation (10) equal to zero is taken as the
answer (i.e., the calculated value of the lowest natural
frequency).

In the example, the 22.88-inch-length aluminum-core beam, the

final value of the lowest natural frequency f was 308 cps. This
gave bZ - 494, •< - 0.2106, and fl - 0.2136.

Having determined the value of the frequency f, the next step in

the calculations is the calculation of the node location xN. The
procedure for doing this is as follows:

(1) Assume a value of £ and use it to compute the value of

the left-hand side of Equation (8), repeated here for
convenience:

cosh h<6 - AVsinh h*C€ + (1/^) cos hflt -

where frln bflt - 0 (8)

2
.2)/(^2+.2)

y - (cosh b«f - COB ^3 ) / ( Asinh b-< - iTsin hfi)

(2) The above procedure is repeated until, by iteration, the

magnitude of the left-hand side of Equation (8) is as
close to zero as possible. The nearest value of € , to
three significant figures, to the one which makes the
left-hand side of Equation (8) equal to zero is taken as
the dimensionless value € corresponding to the node
location.

(3) Finally the distance of the node location from the end of
the beam is x» ■ L^N'

In the example, € - 0.225 and x - 5.15 inches.

23
 «. Reiulti «n«f Evluatlon

The results <K the experimental neaeurencnts and the calculations

of the lowest natural frequency and node location for each beam
are summarised in Table 4 and in Figures 5 and 6. It is empha-
sised that the calculated values are based on static values of
the appropriate mechanical properties (facing modulus of elas-
ticity and core shear modulus).

As can be seen in Table 4, the greatest discrepancy between measur-

ed and calculated values is only 6.0 per cent. This is quite good
in view of the scatter in facing thickness and in total beam thick-
ness from beam to beam, since the calculations were based on mean
values of these quantities.

It is interesting fo note that all of the methods of calculation

of sandwich-beam natural frequencies utilising an energy approach
are based on the assumption that the node locations are the same
as for a simple Bernoulli-Euler beam, which has no shear flexibi-
lity and no rotatory Inertia. This location is 0.22L, which does
not compare very favorably with the locations determined experi-
mentally in this project; these ranged from 0.225L to 0.235L.

The experimental1" determined location of the node was different

at each end of e h beam; however, the maximum difference from
one end to the other end of the same beam was only 2.9 per cent.
This Is attributed to small variations in the mass and stiffness
properties along the length of the beam (see Tables 2 and 3).
For example, here was a variation in total beam thickness of
approximately 0.005 inch from one end to the other on the beam
which had a 2.9 per cent variation in nodr location.

In view of the good agreement (within experimental accuracy)

between the measured values and those calculated on the basis
of the static ni'dull, it was decided not to go into the iterative
procedure described in section A3c for calculating the dynamic
moduli. In other words, it was concluded that the dynamic elastic
modulus of the fiberglass facings and the dynamic shear moduli of
both the aluminum core and the fiberglass core were the same as
their respective static values. This is in agreement with previous
work by Richter (reference 36) on fiberglass in the form of tubes.

4. Damping

The literature on the internal damping properties of macroscopically

homogeneous materials is voluminous. Although previous investigations
of vibration damping in sandwich type construction apparently have
been limited to a few independent investigations, in each of these a
method was developed to predict the damping behavior of the composite,
assuming knowledge of the damping properties of each of the component
materials.

24
 tt
>
i » r» O
m i ft O
oo »1 1
O I 3 •-• ft
n m
c i
n
it
I
3
a a
n r
n <
3
< K) ^- H- ^- ►-• tsj »— ►— r- ►-> 90 OS
c fi) ^- OD (T vyi U) N> vO ^J iT f- r» ft
n
at C QO ««j 91 Q 9> 00 0D ai K- sO
vO X» ISJ fO SO 00 f- 00 f> H-
CO
B)
C
n
a n
o «
§
m
U) w -^ O- -J
O O O O O O >C vO O O
at 3 NJ^- »— LJ <r ro so ao o v/i ■
I 1
M
< 8
H u> ^ i>n 91 ^J Ul ** ^" (T ^J
c* O ►- ^- NJ f" O P ^- O ^ T3 n
a- sO 0> »O vy » OS ^ ^1 00 tsj w •
ft
«
i
ft
b
o 3 C
U ts> Ul W U* 9 hj ^. U, H- •— ft
3 3
oo U» ~J o* ^ o O 00 00 u> o O
I
C
I
C
n
Ui^U*V«*W ui t^ *» V.J u» n ft
3 1
►- (-> >o *> ►- M Wi P 9 U»
■-• t« O ^ U» 9 o ^ o* 9
I 8.z
V^^>U*U>U> tJ<^U>UtW
3 ^-
O U> 00 sy NJ NJ <J> O »-J *• a. ♦
SJ ^- ^ ^- ro UJ O so ^ ui
3
i
O ft
£> £• Ul l*> W en «> OJ u> w *- 8)
3 ^- «
»- t» ^ (T V«J O C
V/l w« >o u> w ai -o t> ts) f^ 1
ft
a
•»i
1 i i i i
U) OJ NJ t^ O
1 1 1
K) O N> i^ O
1 1 PI i
J ^- rsj oo <y
S
>2 3
FT» a
•g i i i i i I 1 I I I
r- ►- O NJ NJ H- N) ^- w N»
^ ** *-n U) OD ui 00 w v^ (^
 800 1 I

+ LEGEND
u
w V \
700 ^
>- • EXPERIMENTAL
o
UJ •f CALCULATED
3 600
O ^
UJ
\
Ü.
500 ^
i
V^ALUMINUM CORE
\

t 400
HRP-F IBER6L ASS CC ..>
^
uj 300 •s^
*
o
200
12 14 16 18 20 22 24
BEAM LENGTH (inches)

Figure 5. The Effect of Beam Length on Lowest Naturel Fre-

quency for Sandwich Beams with FRF Facings and Honeycomb Cores
of Aluminum and HRP Fiberglass.

26
 LZ
NODE LOCATION. MEASURED FROM END (mchtt)
1 ' 1
X .
T >
O 1
k » ><
a» r *e
m c :•
\ ^ •
\ » « PI
A n
V r O -
o \
c z ^
►* O M» 8 o
n ft.«
H
\
3* «» C
»1 X
fit ft N
r»
3r•
»
o
9 rr ft
00
3B PI

k \
ft \
Is
B M
0 O
3 »fl H»
ft »1
«< ft C0
n ^ ft
NODE LOCATION. MEASURED FROM END (inchtt)
Mlo r1
o*< ft oi * »
o 9
1 H»00 ts>
ft O rr 1
• 1 3"
X •
1 x^ o n»
^x > j< —|
I?
n » OB \ r \0 r-
3* rr 1 \ n m n
r* ^i
c a A
08 O
ft 9 \ P»
\ H 5 J>
X 1*1 ' O —<
o
i V ;
z r
8 ^V
H
X
\
T *
i°
n
| \
w
I» V
O
\
rv) u
Before diecusiing these previous InvestlgAtione and their relation
to the present study, it is well to describe the various ways of
■assuring or specifying vibration damping. This is done in the next
section.
a. Measures of Vibration DampinR
There are four basic ways of measuring or specifying vibration
dsmping (reference 45):
(1) Energy dissipation under steady-state cyclic conditions.
(2) Resonant amplification factor under steady-state cyclic
conditions.
(3) Phase lag between stress and strain under steady-state
cyclic conditions.
(4) Decay of free vibrations.
Energy dissipation under steady-state conditions is usually
determined in terms of energy dissipated per cycle. However,
since the energy dissipated per cycle depends upon the sice,
shape, and stress distribution per cycle, the energy dissipa-
tion is usually specified in terms of specific damping energy,
which is usually considered to be a more basic msterial param-
eter. Specific damping energy is defined as ehe energy dissi-
pated by damping in a unit volume of material at a given stress
level. It is usually specified in units of inch-pound/cubic
inch-cycle.
The resonant amplification factor is defined as the rati.o of
the response to the excitation, with both the response and the
exitatlon specified in terms of the same units. Unfortunately,
the resonant amplification factor is dependent upon the system
configuration as well ss upon the damping property of the
material, so that it is not considered to be s basic material
property.
The stress-strain lag ii usually specified in terms of a phase
angle (hK) in radians. Although this is a dimensionless factor,
like the two preceding measures of damping it can be determined
directly only by measurements made during steady-state forced
vibration. Measurements made under these conditions are likely
to be inaccurate due to losses in the excitation equipment.
The decay of free vibrations is ususlly specified in terms of
the logarithmic decrement $ , which is defined as follows:
S mln (al/al + i) (38)

28
 where ai la the amplitude of a particular quantity (deflection,
atrain, or stress) in a particular cycle i (aee Figure 7) and
ai + i is the amplitude of the succeeding cycle. Since the
logarithmic decrement of nearly all materials ia a very small
quantity, it ia difficult to meaaure a. and »i + i with auf-
ficier accuracy to calculate 6 reliably by using Equation (38)
Thus, it ia customary to assume that 6 does not change signif-
icantly over a few cycles n (aee Figure 7) and to calculate €
from the following equation:

5 - in In -a (39)
i + n

The procedure used for determining damping in thia project waa

to meaaure the decay of maximum atrain level and to apply
Equation (39), using atrain anplitudea.

f
jCYCLC L
CYCLE i ♦ I
CYCLE I +n

MEAN
VALUE TIIIE—•

DEFINITION: •! InfO^/o^i)
PRACTICAL METHOD OF COMPUTATION
S.^lalOt /o I4l )

Figure 7. Decay of a Damped Free Vibration IIlustrating Quan-

tities Related to Definition and Practical Calculation of the
Logarithmic Decrement.

29
Early •xpcrlmenti by Klmball and Lovell (reference 20) on the
duping of a wide variety of materials indicated that the energy
dlsilpation per cycle was proportional to the square of the
vibration-displacement amplitude (and thus, also proportional
to the square of the stress amplitude). Thus, damping of this
nature is often called material damping, hysteretlc damping, or
structural damping. However, in order to avoid confusion in
this report, this type of damping will be referred to as Kimball-
Lcvell damping. More extensive studies conducted by Lasan
(reference 25) have indicated that the specific damping energy D
can be more generally represented by the following relation:

D - CDOT" (40)

where Or is the stress amplitude and CQ and m are temperature

dependent material constants for stress levels up to approxi-
mately 80 per cent of the endurance-limit stress under completely
reversed stresses.

The specific damping energy D is related to the logarithmic decre-

ment 6 by the following approximate relationship:

D - 6<r 2/E (41)

where E is the dynamic elastic modulus. Thus, by comparing
Equations (40) and (41), it is seen that for a material with
Klmball-Love11 damping (m - 2), the logarithmic decrement is
independent of stress amplitude. Furthermore, if the exponent m
increases with stress to values greater than two (as it does for
many metallic materials for stress amplitudes approaching the
endurance-limit stress), £ also Increases with stress.

A final way of specifying damping properties is in terms of the

loss modulus. This concept is applicable only to materials with
Klmball-Love11 damping, i.e., m ■ 2 in Equation (40), since for
such a system the sum of the damping force and the spring force
(force due to material stiffness) can be represented by a single
complex spring constant. In terms of materials properties rather
than specific system performance, this implies a single complex
modulus E,

E - E + iEd, (42)

where E is the dynamic elastic modulus (also known as the storage

modulus and the stiffness modulus), 1 ■ /^T. and the E. is the
damping modulus (also known as the loss modulus). The ratio E./
E is called the loss factor. Since the loss factor is also equal
to the tangent of the phase angle lag between stress and strain,
it is sometimes called the loss tangent.

30
 The loss factor is related to the logarithmic decrement by the
following approximation:

Ed/E » f/tr. (43)

This relationship permits calculation of the damping modulus from

values of the dynamic elastic modulus and the logarithmic decre-
ment, as follows:

Ed - E A/IT. (44)
b. Literature Review: Vibration Damping of Sandwich Structure

Among the earliest analyses of vibration damping of ordinary

(three-layer) sandwich structures were the approximate analyses
of Plass (reference 31) and of Ross et al (reference 37). These
analyses treated sandwich plates, and the approach used was to
determine the total effective stiffness of the composite plate
and then to replace this stiffness by a complex stiffness.

A more exact analysis of damping in sandwich plates was carried

out by Yu (reference 49), uaing a differential-equation-of-
motion approach. His analysis incorporated loss factors asso-
ciated with flexural motion of the facings and with shear motion
of the core; so in effect he assumed that these materials behaved
in damping as Kimball-Lovell materials. Unfortunately, Yu's
theory is rather complicated to apply and so far as known has
never been verified experimentally.

