[b

AMMRCTR85-18 JAD

RELIABILITY AND LIFE PREDICTION

METHODOLOGY - M60 TORSION BARS

AD-A 159 197

R. S. BARSOUM, W. M. BETHONY, R. H. BROCKELMAN, H. P. HATCH,

C. F. HICKEY, Jr., W. T. MATTHEWS, and D. M. NEAL

C..-

June 1985 ,

4 DTIC
Approved for public release; distribution unlimited. S ELECTE

ARMY MATERIALS AND MECHANICS RESEARCH CENTER D

Watertown, Massachusetts 02172-0001

85 9 12 050
The findings in this report are not to be construed as an official
Department of the Army position, unless so designated by other
authorized documents.

Mention of any trade names or manufacturers in this report

shall not be construed as advertising nor as an official
indorsement or approval of such products or companies by
the United States Government.

DISPOSITION INSTRUCTIONS
Destroy this report when it is no longer needed.
Do not return it to the originator.

---. a - . a . . .. a7-
 89CUDITV CLASIICATION OF THIS PAGE 11mm OD. Aster. ______________

REPOIT DOCUMEN4TATION PAGE _________________

1. 111KI MUMMER -P. OVTCCE"f9N,4@ 5 CAT ALOG NUMBER

AMOIC TI 85-18
4. TTLE md At~q)S.
Ad V- bq
('ii 1
TYPE OF REPORT & PERIOD COVERED

RELIABILITY AND LIFE PREDICTION METHODOLOGY -Final Report

M60 TORSION BARS
*6. PERFORMING ORG0. REPORT NUMBER

7. ATHOR(@) 9. CONTRACT OR GRANT NUMSER~sj

Roshdy S. Barsoum, Wayne M. Bethony,
Robert H. Brockelman, Harold P. Hatch,
* ~Charles F. Hickey, Jr., and Donald M. Neal ____________

9- PERIORMING ORGAMORATION11 NAME AND ADDRESS 0S. POGRAM ELEMENT. PROJECT. TASK

*Army Materials and Mechanics Research Center D Arjet

WORK NIT 02AH42I
Watertown, Massachusetts 02172-0001 /Prjc:I110A4
AMXM-SMMAgency Accession: DA 0G9431
11. CONTROLLING OFFICE NAME AND ADDRESS 12. REPORT DATE
U.S. Army Materiel Commnand (AMC), June
Alexandria, VA 22333 13. NUNUER Of PAGES

14. MONITORING AGENCY NAME & AODRESM(Ei dillftmt k-m Cm~OUNJI081i

O11000) IS. SECURITY CLASS. (.4 mhi.
reow)

Unclassified
IS. DCASSOPI /(DOWNGRADING
SQIDUL Aoession For

16. DISTRIBUtTION STATEMENT (of Ole Report) SI R

DTIC TAB

Approved for public release; distribution unlimited. Unanmnounc ed

Justification

17. DISTRIBUTION STATEMENT (olth. ab.,ate .snt.,.ed Block 20.11at ralItem Rep..) D~rbai
Ava ilability Co, en6
D tAvail and/.

1S. SUJPPLEMENTARY MOTESDit pesj

A umayo the report was presented at the Army Solid Mechanics Conference 1011
in Rbode Island, October 1, 1984.
INSPECTL'
I9. IKEY WORDS (CWu'buem anorevers aide of neessar ad a*.agp by block amober)3
4Reliability, Torsion bars Nondestructive testinE
Life epcay)Probability,, Random vibration*
*Failure, Mechanical properties1 Maintenance.
Fracture (mechanics).
J
01. ABSTRACT (ConM..r H wesfr and Ideat#*
an revese .4*14 bp block aMbo,)

(SEE REVERSE SIDE)

W FOM I147n EDITION OP' I NOV 65 IS OBSOLETE UCASFE

SECURITY CLASSIFICATION OF THIS PAGE (When 0018 Etft.d)

p.%
 UNCLASSI PIED
UCUOITY CLASPICATOW OU NIl0"C
P10a3 a... rnero.

Black No. 20

ABSTRACT
The life prediction and reliability assessment of the M60 tank torsion
bars were obtained from applying a methodology involving several disciplines.
The disciplines included: structural mechanics, fracture mechanics,
statistical reliability, mechanical testing, nondestructive examination
(NDE), quality assurance (QA), and metallurgical and fractographic
evaluations.