One of the most significant Investigations of damping in sandwich

construction is the analysis published by Keer and Lasan (ref-
erence 19) in 1961. This consisted of both theoretical and
experimental studies. The theoretic«! approach used was to first
carry out a vibration analysis ot u '.earn with free ends and
central point excitation. Frcm this analysis, they determined
the bending-stress distribution in the facings and the shear-
stress distribution throughout the core. Then, assuming that
both the core and facing materials had Kimball-Lovell damping, the
total energy dissipated per cycle was determined by integration
over the entire volume of the materials. The core damping prop-
erties were obtained experimentally using the procedure previously
established by James and Norris (reference 17). Keer and Lazan
also made some calculations and carried out some experiments to
indicate the relative unimportance of external damping due to
aerodynamic friction.

Keer and Lazan measured the total damping (neglecting air friction)
for (a) four different beam configurations with 2024-T3 aluminum-
alloy facings and aluminum honeycomb cores, (b) one with 2024-T3
facings and a balsa core, (c) another with 2024-T3 facings and a

31
paper honeycomb core, and (d) another with facings and core of
glats-flber-reltiforced plastic. There was very good agreement
between theory and experiment for material combinations (a),
(b), and (c), above, at maximum stress levels up to 1,000 to
1,200 psi. Above this stress range, the experimentally measured
damping values began to increase at a gradually increasing slope,
as would be rxpected in view of the similar behavior of homoge-
neous 2024 aluminum alloy. Theoretical damping values were not
determined for the glass-fiber-reinforced plastic beam (d) due
to a lack of sufficient data for the component materials.
However, it is curious to note that unlike the other material
combinations, this material combination had a constant exponent
m of approximately 2.25 over the entire stress range covered.

It is somewhat surprising that such good agreement was obtained by

Ke«:r and Lazan jtween their theoretical and experimental results
in view of the many gross approximations which had to be made
in their vibration analysis. The largest approximation is the
complete neglect of both transverse shear flexibility and rota-
tory inertia. Thus, the mode shape they used is actually that
of a simple Bernoulli-Euler beam.

The next investigation of sandwich-beam vibration damping, which

was apparently carried out independently of the work of Keer
and Lazan, was that of James (reference 16). His theoretical
approach was basically similar to that used by Keer and Laian,
except that James* vibration analysis was considerably more
refined than that of Keer and Lasan Insofar as frequency is
concerned, since James considered both transverse shear flexi-
bility and rotatory inertia, as described previously. However,
since James useH an energy method based en the same mode shape
as the simple Bernoulli-Euler theory, insofar as computation of
total damping is concerned, his analysis dees not appear to rep-
resent much of an improvement over Keer and Lazan's.

The vibration-damping experiments conducted by Janes are quite

different from those of Keer and Laian. By means of a loud-
speaker, he excited the specimen at its lowest natural frequency.
The specimen was suspended at its nodes by me«v:s of small wires
hooked into the core at the neutral axis. The response was
measured near one end of the beam by means of a microphone, and
the logarithmic decrement was calculated by using Equation (39),
above. Apparently the effect of stress on logarithmic decre-
ment $ was not Investigated, since only one value of $ was
reported for each specimen. All of the -aaterial combinations
tested had 1/2-lnch-thlck, 3003-H19 soft-aluminum honeycomb
cores with 3/8-lnch cells and with the ribbon direction either
parallel or perpendicular to the beam axis. Facings included

32

•
two different thl »neises of both soft aluminum and 2024-T3
aluminum alloy. For tome of these material combinations, two
different adhesive-layer thicknesses were tested. The per cent
differences between the calculated and measured /alues of loga-
rithmic decrement ranged from -77 per cent to +134 per cent.
This large discrepancy cannot be explained on the basis of the
frequency analysis, since the per cent difference for the
frequencies ranged from -0.8 per cent to 4-18.4 per cent. It
is Interesting to note that in all six specimens in which core
ribbon direction was in ihe axial direction, the theoretical
value of p was lower than the experimental one, while in five
of the six specimens with the core ribbon direction oriented
transversely th«> converse was true.

Apparently the first vibration-damping measurements on a full-

scale aircraft primary structural component of sandwich-type
construction were those made recently by Hackman et al (refer-
ence 14). They tested two horizontal stabilizers of conventional
aluminum-alloy construction and two of sandwich construction. The
sandwich construction consisted of both facings and core("Ray-
pan" triangular fluted type) of glass-fiber-reinforced plastic
(GFRP).

In the first bending mode, which occurred in the frequency

range of 18-22 cps in all cases, the GFRP stabilizers exhibited
lower logarithmic decrements than the aluminum-alloy stabilizers.
This was also true for the average values of logarithmic decre-
ment for the second bending mode which occurred at 70-77 cps in
the aluminum stabilizers and approximately 57 cps in the GFRP
stabilizers. It should also be noted that there was considerably
greater difference between the logarithmic-decrement values for
the two aluminum stabilizers than between the values for the two
GFRP stabilizers. This can be attributed to differences in rivet
fit, which in turn affects the amount of slip damping.

At first, the fact that the values were lower for GFRP seems
surprising, since it is generally true that nonmetallic materials
exhibit more internal dampi. g than metals. However, it is
believed that the difference here is again due to the effect of
slip damping in the riveted aluminum structure. If the aluminum
stabilizer had been of sandwich-type construction (and thus, not
riveted), the damping in the aluminum structure would have been
considerably lower. Another point is that the frequency ranges
Involved in these tests were lower than those usually encountered
in vibration-damping tests for material evaluation purposes. It
may very well be that the effect of frequency on the $ values of

33
 •luminua and GFRP is luch that at low frtquanclcs .he two 6
veriui frequency curves cross. However, this cannot be ascer-
tained quantitatively from the results reported by Hacknan et al,
since they did not run any damping tests on simple coupons and
since they varieu frequency and node together (i.e., they did
not have a series of specimens in which they could run tests in
the same mode yet at various frequencies).

c. Sample Calculation

Equation (39) on page 29 was followed in calculating the values of

logarithmic decrement. The sample calculation presented in the
subsequent paragraphs will illustrate the procedure employed,
starting with the reading of the photographs.

The photographs obtained in the damping tests resemble the one

shown in the accompanying figure for the fiberglass specimen
excited at 603 cps (Figure 8). After the wave heights were
measured as explained in the experimental procedure, the cali-
bration constants were calculated. This was done by dividing
the value of the simulated strain (93 M in./in.) by the differ-
ence between the heights of the modified carrier (under 92M
in./in. strain simulated by a resistor shunting an active Strain
gage) and the pure carrier. For the test conditions of the
photo in Figure 8, the calculation becomes:

93
20.88.
24.876 - 20.421

Next, a plot was made of facing strain versus wave number for
each photograph. A smooth curve was drawn through the points,
and the strain levels were chosen on which to center the log
decrement calculations. The curve for the previous photograph
is presented in Figure 9.

For convenience, n was taken equal to 10 in Equation (39), and

thus the ratios a^/a^ + n were obtained from the peak strains
for the 5th wave on either side of the reference strain. Using
the 300 4«. in./in. strain level (1,000 psl) for the sample

1
The calibration resistor could either reduce or enlarge the
height of the carrier. It is ehe absolute value of the
difference that is sought.

34
 sc
ft -v X i
M 1 f» H-
H. o » oo
rt a « e
» e c i
r* n m m
o- a
» *< rr
rt M H
H.*' ^
rr H« H>
"S^SS
g««,-
rr 00 a •«
c ^ n 0
1 0) o ^.
» a » »
K- • v< »1
1 H- H-
H« n 3 Q
i o oo a
n -i
us re » no
C r» *
A • H 0
Ö -O • rr
n (» H- o
•< O 3
►* «1
O | ■ 0
3 Oil
O « C K-
« H 1 i
• »do
STRAIN (JI in./in. )
o o
o o
[ r
% iīA
a f"
I« K^
3* (A O
A n » ^
1 •—
■« m
3* ** t
O 3 i-»
rt ■ 0
O rt
o O y ^ — —
rh 3 rti
«
•»3 M ft
K- n 3*
A /
1«
r
r» 5?
s
» ]r
oo S
• < VC
A s
1 0
 calculation, the two values of 356 and 247yu i^./ln. were read
from the curve In Figure 9 and he log decrement was calculated
as follows:

5- (0.10) In (356/247) - (0.10) In 1.441

S 0.0348.

d. Results and Evaluation

The measured values of logarithmic decrement are presented in

Table 2 for the aluminum-honeycomb-core sandwich and in Table 3
for the fiberglass-honeycomb-core ssndwich, on pages 12 and 13,
respectively. A comparison of values between these two tables
for corresponding conditions (nominal frequency and stress level)
indicates the following: In fourteen cases the fiberglass-core
sandwich has an appreciably larger log decrement, in two cases
the tog decrements are approximately the same, and in two cases
the aluminum-core sandwich has the larger decrement. From thio
comparison, it is apparent that, in general, fiberglass-core
sandwich material exhibits greater damping than aluminum-core
sandwich material.

The effect of frequency on the logarithmic decrement is shown in

Figure 10. At first glance, the peak at approximately 500 cps
in each case may seem unusual. However, in similar tests on all-
aluminum honeycomb beams, James (reference 16) obtained values
of log decrement which decreased with increasing frequency
from his lowest frequency (459 cps) on up. Whether they would
have peaked at 400-500 cps is unknown, since he did not make any
measurements br>low 459 cps. Also, an increasing log decrement
versus frequency curve was derived by Richter (reference 36)
in the very low frequency range of 0.012 to 1.67 cps for flexural
vibration tests of fiberglass tubes. In view of these limited
examples, it at least seems plausible that a peak exists in the
log decrement versus frequency curve.

The effect of amplitude of stress st the center of the beam,

which should be the approximate location of the maximum stress,
is shown in Figure 11. Although there is some scatter, there
seams to be little effect of stress amplitude on log decrement
for a given frequency. This is consistent with core damping
data at low amplitudes reported by James (reference 16). As
pointed out in section A4a, a log decrement independent of stress
amplitude corresponds to a damping energy proportional to the
square of the stress amplitude. Thus, the slope of 2.2 for the
log-log plot of damping energy versus stress amplitude reported

36

—— - •
 Li
LOGARITHMIC DECREMENT. 6
o o o o o
o o b o o
f\)0 rv) Ol
§
o
o
9 o
m o
o
c
m
0 o> z o
3 OP o o
rt C -<
1
a H- b ft
n 9 0)
Q> a- a ^ o • ■' ■ " ■ ■- ' ■ ■■ ■■ F—^^■B—^ ^- ■ — ■*
o o
1
ft
< •—
(B O
1 *< ft ft o
H H. 9 o
OQ 9 n pa
(0 00
*< ft
< X o • 0»
» o o
c n < O o
(» »< H. 3
w n O"
3 08
• ft
A rr rr LOGARITHMIC DECREMENT. 6
J* C
s n O ft
o 3 ft o o o o
H 1 3 O 6 b
» ft 6 Ö
o r- is»0 8
•— • i o
ft OP s
n
ft
3S r» w
o \
5a ^
a » O
?l ►*■ o
H. O
3
00 O
ft
n o
3 o
m
o tu
c \ *. -
m oo
x
o
-<
0»
o
o
/
o 7
o
i
 DOS

▼ ▼
004
T i 4
w
K
U
o
OOS
• ^
I * 1
t
1 -r
41
• »*¥I0L FWgQUlNCV
S 002
■ 400
T BOO
0 01 . T •00 -J
2o • 700 cpi

1 1 1
0 200 400 600 SOO 1000 1200 1400
MAXIMUM FACINO STRESS (p»i)

ALUMINUM CORE

T
»-
OOS

▲
i
J1
▼
t
▼
s 0.04

II
(1 •
'
•^ i > •
u 009 ■ (I
(1 [
SYMAfit 1 RC9VCK£v
| 002 il SOO
x ■ 400
▼ 900
* 001
i 4-
•
• 00
700 CM

1 1 1 1
200 400 600 600 lOOO 1200 1400
MAXIMUM FACINO STRESS ( pii)

HRP FIBERGUSS CORE

Figure II. The Varletlor. of Logarithmic Decrement

with Maximum Facing Stress at Various Frequencies
for FRP Facing Sandwich Employing Honevcomb Cores.
The data are from Tables 2 «nd 3.

38
 by Ke*»r and L^-an fi^fcr^nce 10) for a beam with flberglaaa-
epoxy facings and fiberglass honeycomb core indicates that they
had a log decrement which increased with stress amplitude.
Unfortunately it is not possible to convert a damping energy
versus stress amplitude plot into a curve of log decrement ver-
sus stress amplitude, so that a quamitive comparison cannot be
made between Keer and Lazan's results and those reported here.

Although log decrement values cannot be computed for Keer and

Lazan's composites, they can be computed for their core shear
damping data. Table 5 summarizes logarithmic decrements for the
shear-damping tests of Keer and Lazan and of James for various
honeycomb-core materials. It is noted that there are few general
trends. For example, Keer and Lazan had slightly higher damping
for tests in Che ribbon direction, while James obtained consider-
ably higher damping for tests perpendicular to the ribbon direc-
tion. Nevertheless, all of the values reported in the table fall
between 0.0080 and 0.0282, a surprisingly narrow spread. It is
expected that the composite would exhibit considerably higher
damping than the component materials, as exemplified by the work
of James (reference 16). This tends to substantiate the test
results obtained in the present project. Further substantiation
is shown by the close correspondence between the log decrement
values reported here (0.0271 to 0.0486) and those reported by
Hackman et al (reference 14) for a horizontal stabilizer of all-
fiberglass honeycomb-core sandwich construction (0.033 to 0.048).

TABLE ">
SMKAK DAMPING IN HONKVOMB-CORE MATERIALS
Honeycomh Ribbon Cor« Cell Shape Load LoRarithnlc Data
Cort- TMckneii Thlckneit and Sice Orlen- Decreaent Source
Mat.rial UaJ LtaJ UaJ tatlon AiL (Ref. No.)

Aluminum 0.002 Ü ,500 He«. , 0.2S -K 0.0094

Aluminum 0.002 0 soo Hex. , ().."> ♦R 0.OOB0

FiberKlaat - 0 ,283 Hex. t, - - 0.0098

Aluminum 0.U01 Hex.,, 0.125 ♦R D.OUI

Aluminum 0.002 Hex. , 0.375 -R 0.0094

Aluminum 0.00 Hex. , 0.375 ♦R 0.0282

Aluminum 0.003 H«x. , 0.375 •R 0.0108 16

Aluminum 0.003 Has. . 0.375 ■fR 0.0160 16

The syaboli ♦R and -R indicate pirpendtcular and parallel to the ribbon direction,
raapectivcly.

39
5. Conclusloni and Rccomncndatloin

The major conclusions drawn from the research on dynamic moduli and
damping are as follows:

(1) Using the Improved analysis presented In this report, It Is

possible to determine both the facing dynamic modulus of
elasticity and the core dynamic shear modulus from measure-
ments of lowest natural frequency and location of the nodes
on a single beam.

(2) Using this approach In conjunction with experiments In which

the lowest natural frequency and associated node location
were measured, It was found that the dynamic moduli were
Identical tc the respective static moduli, within experimen-
tal error and over the range of frequencies of 300 to 700
cps and stress levels of 270 to 1,270 psl.

(3) Measurements of the logarithmic decrement for decay of free

vibrations gave values ranging from 0.023 to 0.042 for the
fiberglass-epoxy facing sandwich with aluminum honeycomb core,
and from 0.027 to 0.049 for this type of sandwich with fiber-
glass honeycomb core. In general, for the same conditions of
frequency and facing stress, the fiberglass-honeycomb-core
sandwich gave the largest damping values.

In the course of carrying out the research program reported herein,

the following logical extensions of the work on dynamic moduli and
damping have naturally occurred to the present investigators:

(1) The experiments should be extended to higher vibration modes.

For the sake of expediency, the present program was limited
to the first mode; however, there is no reason why it could
not be extended to higher modes. This would provide additional
chec' s of the analytical methods presented here. Also, It
would permit separation of the effects of mode number and
frequency per se. This could easily be accomplished by design-
ing one series of specimens so that several selected frequen-
cies would correspond to at least three different modes on
different specimens and another series to cover the range of
frequencies for a number of selected modes. Cross plotting
would permit separation of effects of mode and frequency from
frequency per se. It would be expected that mode number would
have a major Influence on damping since damping is a direct
function of the stress distribution which in turn depends upon
the modal shape and, thus, the mode number. This would be
quite important in connection with designing fiberglass sand-
wich structures to withstand wideband-frequency excitations.

40
(2) Experiments should be conducted on been specimens with lower
values of lowest natural frequency. This could be accomplish-
ed either by going to free-free beams with a rigid mass attach-
ed at each end and the same maximum beam length limitation as
encountered here or by using free-free beams without attached
masses but of much greater length. The first alternative
should permit attainment of a lowest natural frequency near
90 cps. However, this is still high compared to frequencies
of interest in many airframe applications. Thus, a combina-
tion of the two approaches would probably be required.

(3) A theoretical analysis should be made of the damping in a

sandwich beam using specific damping data for the component
materials and stress distributions based on the refined modal
shapes obtained by the vibration analysis presented in this
report. This would be one of the bases for the optimisation
described below.

(A) Having established a firm basis for the analytical assumptions

and having determined the material properties from the sand-
wich beam work, analyses should be made for vibration of sand-
wich panels and sandwich shells. (Limited analysis, mostly of
an approximate nature, has already been carried out for vibra-
tion of rectangular sandwich panels and cylindrical shells.)

(5) In conjunction with the results of the present research in

regard to dynamic moduli, damping, and fatigue, as well as the
results of items 1 through 4 in this listing, optimum material
combinations and sandwich configurations should be determined
for various basic structural elements subjected to various
types of support and dynamic excitation.

41
 DISCUSSION

PART B - FATIGUE TESTS

I. Introduction

In the fatigue test phase of the investigation there were two objec-
tives: (1) to establish S-N curves for use in designing primary air-
craft structunl components and (2) to provide a comparison of two
types of fiberglass-reinforced sandwich structure, that using the
aluminum honeycomb core and that using the HRP fiberglass core.

Since there are many combinations of fibergl«4S fabric weaves, thick-

ness of laminates for facings, core thicknesses and cell sizes which
may be combined with a multitude of combinations of flexure, tension,
compression and shear loads, it was impossible to investigate every
phase of fatigue. To completely investigate fatigue at this point
for sandwich construction would be unwarranted because of the many
other factors being investigated, here and elsewhere, which will
affect the fatigu« characteristics of sandwich construction. Thus,
this program was aimed at finding preliminary information which
would affect the choice of materials to help optimize other strength
and fabrication characteristics in the total program.

It was, therefore, decided to investigate pure flexure loading only.

The only tests somewhat similar to these reported here were per-
formed by Keer and Lazan (reference 19), who reported four specimens
of a comparable type. The results showed that the fiberglass facing-
fiberglass core combination sandwich construction was inferior to
aluminum facing-aluminum core and aluminum facing-balsa core combin-
ations. In addition, Keer and Lazan1s specimens were vibrated at
resonance, and the specimens were double cantilevers driven from a
center point on an electromagnetic shaker. This produced a combin-
ation of flexure and shear.

The reader is cautioned not to draw sweeping conclusions from fatigue

data now available in the literature because fatigue information is
incomplete and rather specialized. For example, there are numerous
tests on special applications which indicate the superiority of fiber-
glass laminates in actual structures. Typical of these is the inves-
tigation of rotor blades for helicopter use reported by KaaMn Aircraft
Corporation (reference 18). They report that for a composite sand-
wich construction rotor blade, a fatigue strength of three times the
comparable production model using wood was obtained. This added
fatigue strength was achieved with considerable reduction in total
weight of the part. In addition, a multi-ply fiberglass laminate
bom ed sheath over a wood rotor blade substantially improved its
fatigue strength characteristics.

42
 On the other hand, most fatigue research indicates that the fatigue
strengths of fiberglass structure expressed as a per cent of the
ultimate static strength are substantially less than metals. Typical
of this is the report by Bert and Hyler (reference 3), which shows
that a flajs-filiment-wound, epoxy-resin composite had a fatigue
strength at 100,000 cycles of approximately 31 per cent of the ulti-
mate tensile strength as compared to a value of 76 per cent for
titanium alloy at the same number of cycles.

The point is that fatigue strength?, especially those expressed by

ratios or percentages, are not the only criterion on which a material
should be Judged. Careful consideration by the designer of fatigue
strength-to-ultimate static strength ratios, strength-tu-weight
ratios, and the economics of fabrication will decide this issue when
complete data are available. The reduced notch sensitivity, high
strensth-to-weight ratio, directional strength versatility, and
lower tooling costs give fiberglass an advantage in this comparison
in the overall picture. The present project Is thus aimed only
at filling in one facet of the total required information which
will aid the designer in making his decision.

2. Description of Experiment

a. Test Specimens

The test specimens were cut from the panels described in Part A
for the dynamic moduli and damping investigation (section A2a,
P«ge 2).

in the initial fatigue tests, the specimens were cut 3 inches

wide and approximately 15 Inches long using an abrasive-blade
table saw and then ground to final smoothness. This gave a
specimen having 18 cells of core material across its width.
The specimen was clamped syometrically in the fatigue apparatus
(described later) with 4 inches of the specimen in the clamps
at each end, leaving a 7-inch flexure tone in the center portion
of the specimen. In these early tests it was learned that the
specimens failed near the clamps because of stress concentrations
at the edge of the clamp. In addition, there was crushing and
abrasion in the clamp area. On the basis of this preliminary
testing, the final specimen was developed. None of the prelimi-
nary tests were Judged valid and are excluded from the report.

The specimen finally selected for testing is shown in Figure 12.

It retained the overall length of 15 inches, but the center of
the specimen had a narrow section 2 inches wide with 12 core cells
across the width. This narrow throat forced the failure to occur
in the center section. In order to prevent crushing in the clamps,
the core was cut out of the ends of the specimens for a distance

43
 AT ;

T'^

H-t.T«?"

1/4
OftlLL

i"
S- I" 5"

d"

^^-FRP--FACINO
1
OAK BLOCK |i
in nl mil I'lHONEYCOMB CORE 1 HI OAK BLOCK 0815"
T

Figure 12. Fatigue Specimen!. A, Kiln-Dried Oak Blocks To

Prevent Core Crusning; B, Silver Plate Contact for Deflection
Measurement; C, Honeycomb Core Material.

44
 of 4 inch«! and kiln-dried o«k block« war« glued between eh«
facing laa^nataf with an «poxy resin (EPON A-6). All edge» of
Che facing ware ground or hand lapped to proper dlmrniions to
eliminate •cratches and other Inperfeet Ions which sight affect
the fatigue results, .wo holes were drilled In the ends of the
■pecimen for alignment In the fatigue aachlne.

All the speclnens had a snooch glassy finish with very few i»-
perfectlons visible to the eye. Microscopic exaalnatlon, how-
ever, shows scattered smill air Inclusions a few thousandths of
an Inch in diameter. On occasion these were at the surface of
the facing, resul.lng In small pick holes. These •mall imper-
fections were not uniformly distributed In all specimens and prob-
ably provided stress concentrations which had an undetermined
effect on the fatigue life of each specimen. This matter Is
discussed later.

b. Fatigue Testing Equipment

Fatigue Electromagnetic Equlpmei t. The fatigue equipment, was

designed around a Model C-IOE vibration exciter manufactured by
MB Manufacturing Company, Division of Textron, Inc. The exciter
was fitted with a apeclal flexure device, bolted to Its top, which
held the specimen. The general set-up Is shown in Figure 13.

The Model C-IOE exciter (A in the figure) provides a reciprocating

movement by means of an induced electroiragnetlc force. The baaic
internal structure consists of a stationary field coil and a
driver coil attached to the moving table. The table is mounted
in a series of restraining rockers and rubber mounts such that a
linear vertical motion is provided. The exciter has a capacity
of 1200 pounds force for continuous-duty sinusoidal input with
a maximum amplitude of 1 Inch or a maximum velocity of 70 inches
per second. The frequency range is 3 to 3,000 cps.

The C-10E exciter is controlled by a Model T1SI electronic ampli-

fier (B in Figure 13). This consists of a Model T104A automatic
vibration exciter control, a power amplifier, and a DC field supply.
The R104A unit consists of an oscillator, a frequency sweep unit,
a vibration meter, an automatic level control, and an automatic
displacement-acceleration transfer unit. Although there are many
autoratic features to this equipment, manual control am. adjust-
ment were used for these experiments.

The frequency of operation was monitored through a Hewlett-

Packard Company Model 521G frequency counter, accurate to the
nearest cycle per second. The counter was plugged directly into
the T1S1 console unit.

45
Figur« 13. General View of Fatigue Testing Apparatus. A,
C-10E Electromagnetic Exciter; B, T151 Control Console; C,
Fatigue Device Bolted to the C-10E Exciter.

46

I.
Thj fatlgu« de< let w«s •pcclally d«tlgn«d and built fro« aluml-
ma to aounc on top of tht C-IOE excittr. It !• thovn in Pigur«
14 with a ipociatn claapod in place. Basically, this provide« a
aiapla baaa with tha Unk-typ« vartical and-supporta (A in Figur«
14) boltad to a larga ttaal ring which in turn la boltad to tha
excitar fraaa. The load is appliad through two sets of center
linka (B in Figur« 14) boltad to tha aoving table of th« «xclter.
Each link and waa fitted with heavy-duty roller baaringa. Becaur ■
tha four set of links supporting tha apaciaan clamps provide a
statically unatabl« force ayataa, on« aat of rigid linka (no lower
bearings) coupled to a stabilizer bar (C in Figur« 14) was neces-
sary to eliminate sid« sway during operation.