The methodology was applied in order to obtain a procedure for increas-

ing torsion bar life and enhance the reliability of the M60 tank. U.S. Army .
reliability performance is commonly measured in mean miles between failures
(MMBF). In order to introduce a more meaningful measure of component
acceptability, this report describes methods for obtaining minimum life
estimate at a specified probability of survival using the Monte Carlo pro-
cess. This includes predicting remaining life with known risk and an
approach to perform trade-off costs or redesign for increased life. The
result is a more meaningful measure of component acceptability. The
reliability versus bar life computation indicated a negligible amount of
life after a bar flaw was initiated in the bar. The minimum life estimates
prior to crack initiation were in good agreement with frequency of actual bar
failures recorded at Aberdeen Proving Grounds. C j, ' U,

UNCLASSI FIED :
r
SECUITYl CL.ASSIF'ICATION Of THIS P AGEI (Who Io)* F'--
telo,, ,

% "v
%.-.
• - ,. - .A - ,
 CONTENTS

Page
I. INTRODUCTION. .. ..... ................ 1

II. RELIABILITY AND LIFE PREDICTION METHODOLOGY ..... I

III. MATERIAL CHARACTERIZATION DATA. .. .... ........ 2

Fatigue Data. .. ............ ...... 2
Metallurgical Evaluation and Other Mechanical
Properties. .. ............ ....... 3
IV. FAILURE ANALYSIS .. .. ........... ....... 4

V. STRESS ANALYSIS .. ...... .............. 4

*Bearing Stresses on Spline Teeth .. .......... 5
Torsion of Nonuniform Bar. .. ..... ....... 5
*Evaluation of Stress intensities .. .......... 5
Redesign - Reduction of Stresses in Spline . . . 5

VI. SIMULATION PROCESS AND RELIABILITY EVALUATION DUE TO

% RANDOM LOADING
Random Loading - Amplitude Displacement Model. 6
Fatigue Crack Growth Law for Estimating Torsion
Bar Life. .. ............ ........ 7
*The Monte Carlo Process. .. ..... ........ 8
Fatigue Life - SIN Curve Analysis. .. ......... 9
Torsion Bar System Reliability .. ...... .... 9
Remaining Life - Reliability of Operation After
Specific Number of Cycles .. .. ........... 10
Results of the Reliability Study. .. ........ 10

VII. NONDESTRUCTIVE TESTING CONSIDERATIONS. .. ........ 13

VIII. CONCLUSIONS AND RECOMMENDATIONS. .. ........... 13

*IX. TECHNICAL GAPS. .. ..... ............... 14

X. ACKNOWLEDGMENTS .. ...... .............. 15

APPENDIX A. TORSIONAL STRESSES IN A NONUNIFORM BAR. . . . ....

APPENDIX B. ST. VENANT TORSIONAL STRESSES OF MULTIPLY

CONNECTED REGIONS .. . ................. 3
 I. INTRODUCTION
The M60 tank (Figure 1), which was developed in the 1960's, is still the prin-
cipal Army heavy tank. It is currently undergoing a major "Rebuild" program to
increase its effectiveness. The M60 program manager has expressed concern regarding
* the reliability of various components of the overhauled tanks as a result of failures
* experienced during proving ground tests. In a report to Congress, January 29, 1980,
the GAO expressed a similar concern on the reliability of other systems. This study
* is an effort by an interdisciplinary task group assembled at AMMRC from specialists
in material science, engineering, structural mechanics, nondestructive testing, and
statistical analysis. The charter of the task group was to develop a methodology
for life prediction of structural components in the M60 tank. The torsion bar
* (Figures 2 and 3) was chosen because of the availability of a data base obtained
from a comprehensive research at AMMRC, proving grounds and overhaul depots. It
should be noted that the torsion bar performs like a spring by providing the neces-
sary motion for the wheel to follow the terrain. The M60 has six torsion bars on
each side. The front and rear wheel arms have shock absorbers and stops, while the
intermediate wheel arms have only stops to limit their motion. Torsion bar failure
can lead to reduced mobility and track throwing, or failure to other overloaded bars
and eventual total loss of mobility of the tank.

The task groups made visits to manufacturing facilities, the overhaul depot,
field test sites, AMSMA, and TACOM in order to obtain data and background information.
Detailed failure analysis metallurgical examination, material testing, NDE, detailed
stress analysis, and a comprehensive probablistic analysis were completed by the
group. The paper will discuss in detail the above studies, including major findings
2 of a sensitivity analysis, and the final conclusions with regards to design improve-
ments, material requirements, and the feasibility of predictive methodology for tor-
sion bars. The methodology used here circumvents the present deterministic approaches
* used in establishing design life. It simulates variability of loading and material
by introducing Monte Carlo methods. 1

II. RELIABILITY AND LIFE PREDICTION METHODOLOGY

*The approach taken to developing life prediction methodology was basically to
* obtain as much data about the torsion bar as the program would allow, analyze the
data, perform the reliability study, and make recommendations.