Tha specimen waa clamped in tha mountings (D in Figur« 14) sup-

ported by th« vertical linka. Vartical pina in thaaa mounts
aaaurad proper alignment by fitting into accurately drilled
holaa in tha specimen. On« pin waa extracted aftar specimen
installation in order to eliminate any axial forces on tha speci-
men. Clamping in tha mounts was accomplished by a top plata at-
curad with four cap screws. Each of the top plates waa alao
fitted with four shear pina to inaure that the clamps functioned
aa rigid bodies. The clamps (mounts and top platea) were lined
with 1/32-inch-thick replaceable rubber sheets to prevent abraaion
of the clamped portion of the speclmenn. All cap acrewa were
torqued to 100 Inch-pounds at the beginning of each teat to
provide uniform mounting condltiona for the specimens. As
stated previously, the specimens (H in Figure 14) were 15 inches
long with 4 inches clamped at each end, leaving a 7-inch teat
cone loaded under a conetant-type bending moment diagram (no
shear).

An Endevco Model M2110 piezoelectric dynamic force gage with

the accuracy of tl per cent (E in Figure 14) was mounted in
the rigid set of loading linka to measure one of the reactiona.
The signal from the force gage was fed into a cathode follower
amplifier for impedance matching and thence to a Tektronix oscil-
loscope (D In Figure 13) for read-out. This system waa used to
maintain a conatant maximum force at the specimen loading points
throughout a teat by manual adjustment of the exciter control.

Dynamic deflections were measured by a micrometer gage (F in

Figure 14) mounted on a atanchion which straddled the specimen
at Ita centerline. A small copper spring with a silver tip was
fitted to the end of the micrometer. While the specimen waa
vibrating, th« apring tip was lowered to make contact with a small
silver plate (3/16 inch by 3/16 inch In size, see B in Figure 12)
held on the specimen with an adhesive at the point where deflection
measurements were required.

47
1 0t
Tg 2
ffl
►* as •—
llfB
■ * ^
IH:
3-If
■O O rr H
1 »1 ^ •
H. O f •
b « H- r»
OQ an.
00 ••
SA 0
<• >• Cd A
w ^"V ►*
■o - r1 o
» _ o »
n X *
fSff •"t p* h-
- oo ^
r
 Contact between tht spring tip and the silver plat« on each cycle
of vibration conpleted a •Imple scries circuit, th« signal of
which was displayed on an oscilloscope. This provided a simple
visual SMchanlssi to measure th« dynamic maximum deflections with
an accuracy of 0.001 Inch.

• Dynaalc Strain Equlpmsnt. Th« dynamic strain-gage circuit for

th« fatlgu« tests was Integrated with that used for th« danplrg
measurements described In section A2b. By a sat of switches the
strain measurements could b« mad« on either th« damping specimens
or th« fatigue specimens. As previously stated, he system in-
cluded an AC bridge circuit, an AC amplifier, and an oscilloscope.

A carrier frequency of 1,000 cps was determined to b« th« opti-

mum for the fatlgu« circuitry. Th« signal on th« oscilloscope
thus consisted of th« amplified carrier signal on which was
superimposed the dynamic strain signal from th« specimen. As
for th« damping program, calibration of th« oscilloscope trace
was accomplished by shunting a strain gag« with a resistor sim-
ulating strain.

During a test it was necessary to read the scop« visually whil«

adjusting the amplitude of th« C-10E exciter, and this in Itself
led to significant «rrors because of the small variations pos-
sible on the displayed carrier wave. At the higher specimen
facing stresses of 20,000 to 25,000 psi, reading accuracy could
only b« ±10 per cent. At th« lower stresses, because of smaller
changes in the carrier signal, the error was increased to 120
per cent.

There were also considerable problems with the strain gages.

Often the life of the strain gage under these severe loading
conditions was so short (a few thousand cycles) that there was
insufficient time to relate the strain measurements to the force-
gage signal. A variety of adhesives, types of strain gages, and
techniques of application were tried, but either the gages had a
short life or the adhesive allowed the gage to yield relative to
the facing material. This method was abandoned after testing ten
specimens because of the difficulty encountered and the unreli-
ability of the results. These specimens are not reported herein.

Static Loading Equipment for Fatigue Specimens. As an alternate

solution to dynamic facing strain measurements, dynamic deflec-
tion measurements were taken by the micrometer gage previously
described. The procedure followed consisted of relating the
dynamic deflection to the force-gage reading while the test was
running. Since the force-gage reading could be displayed full

49
 teal« on kh« «cop«, the fore« could bt read wich an accuracy of
-2.5 par cent, a considerable Improvenent over the itrain-gage
Mthod.

In ordar to establish chc relation between the deflection and the

facing st.ain, a special testing device was made to load the fa-
tigue speclncns statically The equipment employed, including the
Instron testing machine, is shown in Figure 15. The configuration
of the apparatus was identical to the fatigue device mounted on the
C-10E exciter except for the fittings required for attachment to
the Instron. Again, each end of the specimens was gripped between
a mount and a top plate, both lined with 1/32-lnch-thlck rubber
pads and screwed together with cap screws, and each torqued to
100 inch-pounds.

Deflections were measured by a Baldwin Model PD-IM differential

transformer type deflectometer and automatically plotted versus
load on the Instron x-y recorder (F in Figure 15), an integral
part of the Instron machine. At the same time, stresses were
read by a Baldwin digital atrain recorder (G in Figure 15). The
load versus deflection curve on the Instron chart was blipped at
the desired points, and thus complete deflection versus facing
strain data were accumulated.

c. Experimental Procedure

In constructing a fatigue S-N curve, the Investl* itor is faced

with a mismatch between the designer's needs and the limitations
of laboratory fatigue equipment. The designer's needs are dic-
tated by the loading requirements on the structure which he trans-
lates into the required stress capability. Thus, the S-N curve
(stress versus number of cycles of load) is the type of informa-
tion required for design.

The investigator finds it difficult to keep a constant stress in

the specimens in the laboratory during the repeated load in cer-
tain cases. He may impose constant repetitive deflections or
constant repetitive strains, but these may not be translated
directly into stresses unless the material properties and geom-
etry of the specimen remain unchanged during the test. This
is not usually a problem in tests involving direct tension or
compression repetitive loads, because with fixed specimen geometry
and the capability of maintaining constant repetitive loads on
the specimens, there will be repetition of a constant maximum
stress even If the modulus of elasticity varies over the duration
of the test.

In the case of this experiment, the specimens were flexure spec-

imens and the methods of maintaining constant stress are indirect.

50
 K
n a ft*m
3* c
ft i-* n
O »1
-mm
a•
m H- ^^
^- • • c
oSÄS
00 Ml'Ö Hl
H> 0 « O
rr H ft »1
M H ■ (ft
»1 II rr
W 3 •»
ft O •• r»
1ft H>
ai H» n n
►* ^- - n
9 ft »-
ft PI »^
o I
O ft (ft Q
r1
n rt c S
a ft TJ Q.
» i ■o >*
1-09
H 00
•fl ft
- Si ij
^11
«>^tft
a0n
§22
I
M H.
a*
9 9^
• 00 >
ft <•
1 f
0 H« O
9 9 H-
ft •• "O
• •
ft tn
H» • ft
9 O
h-1
a H-
It AS known that the modulus E of tht flbtrglatt laminate facing
materials changes with tht number of cycles of stress. In tests
reported herein, the modulus decreased as much as 33 par cant
over tha specimen Ufa. Thus, If constant repetitive deflections
ara imposed on tha specimen, tha stress will dacraasa with tha
dacraaalng modulus E over tha duration of tha test. An alterna-
tive la that of tracking tha strain on tha specimen, but since
convarting strain Information Into stresses is alao dependent
upon tha varying modulus, this could not be obtained without ex-
cessively complicating tha Instrumentation and running a separate
Investigation on modulus variation. Time would not permit this
extra phase in tha present investigation.

This problem was circumvented by a combination of strain, deflec-

tion and load measurements. In tha feasibility portion of the
experiment, two methods ware uaad and evaluated. One method re-
latad dynamic strain-gage measurements to the force-gage readings,
and tha other method related dynamic deflection measurements to
the forca-gaga readings.

In tha strain-gage method, the amplitude of vibration of the

machine waa rapidly increased, thus increasing the deflection
of the fatigue specimen, and at tha same time the dynamic strain
signal on tha oscilloscope was read. This was an axcaadingly
tricky procedure. When the proper strain was achieved, the oscil-
loscope was switched to tha force-gage circuit, and a force
reading taken and then maintained throughout the duration of the
test. Because of the difficulty in obtaining accurate strain
readinga (see second part of section B2b) and also maintaining
■train gages serviceable for a long enough period to make the
switch to tha force Rage, this method was abandoned.

At an alternate solution, one which proved to be much more feasi-

ble and accurate, a dynamic deflection measurement was made while
the amplitude of vibration was rapidly increased. When the
proper deflection was achieved (measured with an accuracy of
0.001), a force-gage reading was taken and then maintained for
the duration of the test. The dynamic deflation measurement
waa previously deteribed, on page 47.

In order to carry out thit procedure, a relationthip between

deflection and facing ttrett had to be established. To do this,
three specimens for each type of sandwich construction (aluminum
or HRP core) were tested statically in the static loading device
previously described (last part of section B2b). The fixture
provided, aa near as possible, identical conditions of clamping
and load application to that which the fatigue equipment would
give.

52

I l
In each test the fpecimen wa« loaded CO approximately 30 per
cent of the ultlnate flexural load to seat the specimen in the
teet fixture and to remove Initial dlecontlnultlea from the de-
flection Lun'e. The specimen was then loaded continuously at a
rate of 0.05 Inch per minute until a strain equivalent to 25,000
pel stress was recorded on the facings. Complete strain, deflec-
tion and load data were taken. The curve "at ie-run four times
for each specimen, and on the final run the specimen was loaded
to failure. The final failure strain w<*s recorded and converted
to ultimate failure stress based on a modulus of elasticity of
3.33 x 10° psl. The modulus was an average value based on ten-
sion and compression tests carried out on a laminate made at the
same time as the sandwich panels (aee aectlon A2a, page 3).

The data from the three specimens for each type of sandwich
construction (a total of 12 stress-deflection curves) were
averaged to establish the deflection-stress relationship.
All the curves were almost Identical, which was an Indication of
the careful fabrication and machining process which was used to
make the specimens. Nevertheless, the averaging of the results
did perhaps serve to eliminate some minor experimental error and
minor difference In specimens. The curves for both types of con-
struction were also nearly Identical. The following equations
were obtained from the curves for use In calculating the dynamic
stresses of the fatigue specimens:

Aluminum Core: Facing Stress (psl) - 7022 x deflection (Inches)

HRP Core: Facing Stress (psl) - 7072 x deflection (Inches).

It must be pointed out that there may be some Indeterminate error

In this process because the dynamic moduli and the static moduli
may not be precisely the same. However, in section A3e it waa
concluded In the damping Investigation that there was no signif-
icant difference between the two moduli for either of the core
jiaterials or the facing laminates at the rates of strain imposed
in the damping vibrations. The ratea of strain were estimated
to be no more than three times as great in the fatigue tests as
in the damping investigation, so that it is believed that the
moduli are not significantly different. Nevertheless, this is
the best relationship that could be established at this point.

A frequency of 40 cps was used for the fatigue test program.

This frequency was chosen on the basis of the exciter's capa-
bility and a reasonable maximum speed for the specimen. At the
highest stress (25,000 psl), a total vertical movement of the
center of the specimen of over 0.7 inch was necessary, and the
specimen could be observed only as a blur while under test. At
this speed, the shortest failure time occurred in 5,743 cyclea, or

53
only 2.4 minutes, while the longest tetti mede were 10 million
cycles st 10,000 pel, or 72 hours of continuous operstlon. The
fstigue machine required continuous monitoring and adjustment
during each test.

The 40-cps opersting frequency was near the resonsnce of the total
mechanical system. The system consisted of two primary masses,
the exciter table (approximately 20 pounds) and the fatigue de-
vice and specimen (approximately 15 pounds). A spring constant
was provided by sn internal spring, the specimen itself, and
two auxiliary springs screwed to the table at the bottom and
into the stanchion crossing above the specimen, which vas in turn
fastened Co the frame of the exciter. The two auxiliary springs,
each with a spring constant of 900 pounds per inch, were necesssry
to adjust the frequency of the system upwards to 40 cps and to
give the system stability during the development of the fatigue
process. Operating the system at resonsnce was favored to avoid
overtaxing the exciter for the long test period, since this proce-
dure would substantially reduce the force requirement.

It is emphasised that the system as a whole operated at reso-

nance. There were no resonant effects on the specimen itself
because its natural frequency would be in the vicinity of 800
cps or more in a simply supported state. Since both deflection
and link-reaction force were measured dynamically, and even though
the flexibility of the specimen increased over the duration of
the test, constant repetitive moment was applied to the ends of
the specimen; thus it may be stated that constant stress was
schlaved in the specimen facings throughout the duration of the
test.

The deflection of the specimen increased as much as 35 per cent

during a test. On a fsw speciu.ns, dynamic deflection measure-
ments were made periodically during the duration of the test.
The test had to be a relatively long one (at least two hours) in
order to record a sufficient number of points to plot a curve.
In a test lasting only s few minutes there were too many required
data to record and too many adjustments to make to take time for
these extra data. Figure 16 shows a typical curve for deflection
increase for a specimen (aluminum core) loaded at a facing stress
of 14,000 psi.

Since only scant deflection increase dsts were taken, no com-

parison of deflection increase (decrease in modulus) can be made
between stress levels or between specimens with the two types of
core material. It was noted, however, that as the stiffness of
the specimen decreased, the resonant frequency of the system
changed and the exciter had to work harder to maintain constsnt
force at the constant 40-cps frequency. This was apparent through

54

1 I
 observationi of the change of plate current required at displaced
on the control console, indicating the change in input power.
This, in fact, became a way of detecting approaching failure of
specimens. When the plate current changed from a constant slow in-
crease to a sharp increase requiring frequent adjustment of the
vibration amplitude and reaction force, failure was imninent.

All the fatigue tests were run at ambient temperr.tures which

ranged between 60 and 82 degrees Fahrenheit. Past research (ref-
erence 4) has shown little temperature effect on fatigue results
for such small temperature variation. The specimens themselves
became hotter than room temperature during the test due to energy
absorption but became only warm to the touch. There seemed to be
no detectable correlation between temperature and fatigue life.

1* 30 —-

•
bJ

2 20 y
o /. STARTING DEFLECTION - 0.198 in.
z
f •4
EFLECTION
5

/
25 50 75 100 125 150
TIME(minutff)

Figure 16. Typical Variation of Specimen Deflection with Time

During Fatigue Testing. This specimen was constructed of HRP
fiberglass core and was loaded 14,000 psi at the rate of 40 cps

55

•
 A total of 47 specimens were tested. Thirteen of these were ex-
perimental specimens used to adjust the equipment and standardize
the procedures. One was used to check the effect of a particular
"wave" form in the fabric due to an error in fabrication, and 33
were used to establish the S-N curve described later.

3. Test Results and Evaluation

a. S-W Curve and Discussion of Fatigue Data

Fatigue tests were run at the following six stress levels in

the facing laminate: 25,000; 20,000; 17,000; 14,000; 12,000;
and 10,000 psi. Three specimens were used to establish each
point on the S-N curve for each type of sandwich construction
tested, except for the 10,000 psi level where only three specimens
vere tested for both types. A sunnary of the results is shown in
Tables 6 and 7 for each type of construction.

The numbers of cycles to failure at each stress level were

weraged to plot the S-N curve. The resulting curve is
presented in Figure 17. It will be noted in studying the
data at each stress level that there is a considerable spread;
yet, despite this spread in the data, the average value at
each point falls on a smooth curve. The data for both the
aluminum-core specimens and the HRP-core specimens seem to fall
on the same curve; in fact, it did not seem feasible even to
distinguish between the two materials in drawing the curve.
This was not surprising because the type of flexure test used
with constant moment on the test span gives only tension and
compression in the facings with no shear stress in the core.
However, there may have been some incidental uncalculable shear
stress at the glue line or in the core due to small amounts of
torsion or undetected unsymmetrical loading, but this would be
minor compared with the flexural stresses.

All of the specimens failed in the fatigue machine except

three. These included two aluminiuu-core specimens and one
HRP-core specimen loaded at a facing stress of 10,000 psi.
All of these endured 10 million cycles or more. It did not
seem necessary to establish this point with more specimens
since the S-N curve, if projected, indicated the life of i
specimen loaded at 10,000 pni to be about 15 million cycles.

Although the curve seams to be gradually flattening, it does

not necessarily indicate that the material has an endurance limit.
Moat nonmetallic materials do not have an endurance limit. The
term fatigue strength is usually used in such cases and defines the
stress that may be endured for a given number of cycles of load.
From the curve, one can say the material has a fatigue strength
at various numbers of cycles of load N as listed hereafter:

56

•
 u
>>>
< l-J ^-
>>>>
< ^- ■- ^-
>>>>
<^rt^ri^rt
y>>>
< i^ N^ ^«
> > ► > >>>>
< » w> «>
>>>>
< W N> ^
■tO<o It 00 -J »■ <tw«-w It K) ^« O 4 It It
1
a
» • •> i
I ?| I ■ it I -:
1 is ■
0
H Kl Kl K» t/» w £ <
^1 vj ^i *j o o o — w • i
% i
so K o NI at <•
N» »J »- *
OB O » *
It •I
••1 -n
00
It 5
1
o »*■ ooo
w»
i s
Wl «««
•*•r» o o o
M
in
%
1 5 n S5
**■ 0
1 5K
f »-I OOO WS^tWW WK)^-N*
3
r>
i n s W NI UJ f" (0 % a# w
^ QD U« W
^-> % Kl «0 W ^ I ^J "^i «* t^ (^NIWW

^1
8w o <y < ?W ^ Kl ~J WJ * (X
8 s§8: ig! W O O i Kl ^ ^ O NI OD O
B
*
£ I
c
>
•
•»1 W)
=■
"fl 1 O
rv
K»
3 o
k « i o
M
1 ►*
I"
> -n "* ► •>» i ■>» > -fl -«l "«
< ■- ►- <»«»«^« « at -j o- ^ «- 4 w K> »
• M — O ■< I< It
"1
s »I
E !
► J IJ M MM £ «• £ W<
OOO W< W< OB ~J Oi O I» »
I ^-J *- ^- <* -;
«I wi Oi Oi «•■ It • »•
? § §§§ §§§ §§: W W W* Kl
i M it a
o o o OOO o o o OOO VI Ut Wl
r* m n
^- ^- — Kl K» Kl ■ ■ •
in o o
««« OOO
a. *~ n
H
^ ^j O w •-
z
v « a « <r ••»«>►-
(
(r ^ ^ öD <^rwKi^ QwO^
w O Ö -J o OD •- ^W^IKI 9* <& & t*
^ o *■ s IJ w O O * ^/i^^Kf Oi^tMW
>
O o o o
iff
« s »»
t -*
w n M 0
s ir s o
n o
It >•
VI r
I
I«
 85
FATIGUE STRESS/ULTIMATE STATIC FAIL. STRESS
p P p o o
5 s o o i
FACING STRESS (pti x I0"3 )
oi
us >n — iß m o o
O 1
If
o (0><
1

-
W •
rr
l

1
• OF

ERA6E

C •*>
SPE
■•• HRP
ALU

.900 pi
CING Fi

0 ft
r» r»
C p.
1 09
n C
A
CA
C M
■ siii«! :
O* •
*— SB .- H ^ 2o - f/
| o» 2 > Eo A \
2w 2 «m /
^3 o n^ * </» / -
-<
O Mt o 0» H « > /
o r — m < M
m o m i _
1 0) O * 1
i ■
•n 5
o / -
^- ft oi m j ' -_
» «I
A» -
|| ,i
1 9 0 y
ft oo ? » /
r-S r
a - J -
8a x
H- o m
9 9
00 A
• •<
o 3
- ^
I
n
1
y
A
/
- -
/
- -
/
-
,/
■
 Fatigue Percent of Number
Strength Ultimate of
Strength Cycles (N)

19,000 387.
in 106
l0
14,500
10,000 207. 5
10' plus

b. Description of Types of Fatigue Failure

Figure 18 on the following page Illustrates some typical fatigue

failures. This Is a group of aluminum-core specimens loaded
at 14,000 psl In the facings. The failures were approximately
perpendicular to the direction of the principal stress and
followed the fabric fill direction (across the width direction).
If the weave was slightly canted from perpendicular, the failure
line followed the slant of the weave.

The majority of the failures occurred In the center "Inch" of

the specimens but rarely at the narrowest point. The point of
failure seemed to be Influenced by edge defects In some cases;
I.e., after failure, four or five small cracks were noted extend-
ing into the center of some of the specimens a distance of
approximately 1/32 Inch from the edge, especially in the
center Inch of the specimen in an area adjacent to the failure
line. All of these short cracks developed during the test, and
apparently the most critical crack was rapidly propagated across
the specimen at failure. At the high stn ses, a sharp "report"
was heard at failure as the crack propagated across the specimen
in apparently one quarter cycle of load.

Simultaneously with the facing failure, the core material ruptured,

with a vertical crack penetrating from the facing rupture to the
opposite unruptureJ face. This is the only phase of the test
where one type of core seemed to have superiority over the other.
Here the aluminum core had only the vertical crack, while the HRP
core became more shattered and the facing peeled from '.he core
for about 1 to 2 Inches back from the facing break die to
acceleration forces in the fatigue machine. There were some
diagonal cracks originating in or passing through the core per- *
forations in some of the specimens. The cracks were at about 45
degrees. It was impossible to determine if these cracks were a
function of the fatigue process similar to those reported by
Apodaca and Preston (reference 50), where progressive cracking of
core walls was reported, or if they were a function of the final
rupture. At any rate, it appeared that the energy absorption
characteristics of the aluminum core would thus be superior.

59
 09
H. ff» H •<
■ O (» H>
I» vO «• 0Q
3 « »T C
m <* n n
•-* d t-
U) O r» 00
>r^
O CD
*» • OX
• ^' Q"«
■ O H-
Pi n
a o « ^
i .Mimt
r» —•
> n "1
P»
/-s 2 >£
a « OP
•»J rr C
» 0 M A
l— C "O
M 9 • 9
H> • n •
-»
0Q O § M
n» A C
to S H
rr »fl «
►- H- N>
■ (^ K< M •V
ft.'
., I «
« H> H> e
« 3 rr •
§» S* ft
•• 9 P 9
^852.
•3 A
oo so*
o o _*<
N» 3 •»!
ft) ft
*» 11 O* 3-
§0 1 A
r ►*
3 0) P»
P» 00 rr g
3 •• H> H>
a o i->
0B 3 •<
to-
O 0
O O A «>
Ut<• ft)
W A '/>
. P» O -3
W* 3 ft A
o a. A
O »
><
»A
H. 3

**r
n M M «
^•3 A
A A a
(0 A
Microscopic examination of the facing surface after failure showed
a pattern of small microcracks. The cracks were perpendicular to
the direction of the principal stress, following the fill direction
of the fabric. The depth of cracks could not be determined, and
they may not have penetrated beyond the first layer of fabric. It
was not possible to observe with the microscope (100 power) the
relationship of cracks to the cross-over points between the warp
and fill strands because the strands are visually merged with the
resin.

Apparently these microcracks have been observed before. Wilson

(reference 48), for example, has reported the accumulation of
interlaminar air blisters in filament-wound air-storage bottles
cyclicly pressurized (4 cycles per minute) at as low as 3,000
cycles of pressure. These blisters could have been due to the
formation of microcracks and the consequent penetration of the
pressurized gas into the laminate. Broutmen (reference 7) also
has studied these cracks for filament-wound plastics. He indicat-
ed that the cracks propagate to form larger cracks when the mate-
rial is subjected to creep and fatigue loads and that these cracks
can exist well in advance of ultimate fracture.

Scattered throughout the resin are small air bubbles ranging be-
tween 0.002 inch and 0.005 inch in diameter. These bubbles did
not seem to influence the crack pattern; i.e., they were not the
origin of the microcracks. Although sometimes cracks terminated
in bubbles or passed through them, there were many bubbles which
were bypassed by cracks. The bubbles were only faintly visible
to the naked eye, and the surface of the specimens could usually
be described as smooth and glossy. Some specimens exhibited more
air bubbles than others, but there did not seem to be a relation-
ship between the fatigue life and the number of air bubbles. On
the contrary, some of the specimens with the greatest number of
bubbles had the longest fatigue life.

The microcracks were spaced on the order of 1/64 inch to 1/32 inch
or less and, when magnified, resembled a group of straws laid par-
allel to one another in a sort of lattice with very few perpendic-
ular connections. The crack pattern was similar to the crack pat-
tern observed in simple tension specimens. It could not be deter-
mined whether these cracks developed during the fatigue process or
whether they were a function of the final failure process.

Werren (reference 47) has reported similar crazed formations in

polyester laminates when tensile stresses beyond the initial
proportional limit are applied. He hypothesizes that cpoxide

61
 laminates will probably craze also. Thus the resin gives less
lateral support to the glass fibers on successive compressive
loads, and failure may occur at a fewer number of cycles because
of the more critical lateral buckling stresses in individual
fibers.

If. seems that more research should be done in the uture on the
formation of microcracks and the part they play in the fatigue
process. Perhaps there are improvements in the weave or type of
resin used which may eliminate them and, thus, Increase fatigue
life.

Specimen A in Figure 18 Illustrates the failure of a specimen

(A21, Table 6) which had a "wave" distortion of the fabric
inadvertently produced In the fabrication process. Note that
the failure line followed the "wave" fill line of the fabric.
The fatigue life of this specimen was substantially below the
average fatigue life of the other aluminum-core specimens in
this group (14,000 psl): if endured only 38 per cent of the
average fatigue life. Another similar specimen (A8, Table 6)
was the highest in Its group (17,000 psl), so that no conclusions
can be drawn about such fabrication distortion defects.

c. Comparison of Results With Past Research

The only fatigue tests similar to those reported herein were

performed by Keer and Lazan (reference 19). Only four specimens
were reported on fiberglass facing-fiberglass core combinations,
but several, other types of sandwich constructions were reported.
The tests were performed on an electromagnetic exciter with a
double cantilever specimen mounted symmetrically on the machine.
The specimen was vibrated at resonance, and a combination of
flexure and shear was produced In the specimen Instead of pure
flexure as in the present tests. The four specimens established
a smooth, but flat, S-N curve. All four specimens failed between
12,000 and 300,000 cycles of load at facing stresses ranging from
9,000 to 7,000 psl. None of the specimens In the tests given
herein has such low fatigue life--at 10,000 psl, thl liwest
stress used, the life exceeded 10 million cycles.

No data on fabrication details were given, so that a direct

comparison of results Is impossible. In the resonant system
used by Keer and Lazan, the natural frequency dropped off near
failure as cracks developed. This gave more time to observe
failure. The failures reported were similar to the tests
reported herein; i.e., facing failure began with the breaking
of many fibers of the facing (possibly similar to microcrack
formation) and the development of a fragmented white portion

62

•
on the edge of the beam (delamination and edge crack) which
moved toward the center, slowly at first, but finally resulting
in a rapid failure. The frequency and load were maintained
constant right up to failure in the tests reported herein, so
that failure was quicker and of a more catastrophic nature.
Other than having similar modes of failure, the tests do not
seem to be comparable.

Other fatigue tests reported on sandwich construction are

those by Werren (reference 46, parts C and H) and James and
Norris (reference 17), both performed at the Forest Products
Laboratory (FPL). All of these tests were aimed at core
shear fatigue and were not comparable to the flexure tests
reported herein. Nevertheless, some interesting conclusions
were drawn which may affect the direction of USAAVLABS work.
Both FPL investigations used heavier aluminum core. Werren
used 3/8-inch cell size, perforated aluminum foil, 2 or
4 mils thick, laminated between 2024 clad aluminum alloy
facings with a high temperature, thermoplastic, thermosetting
resin. James and Norris used aluminum core with 3/8-inch cell
sise with 2-, 3-, or 4-mil thickness foil bonded directly to
the test fixture. In both cases the glue line bonding the core
to facings or fixture was generally unaffected by the fatigue
loads ea-ly in the test, causing diagonal cracks in the core.
This caused a slight load drop in the test, but did not cause
failure. Final failure was a combination of diagonal cracks
and buckling of core walls, with core-to-facing glue line failure
as a secondary effect.

The James and Norris work indicated that the shear fatigue
strength at 100,000 cycles averaged 40 per cent of the static
shear strength. Also, the fatigue strength at low frequency
(15 cps) was substantially higher than that obtained at high
frequency (200 cps) by a factor of 1.75. The fatigue limit
ranged from 20 to 23 per cent of static strength. Warren's re-
sults for aluminum core were higher, indicating a fatigue strength
of 23 to 36 per cent of static strength at 30 million cyclea of
load depending on the plane of loading (ribbon direction or
perpendicular to ribbon directi n).

In a supplement to fie first report, Werren (reference 46,

part K) la.er reported more tests where core-to-facing bonds
were more critical. However, given adequate bond, a fatigue
strength of 38 per cent of the static strength could be expected
at 30 million cycles of load. Since the core used in the test»
reported herein was only 1 mil in thickness and had a density
of no more than 50 per cent of those in the other investigations,

63
additional shear fatigue tests on aluminum core are needed. The
investigation of the integrity of the core-to-facing bonds should
also be a part of the program.

Werren (reference 46, part K) also tested fiberglass-honeycomb

core laminated between 6-ply fiberglass facings or aluminum
facings in shear fatigue in a similar method to that described
above. The tests showed a fatigue strength at 30 million cycles
of load of about 42 per cent of static strength. No weakness of
glue line was indicated. Failures began by shear cracks either
parallel or perpendicular to the flutes which gradually progressed
during the test. The fatigue strength was affected very little
by the direction of loading (ribbon direction or perpendicular
to ribbon direction).

The only other fatigue tests on sandwich construction are those

reported by Apodaca and Preston (reference 50). Several types
of metal-honeycomb sandwiches were tested. A double cantilever
specimen similar to Keer and Lazan's (reference 19) was tested
at resonance in an electromagnetic vibrator. Thus, the specimen
was subjected to both shear and flexure. The pertinent specimens
used aluminum core material (S052-H39) with 3/16-incU cells and
a 2-mil foil thickness. There was considerable scatter in the
results, but a core shear fatigue strength of 27 per cent of the
ultimate core shear strength was established at one million cycles.
The shear fatigue strength found at the Forest Products Laboratory
(reference 17 and 46) for pure shear is substantially higher
than those reported by Apodaca and Presten. This leads to the
hypothesis that flexure and shear combined loads will lead to
■hear fatigue failures at lower numbers of cycles of loads at a
given core shear stress even if core shear fatigue failures are
the final result. Thus, the area of combined flexure and shear
fatigue loading needs further investigation.

There has been a great variety of fatigue tests on laminates of

various thickness, types of fabric, and adhesives reported in the
literature over the past few years (references 1, 4, 5, 7, 9, 21,
28, 32, 39, and 47 are a sample of these). Some of this work
bears directly on the present results. Boiler (reference 4)
reported tests on a series of 23-ply fiberglass laminates using
181-Volan A fabric and alkyd-styrene or polyester adhesives. The
results indicated a fatigue strength of 9,000 psi (FS/UTS - 0.22)
at 10 million cycles and 20,600 psi (FS/UTS ■ 0.S1) at 10,000
cycles. These results are slightly below the fatigue strength ob-
tained in this project, but Pusey (reference 32) has indicated that
polyester laminates are inferior to epoxy laminates in fatigue.

64
 Werren (reference 47) and Kimball (reference 21), both Forest
Products Laboratory Investigators, supplemented Boiler's work.
The specimens were 24- ot 26-ply tension-compression specimens
of 181-Volan A fabric, with an EPON 828 and curing agent CL
adhesive system. The tests were aimed at comparing notched
specimens wit.i unnotched specimens loaded at various angles to
the warp direction. Several S-N curves reported by Werren senm
to be comparable to the results reported herein for sandwich
facings. The fatigue strength to ultimate static strength
ratio ranged from 0.27 to 0.41 at 10 million cycles and 0.35
to 0.48 at one million cycles. The FPL results, whether express-
ed in ratio form or in psi, exceeded the fatigue strengths for
the facings reported herein. Of course, this leads to the suspi-
cion that there is a difference in failure mechanism between a
3-ply laminate and a 24- to 26-ply laminate or that the methods
of tests are not comparable at all.

4. Conclusions and Recomnendations

The major conclusions and recomnendations drawn from the fatigue test-
ing phase of the research are as follows:

a. The composite sandwich construction of the type tested does not

have a fatigue endurance limit. Instead, a fatigue strength of
the facings at a given number of cycles of load may be used to ex-
press its fatigue characteristics. For these tests the fatigue
strength was 14,500 psi in the sandwich facings at one million
load cycles.

b. There was considerable scatter in the results of the fatigue

tests at a particular stress level due to the great number of
variables present in composite sandwich construction. Hie average
of the data at each stress level allowed the construction of a
smooth S-N curve. Wher eno igh similar tests are made, the prob-
ability of failure should be included in the analysis and proba-
bility of failure versus cycles of load curve constructed for
various stress levels.

c. All fatigue failures for fiberglass sandwich construction loaded

in pure flexure will occur in the facings. The core material and
glue line between core and fatigue were unaffected within the
limits of the tests reported herein.

d. There was no difference in fatigue life between the fiberglass

sandwiches using aluminum honeycomb core and those using HRP
fiberglass honeycomb core.

65
e. Further research should be performed on the fatigue strength of
the glue line between the facing laminate and the core material
using the combination of materials reported in this research.
Past research indicates that the glue line is not a problem for
various other adhesive combinations, but this should be verified
for the present material combinations.

f. Further research should be carried out on combination fatigue

loading conditions; i.e., combinations of flexure and shear In
sandwich construction, and biaxial loading of both laminates alone
and as used in sandwich construction with various types of core
material. This research should not be carried out until more
fabrication research and static tests are performed on singly and
douby curved panels and on full-scale components.

g. At this point, it appears that the sandwich construction with

aluminum honeycomb cores has a slight advantage over those using
HRP fiberglass honeycomb cores, This opinion is based mostly on
the more extensive damage done to the HRP core in the fatigue
failure process even though both types fall on the same S< N
curve.

h. Research on the development of microcracks and methods of

preventing them is encouraged. Although some improvements can
probably be made along these lines by changes in fabric, the main
problem is probably in the molecular structure of the resin. Thus,
the compatibility of resin with fiberglass will probably be im-
proved mainly by changes in the adheslves and glass coatings.

i. The fatigue results are generally encouraging and compare favor-

ably with previous tests reported for fiberglass laminates alone.

66

I .
 REFERENCES

1. Anderson, James A., and J. A. McCarthy, "Pre-preg Reinforced Plastics

in Fatigue Applications," Proceedings. 18th Annual Techrtcal and
Management Conference, Reinforced P.astics Division, The Society of
the Plastics Industry, Inc., February 1963, Section 8-C, pp. 1-8.

2. Anderson, R, A., "Flexural Vibration- in Uniform Beams According to

the Tlmoshenko Theory," Journal of Applied Mechanics 20, Transactions
of the American Society of Mechanical Engineers 75. 1953, pp. 504-510.

3. Bert, C. W., end W, S. Hyler, Design Considerations in Selecting

Materials for Large Solid-Propellant Rocket-Motor Cases. Report
Number 180, Defense Metals Information Center, Battelle Memorial
Institute, Columbus, Ohio, 1962.

4. Boiler, K. H,, Fatigue Testa of Glass-Fabric-Base Laminates Subjected

to Axial Loading, Report Number 1823, Forest Products Laboratory,
Madison, Wisconsin, August 1956.

5. Boiler, K. H., Supplement to Fatigue Tests of Glass-Fabric-Base

Laminates Subjected to Axial Loading. Report Number 1823-A, Forest
Products Laboratory, Madison, Wisconsin, April 1954.

6. Bresse, M., Cours de Mtcanique Appliqulc. Paris: Mallet-Bacheller,

1859.

7. Broutman, L. J., "Failure Mechanisms for Filament Reinforced Plastic

Subjected to Static Compression, Creep and Fatigue," Proceedinga.
19th Annual Technical and Management Conference. Reinforced Plaatics
Division, The Society of the Plastics Industry, Inc., February 1964,
Section 9-C, pp. 1-10.

8. Clary, R. R., and S. A. Leadbetter, An Analytical and Experimental

Investigation of the National Frequencies of Uniform Rectangular-
Cross-Section Free-Free Sandwich Beams. Technical Note D-1967,
National Aeronautics and Space Administration, 1963.

9. Cornish, R. H., H. R. Nelson, and J. W. Dally, "Cooprcsslve Fatigue

and Stress Rupture Performance of Fiber Reinforced Plastics," Pro-
ceedings, 19th Annual Technical and Manageawnt Conference. Reinforced
Plastics Division, The Society of the Plastics Industry, Inc.,
February 19o'*, Section 9-E, pp. 1-22.

10. Dar, S. M., "Vibrations of Rectangular Sandwich Plates With Varioua

Edge Conditions," Unpublished Ph.D. Dissertation, University of
Oklahoaa, 1964.

67
11. Dengler, M. A. ( «nd M. A. Goland, "Transverse Impact of Long Beams
Including Rotatory Inertia and Shear Effects," Proceedings of the
Flret National Congress. Applied Mechanics. American Society of
Mechanical Engineers, New York, 1951. pp. 179-186.

12. Dove, Richard C, and P. H. Adams, Experimental Stress Analysis and

Motion Measuramant. Charles E. Merrill Books, Inc., Columbus, Ohio,
1964

13. Glaser, A. R. "The Vibration of Sandwich Beams," ProceedInRS of the

7th Midwestern Mechanics Conference, Developments in Mechanics.
Volume 1, Plenum Press, New York, 1961, pp. 228-238.

14. Hackman, L. £. ( R. J. Molella, C. L. Stctler, and D. G. Worthington,

Structural Fibar Glass Aircraft Component-Program Results. Paper No.
64-442, American Institute of Aeronautics and Astronautics, 1964.

15. Huang, T. C, "The Effect of Rotatory Inertia and of Shear Deformation

on the Frequency and Normal Mode Equations of Uniform Beams with Simple
End Conditions," Journal of Applied Mechanics 28. Transactiont of the
American Society of Mechanical Engineers 83E. 1961. pp. 579-584.

16. James, William L., Calculation of Vibration Damping in Sandwich

Construction from Damping Properties of the Cotes and Facings,
Report Number 1888, Forest Products Laboratory, Madison, Wisconsin,
1962.

17. James, William L., and Charles B. Norris, An Apparatus for Measuring
Internal Friction and Fatigue Strength of Core Materials Used in
Sandwich Construction. Report Number 1866, Forest Products Laboratory,
Madison, Wisconsin, October 1958.

18. Kaman Aircraft Corporation, Compilation and Analysis of Test Data

on Fiberglass-Re inforcfcd Plastics. USATRECOM Technical Report 64-9,
U. S. Army Transportation Research Command, Fort Eustis, Virginia,
March 1964.

19. Keer, L., and B. J. Lacan, Damping and Fatigue Properties of Sandwich
Configurationa in Flexure. Technical Report 61-646, Aeronautical
Systems Division, U. S. Air Force, November 1961.

20. Kimball, A. L., and D. E. Lovell, "Internal Friction in Solids,"

Physical Review, Series 2 vol. 30, 1927, pp. 948-959.

21. Kimball, K, E., Supplement to Fatigue Tests of Glass-Fabric-Base

Laminates Subjected to Axial Loading-Effect of Notches. Report Number
1823-C, Forest Products Laboratory, Madison, Wisconsin, October 1958.

22. Kimel, W. R., M. E. Raville, P. G. Kirmaer, and M. P. Patel, "Natural

Frequencies of Vibration of Simply Supported Sandwich Beams," Pro-
ceedings of the 4th Midwestern Conference on Solid Mechanics. 1959,
pp. 441-456.
68

1 I
23. Kobayashl, S., "On Vibration of Sandwich Ream," Proceedingg of the
4th Japan National Congress of Applied Mechanics, Tokyo, 1954,
pp. 369-372.

24. Kruszewski, E. T., Effect of Transverse Shear and Rotary Inertia on

the Natural Frequency of a Uniform Ream, Technical Note 1909, National
Advisory Committee for Aeronautics, 1949.

25. Lazan, R. J., Fatigue Failure Under Resonant Vibration Conditions.

Technical Report 54-20. Wright Air Development Center, 1954. Also.
American Society for Metals, Special Publication on Fatigue, 1954,
pp. 36-77.

2b. Mindlin, R. D., "Influence of Rotatory Inertia and Shear on Flexural

Motions of Isotropie Elastic Plates," Journal of Applied Mechanics 18.
Transactions of the American Society of Mechanical Engineers 73, 1951,
pp. 31-38.

27. Mindlin, R. D., and H. Deresiewicz, "Timoshenko's Shear Coefficient for

Flexural Vibrations of Beams," Proceedings of the Second U. S. Nation-
al Congress on Applied Mechanics. American Society of Mechanical Engi-
neers, 1954, pp. 175-178.

28. Nara, H. R., "Some Fatigue Characteristics of Glass-Reinforced

Plastics," Proceedings. 12th Annual Technical and Management Con-
ference, Reinforced Plastics Division, The Society of the Plastics
Industry, Inc., February 1957, Section 5-D, pp. 1-6.

29. Nordby, Gene M., and W. C. Crisman, The Effect of Resin Contenc and
Voids on the Strength of Fiberglass-Reinforced Plastics for Alrframc
Use. USAAVLABS Technical Report 65-66, U. S. Army Aviation Materiel
Laboratories, Fort Eustia, Virginia, November 1965.

30. Nordby, Gene M., J. V. Noyes, and W. C. Crisman, Research in the

Field of Fiberglass-Reinforced Sandwich Structure for Airframe Use.
USATRECOM Technical Report 64-37, U. S. Army Transportation Research
Comnand, Fort Eustia, Virginia, July 1964.

31. Plass, H. J., Jr., "Damping of Vibrations in Elastic Rods and Sand-
wich Structures by Incorporation of Additional Viscoelastic Material,"
Proceedings of the 3rd Midwestern Conference on Solid Mechanics, 1957,
pp. 48-71.

32. Pusey, B. B., "Flexural Fatigue Strengths of Reinforced Thermosetting

Laminates," Proceedings. 12th Annual Technical and Management Con-
ference. Reinforced Plastics Division, Society of the Plastics In-
dustry, Inc., February 1957, Section S-C, pp. 1-10.

69
33. Raville, M. E., E.-S. Ueng, and M.-M. Lei, "Natural Frequencies of
Vibration of Fixed-Fixed Sandwich Beams," Journal of Applied Mechanics
28, Transactions of the American Society of Mechanical Engineers 83E.
1961, pp. 376-381. ^ ~~"

34. Raylelgh, J. W. S., The Theory of Sound, First edition, 1877; American
edition, Dover Publications, New York, 1945.

35. Relssner, E., "The Effect of transverse Shear Deformation on the Bend-
ing of El»«*"!. Ixates," Journal of Applied Mechanics 12. Transactions
of ihe American Society of Mechanical Engineers 67, 1945, A69-A77.

36. Richter, H. P. H., "Photographic Method for Measuring Material Damping

and Dynamic Youngs Modulus at Low Frequencies Applied to A Fiberglass
Reinforced Resin Structure." Proceedings. 18th Annual Technical and
Management Conference, Rtmforced Plastics Division, The Society of the
Plastics Industry, Inc., February 1963, Section 4-D, pp. 1-10.

37. Ross, D., E. E. Ungar, and E. M. Kerwin, Jr., "Damping of Plate

Flexural Vibrations by Mr.ans of Viscoelastic Laminae," Structural
DampIng, American Socle y of Mechanical Engineers, 1959.

38. "Sandwich Constructions and Core Materials; General Test Methods,"

MIL-STD-401A, Military Standard, June 1956.

39. Stevens, G. H., Fatigue Teat of Phenolic Laminate at High Strejs

Levels and Elevated Temperatures, Report Number 1884, Forest Products
Laboratory, Madlaon, Wisconsin, 1961.

40. Tlnoshenko, S. P., "On the Correction for Shear of the Differential
Equation for Transverse Vibrations of Prismatic Bars," Philosophical
Magaalne. 41. 6th series, 1921, pp 742-746.

41. Tlnoshenko, S. P., "On the Transverse Vibration of Bars of Uniform

Cross-Section," Philosophical Magatine 43. 6th series, 1922, pp. 125-
131.
42. Tlnoshenko, S. P., Vibration Problems In Engineering. Third edition,
Princeton, N. J.; D. Van Nostrand Company, Inc., 1955.

43. Tlnoshenko, S. P., and J. N. Goodier, Theory of Elasticity, 2nd

edition, McGraw-Hill Book Company, Inc., New York, 1951, pp. 452.

44. Tralll-N&sh, R. W., and A. R. Collar, "The Effect of Shear Flexibility

and Rotatory Inertia on the Bending Vibrations of Beans," Quarterly
Journal of Mechanics and Applied Mathematics 6, 1953, pp. 186-222.

45. Unpar, E. E., "Highly Damped Structures," Machine Design 36 (4),

February 14, 1963, pp. 162-168.

70
46. Werren, Fred, Fatigue of Sandwich Constructlont for Aircraft. Report
Numbers 1559C, 1559D, 1559E, 1559H, 15591, 1559K and 1559L, Foreat
Products Laboratory, Madison, Wisconsin, 1948-52.

47. Werren, Fred, Supplement to Fatigue Tests of Glass-Fabrlc-Bise

Laminates Subjected to Axial Loading. Report Number 1823-B, Forest
Products Laboratory, Madison, Wisconsin, August 1956.

48. Wilson, Frank M., "Development of Fatlguo-Reslstant Filament-Wound

Emergency Air Storage Bottles," Proceedings, 19th Annual Technical
and Management Conference, Reinforced Plastics Division, The Society
of the Plastics Industry, Inc., Febrar> 1964, Section 1-E, pp. 1-3.

49. Yu, Y. -Y., "Damping of Flexural Vibrations of Sandwich Plates,"

Journal of the Aerospace Sciences. Volume 29, 1962, pp. 790-803.

50. Apodaca, D. R., and J, L. Preston, Fatigue In Shear by Bending of

Structural Sandwich Constructions, Technical Report Number ASD-TR-
61-338, Aeronautical Systems Division, Wright-Patterson Air Force
Base, Ohio, March 1963.

71
 APPENDIX

FREQUENCY AND MODE-SHAPE ANALYSIS

BY TIMOSHENKO BEAM THEORY

A complete derivation of the Timoshenko beam theory for vibration of homo-

geneous beams with flexural and shear flexibility and lateral (translatton-
al) and rotatory Inertia is developed here from first principles. This was
deemed necessary for the following reasons:

(1) The most complete Timoshenko beam vibration analysis reported in the
literature (Huang, reference 15) contains some minor errors.

(2) Inclusion of this derivation makes the complete and correct analysis
available without resorting to numerous references.

(3) It will be used as the basis for sandwich beam calculations discuss-
ed in section A3c.

Derivation of the Diflerential Equation of Motion for a Timoshenko Beam

The Lagrangian method is used here to derive the governing partial differ-
ential equation. For a derivation using Newton's second law, the reader is
referred to pp. 329-331 of reference 42.

All energies are per unit length of beam. The strain energy of bending is

(1/2)EI {hfßlh*)2

where E is the elastic modulus of the facings, I is the area moment of in-
ertia about the neutral axis, & is the slope of the beam deflection curve
due to flexure only, and x is the position co-ordinate along the length of
the beam.

The strain energy of transverse shear deformation is

(1/2)AKG
(£'i
•
where A is the shear area, K is a shear-deformation shape factor to be de-
termined in a later section, G is the core modulus of elasticity in shear,
and y is the total lateral deflection.

75
j
The trans Int tonal kinetic energy is
z
(1/2)M (dy/dt)
where M it the total beam mass per unit length of beam and t is time. The
rotational kinetic energy is
(1/2) J (df/dt)2
where J is the total beam mass moment of inertia per unit length of beam.
Then twice the lagrangian is given by
M (dy/dO2 + J (dftdt)2 - EX (dy/d*)2 • GMtfAz -f) .
Applying Hamilton's principle to the Lagrangian results in the following
set of partial differential equations:

M^Z
2 - GAK/^I
2 - ^]- 0 (A-l)
at (dx dJ
jiJL
2
- EI £#.
2
GAK/Aü - l/f] - 0. (A-2)
^t dx Ux TJ
Solving Equation (A-l) for f? :

r j(dx2 GAK jt2ydx dx JGAK -&d*.

bt

Putting this into Equation (A-2) gives

GAK i JL Ajl dx - 0
J GAK ^t2
Now, differentiating with respect to x to eliminate the integrals results
in

a«4 at2 v <**J^7i? GAK

dtu
or

Equation (A-3) is equivalent to an equation derived by Timoshenko for a

homogeneous beam. The application of Equation (A-3) to sandwich-type
beams is presented in section A3d.

76
In similar fashion, the following equation in y* can be written:

General Solution of the Timoshenko Beam Equation

Due to the presence of the term containing the mixed partial derivative in
Equation (A-3), the x and t variables do not separate exactly in the gener-
al free-vibration case. However, in the present research program, the beam
is lirit excited sinusoidally and then the excitation is cut off. Under
these conditions, it appears to be a very reasonable approximation to as-
sume a sinusoidal wave form shape with respect to time, a.» follows:

y - Y (x) sintft (A-5)

where CJ is the frequency and Y (x) is the modal shape of the beam deflec-
tion curve (to be determined). It should be mentioned that so far as is
known, all previous investigators of both Timoshenko-beam and sandwich-beam
vibrations have made an asnumption equivalent to Equation (A-5).

Substitution of Equation (A-5) into Equation (A-3) gives

2 JM(
Yiv
v HtJ_Y+
Y . / JL+. i_k)
W Y v" +* ^ v - n M
/A 6)
" ^r (GÄK Eiy EIGÄKY"0 (A
-

where the primes denote differentiation with respect to x.

Equation (A-6) can be rearranged as follows:

Equation (A-7) will now be nondloensionalii^d and solved followinp the no-
tation used by Huang (reference 15). First, the following dimensionless
parameters are introduced:

Position parameter £ ■ x/L

2 ML
Bending flexibility parameter b' ■ — ^2
2 , 9 (A-8)
Rotatory inertia parameter r* - J/ML*

Shear flexibility parameter s2 - F.I/GAKL2

77
Then Equation (A-7) becomes

Ylv + b2 (r2 + e2) Y" - b2 (1 - b2r282) Y - 0 (A-9)

where now the primes denote differentiation with respect to 6 .

In similar fashion, the solution of Equation (A-4) can be assumed to be of

the form

ff ' S (x) sin tot . (A-10)

Then, following the identical procedure used above to obtain Equation (A-9),
the following result is obtained:

Slv + b2 (r2 + s2) S" - b2 (1 - b2r2s2) S - 0 . (A-ll)

The solutions of Equations (A-9) and (A-ll) can be written as follows:

Y - C, cosh b«<€ -f C2 sinh b<€ ♦ C3 cos bßC ♦

C4 sin h/3£ (A-12)

and

g • C{ sinh b*e 4- C^ cosh h*€ + C^ sin y9€ -»-

C| cos yf£ (A-13)

where
1/2
2 2 2 2
*. (1//J) |- (r + s ) + [(r2 - s ) + {klhif'f (A-14)

/J- (1//7) | (r2 + s2) + [(r2 - s2)2+ (A/b2^ 2

| (A-15)

In writing the solutions In the form of Equations (A-12) and (A-13), it is

necessary to assume that
1/2
tr2 - s2)2^ (A/b2)] > I . (A-16)
r2+s2
For the sandwich beams used in this project, the inequality (A-16) is sat-
isfied for all frequencies below approximately 10,800 cps, which is well
above the highest natural frequency obtained in the project.

The modal shape functions given by Equations (A-12) and (A-13) involve
eight unknown constants of Integration. However, only four boundary
conditions (two at each end) can be written for a given simple beam con-
figuration. Thus, some additional relationships among the coefficients

7d

11
must be found. To do this, Equations (A-5) and (A-10) are substituted
Into Equation (A-l) with the following result:

Y" + b2s2Y - LS - 0 . (A-17)

Substituting solutions Y and S as given by Equations (A-12) and (A-13) Into

Equation (A-17) gives the following four relationships:

Cj/q - C2/q - (L/hoC) [l - b2s2 (mC2 + r2)J

r T (A-18)
- C3/CJ - C4/C£ - (L/b/J) [l + b282 (fl2 - r2)J .

Equations (A-12), (A-13), and (A-18) are Identical to those obtained by

Huang. When used In conjunction with appropriate boundary conditions,
these equal ions allow solution for the natural frequencies.

Solution for a Simple Free-Free Tlaoshenko Beam

As first pointed out by Dengler and Goland (reference 11), the correct
boundary conditions for a free end are as follows:

and
£
Ö,är
- 0 (A-19)

41 - f ' 0 . (A-20)
d*
Equation (A-19) states that the curvature due to bending Is zero, while
Equation (A-20) states that the shear deformation Is tero. Applying
Equations (A-19) and (A-20) at both ends of the free-free beam ( 6" - 0
and £ ■ 1) gives the following four boundary conditions in terms of the
modal functions Y and S:

S' (0) - 0 Y' (0) - LS (0) - 0

(A-21)
S' (1) - 0 Y' (1) - LS (1) - 0 .

Substitution of Equations (A-12) and (A-13) into boundary conditions

(A-21) results in the following set of simultaneous equations:
•rcj ■•- ac^ - o
•fCj cosh b* + «^CJ sinh b«^ + ^5cj cos ifi - flC^ »in \y$ - 0 (A-22)
(b/L) (0<C2 +/*CU) - CJ - q - 0

79
(b</L) (C, ilnh b«^+ C, cosh b«0 - (b.f/L) (C3 iln hfi - C4
' ' (A-22)
cos bif > - C[ sinh b«C - C^ cosh b#i - C^ sin hfl - Z\ co% hä - 0 .

Using Equations (A-18) to convert the unprlned coefficients In Equations

(A-22) to primed ones results In the following set of four equations In
the four primed coefficients:
C 1 wo
*< CJ cosh b< •♦••CCJ slnh ba< +/*C' cos \>fi - /5c^ sin yf - 0
(A-23)
- (ae-2 + r2) CJ ♦ {fl1 - r2) C^ - 0

- (aC2 + r2) (CJ slnh ba< + cj cosh hmC ) + i/f2 - r2)

•(C^ sin b^ + C^ cos yf) - 0 .

In or let for nontrlvlal solutions to exist for the set of Equations (A-23),
the determinant of the coefficients of Equations (A-23) must vanish. Solu-
tion of this frequency determinant leads to the following transcendencal
frequency equation:
2 2 2 2 2
2(1- cosh b*^ cos b/9) + *>,.,„. [b
2 1 2 L r (r - s ) +
r (i. bVs ) /
(A-24)
2 2
(3r - s )3 slnh baC sin yl - 0 .

Equation (A-24) is identical to that derived by Huang (reference IS) for

this case. In order to determine the normal modal shape Y corresponding
to a particular natural frequency, Equation (A-12) can be rewritten as
follows:
Y/Cj - cosh b*€ + (Cj/C^ sinh h*€ + (C^/C^ cos biftf +
(A-25)
(C^/Cj) sin \y9* .

Use of certain of the boundary conditions (A-23) in conjunction with rela-

tions (A-18) results in the following expressions:

Cj/Ci
2 1 - ->V . where Al?
m , y,
f coah b»C - cos b/>
' fi ' " A sinh b< - ^sin bfl

and r.^i s2
•<2+.

80

I
 Vci •l/*

Substitution of these relations into Equation (A-25) gives the following

result:

Y/Cj - cosh b<£ - AYsinh b»<€ + (I/IT) cos hfi6 -

(A 26)
V . hJ3€
Ysln K^ -

It is noted that there tre two errors in sign in Huang's Equation (53)
which is otherwise identical to Equation (4-26) here.

81
 UNCLASSIFIED
ggcttrity CI«MiHc«tion
DOCUMENT CONTROL DATA • RAD
fS»€uniy clmutlllcsltm ol ml» b^^v ml »btitmti »n4 ind»tm$ mnnotmtton mutt ' • •ni*r*4 *lnm JIM •«•Mil imßoti ia cl—mtlt»dj
I O^iniNATIN 0 ACTIVITY (Coipow *H*ot) I« MCPOffT •■CUNlTV C LAStiriCATlO»
Unclassified
University of Oklahoma Research Institute I» «MOU»

I ntPomr TiTLt

DYNAMIC ELASTIC, DAMPING, AND FATIGUE CHARACTERISTICS OF FIBERGLASS

REINFORCED SANDWICH STRUCTURE
4 OtSCftl^TivI NO*tS (Typ* ol —port an* mclunv 4mf»)

Final
t HUTHOmii) (LmH nam» Um n»m» ,ni. It

Nordby, Gene M., W. C. Crlsman, and Charles W. Bert

c ntPomr DATC
October 1965 81 50
•• CON1 iws;
DA-44-177-AMC-164(T)
6 PKOJfCT NO USAAVLABS Technical Report 65-60
Task 1P121401A14176 »6 OTHKM NfPONT NOfS> (Any ol/i*r number* •>•! may ba »au0>a4
mla rapoti)

10 A VA IL ABILITY LIMITATION NOTICES

Qualified requesters may obtain copies of this report from DOC.
This report has been furnished to the Department of Commerce for
sale to the public.
11 lU^LKMtNTANV NOTII II S^ONSOHINO MILITANV ACTIVITY

US Army Aviation Materiel Laboratories

Fort Eustis, Virginia
II ABtTMACT

Research was conducted to determine the basic dynamic properties of fiber-

glass-reinforced plastic (FRF) sandwich structure suitable for use aa a
primary airframe structural material. The research program was carried out
in two separate parts: (A) determining dynamic moduli and damping, and (B)
determining fatigue behavior. In each part, two types of hexagonal-cell
honeycomb core materials were investigated: 5052 aluminum foil and HRP
(heat resistant phenolic) fiberglass. The selection of the sandwich con-
figuration and fabrication procedures was based upon previous research
carried out at the University of Oklahoma Research Institute for USAAVLABS
(USATRECOM TR 64-37 and USAAML TR 65-66).

DD .HSL 1473 UNCLASSIFIED

Security Classification
 UNCLASSIFIED
Security Clussidcution
tr LINK A LINK ■ LINK C
KCV mono*
MOL l

FRF Sandwich dyiuunlc properties

INSTRUCTIONS

1. ORIGINATING ACTIVITY. Enlcr th* nam* and «Idrvtt 10. AVAILABILITY ^LIMITATION NOTICES: Enter any lim-
of lh» contractor, tubcontraclor, irania«, Dapartmant of De- itations on further diaaeminatton of the report, other than thoaa
fanaa artlviiy or othar erganltalton Ccoiporar» author) ttautng
imposed by security claaalflcation, using standard statement»
tha rapoft.
auch aa:
3«. REPORT SECUHTY CLASSIFICATION Entar lha ovtf
(1) "Qualified requesters may obtain coplea of this
all aacurily claaalficalton or lha raport. Indicata whathat
report from DDC "
"Raatriclad Data" la IncludaA Mariiinc ta to ba In accord-
anca with approprlala i ocurliy ragulallona. (3) "Foreign announcement and dissemination of thi .
report by DDC la not authorised "
26. GROUP: Automatic downuradinc la apaclflad In DoD Di-
ractiva »200.10 and Armad Foi -<•• Induatrlal Manual. Entar (3) "U. S. Government agencies may obtain copies of
lha group numbar. Alao, whan a^iicabl». «hew that optional thla raport directly from DDC, Oner qualified DDC
marking« hava baan uaad (or Group 3 and Group 4 as author- users shall request through
Uad.
3. REPORT TITLE: Entar tha complata raport tllla In all (4) "U. S. military agencies may obtain copiea of thin
capital lattara. Till«. In all caaaa should ba unclaaaldad. repot, directly from DDC Other qualified user«
U a maanlngful tula cannot ba aalactad without rlaaaifica-
ahall requeal through
tlen, ahow lltla clasalflcatlon In alt capttala In paranlhasit
Imatadlatoly (ollowing tha tllla,
4. DESCRIPTIVE NOTE» If appropnata. antar th. type f (S) "All distribution of thla report ia controlled Qual-
report, a.g.. Intarl«, prograta. aummary, annual, or final, ified DDC users shall requeal through
Civa tha Inclualva dataa whan a apaclflc raportlng parlod la
covarad.
If the report has been furnished to the Office of Technical
5. AUTHOR(8): Entar tha nama<a) of authoKa) aa ahewn on Services, Department of Commerce, for aale to the ptibllc, indi
or la tha rapon. Entar last nama, fust nama, m.ddla Initial. cate this (act and enter the price, If known.
If military, ahow rank and branch of aarvlco. Tha nama of
tha principal author la an absolut« minimum raqulramant, II. SUPPLEMENTARY NOTES: Use for additional eiplana-
tory notea.
6. REPORT DATE: Entar tha data of tha raport aa day,
month, yaar, or month, yaar. If mora than ona data appears 11 SPONSORING MILITARY ACTIVITY: Enter the name of
on tha raport, uae data of publication. the dapartmental project office or laboratory sponsoring (pmy
ing lor) the reaearch and development. Include address.
7a. TOTAL NUMBER OF PAGES: Tha total paga count
aheuld follow normal pagination procaduraa, La,, antar tha 13 ABSTRACT Enter an abalracl giving a brief and factual
numbar of pagaa containing informatioih summary of the docun-.enl indirative of the report, even though
H may also appear elsewhere in the body of the technical re-
76. NUMBER OF REFERENCE» Entar tha total numbar of port. If additional spare is required, a continuation sheet
rafarancaa cltod in tha raport. ^hall be attached.
So. CONTRACT OR GRANT NUMBER: If approprlala, antar It is highly desirable that the abstract of claaalfied re-
tha applicable number of tha contract or grant under which ports be unclassified. Each paragraph of the abatract ahall
the report waa written. end with an indication of the military security classification
S6, Sc, fc Sd. PROJECT NUMBER: Enter tha appropriate of the information in the paragraph, represented as ITS) (S).
military depart-iant identification, such aa project number, (C). or (U>
aufaproject number, ayatem numbers, leak number, etc, There is no UmitaMon nn the length of the abatract. How-
ever, the suggested length is from ISO to 22S words.
9a. ORIGINATOR'S REPORT NUMBERS): Enter the offi-
cial raport number by which the document will be Identified 14. KEY WORDS Key words are technically meaningful lerma
and controlled by the originating activity. This number must or abort phraaes i.iat chararterire a report and may be used aa
be unique to this report, indes entriea for cataloging the report. Key worda muat be
«elecleH ao that no security classification is required. Iden-
96. OTHER REPORT NUMBER(S): If the report hea been
fiers, such as equipment model designation, trade name, -nili-
assigned any other raport numbers (either by the originator
tary project code name, geographic location, may be used as
or 6K (fie apenaor), alee enter this numberta).
key words but will be followed bv an indication of technical
contest. The aasignment of links, rules, and weights is
optional.

UNCLASSIFIED
Security Classification

•i« Ml

' I