The M60 task group visited various manufacturers of the torsion bars and obtained
quality assurance procedures, test results, manufacturing procedures, and machining
methods (General Dynamics and FMC). The developer, the Army Tank-Automotive Command
* (TACOM), provided the task group with the requirements and specifications, Depot
- Maintenance Work Requirement (DMWR) 2 described the user experiences. Aberdeen Proving
* Ground (APG) test facilities and the Materials Testing Directorate provided spectrum
load data from a specific course. They also carefully monitored bar failures. The
Anniston Army overhaul depot described inspection procedures information, NDE, and
replacement requirements. A major part of the material data generated, involved
mechanical and various metallurgical properties, and their relation to the speci-
fication. These properties were then incorporated with the data bank obtained from
the QA testing of the manufacturer and published data on the material.
1. NEAL, D. Md.,and MASON, D. S. Determination of StructuralReliability Using a Flaw Simulation Scheme. Army Materials and
Mechanics Research Center, AMMRC TR 81-53, October 1981.
2. Suspension System Components for K60. Depot Maintenance Work Requirement, TACOM, DMWR9-2350.253-1-S.

.5 1
 The methodology then required accurate analysis of the mechanical loading, stresses,
fatigue, and fracture evaluation of the torsion bar.

Finally, all the data is incorporated in a Monte Carlo method in order to

determine the reliability of the torsion bar. This method requires modeling of the
spectrum loads and the material fatigue life with respect to crack growth or stress/
cycles to failure (S/N). The reliability of a single torsion bar (or system) can
then be studied under various assumptions of terrain, speed, improvement in designer
material, inspection procedure, replacement period or any combinations.

III. MATERIAL CHARACTERIZATION DATA

Several sources were used in obtaining material properties. These covered QA

data obtained from the manufacturer, mechanical and metallurgical tests performed at
AMMRC by the task group, published material data, and discussions with the other
torsion bars manufacturers.

The M60 torsion bars are currently being manufactured by General Dynamics (GD,
Scranton, PA) and Machine Products Co., according to MIL-5-45387A, from AISI 8660
steel. It should be noted that the bar specifications do not have an explicit
requirement for fracture toughness or Charpy impact energy and the only requirement
is placed on Rockwell Hardness. The metallurgical data reported here was obtained
from the failed torsion bars obtained from Aberdeen Proving Grounds (APG) and
Anniston Army Depot.

Fatigue Data

As part of the quality assurance (QA) requirements, several torsions bars from
* each heat were tested. The torsional test machine is similar to that shown in
Figure 4. The QA data obtained from GD, shown in Table 1, was that of the number of
cycles to failure of the torsion bar for a specified prescribed angle of twist (420)
which is based on the maximum allowable torsional shear stress of 140 ksi. Table I
also shows the cause and location of failure in each test. This data is represented
*by the probability of failure versus number of cycles to failure (Figures 5 through
8) for the individual ears 1978-1982. The Kruskal-Wallis multi-sample test for
identical populations, showed that there are no significant differences among
individual yearly fatigue test results, therefore, the data was pooled together.
The data shown in Figure 9 is the accumulation from test results on the various
heats of production from 1977-1982. The 1983 data were not included as they represent
* a new bar modification.

The QA specification requires that three bars be tested for every heat with at
least 45,000 fatigue cycles obtained prior to bar failure. The approach currently
used by the manufacturers is to run the test until the torsion bar fails (Table 1).
In many instances the fatigue test machine has failed, due to the excessively large
number cycles required resulting in an increased cost for quality control. In order
to overcome this difficulty a statistical procedure was applied using a Weibull
censored data analysis4 approach. This procedure obtains the design allowable

3. HOLLANDER, M., and WOLFE, D. A. Nonperametric Statistical Method& John Wiley & Sons, New York, 1973.
4. LAWLESS, J. F. Contruction of Tolerance Bounds for the Extreme Value and Weibull Distributions. Technometrics, v. 17, no. 2,
May 1975.
S."

'S.
"-: - I 1
•% - . .'- .. -'." ,;- ; , ",4.
. ",. . "-. " . ". . " ". .-
. "- ..-. ,-,".".. .. . .,.." .. , ", -- .' ,-.
values for failure data plus run-out values. The run-out values are non-failed bars
with only the information that they have not failed at a specified number of cycles.

Using the QA data from GD it was determined that the maximum number of required
cycles can be reduced to 110,000 as compared to the maximum of 570,000 currently
being required to cause failure of the most durable bar. A comparison of material
design allowable values obtained from both complete sample (all data) and censored
data (only failed data below 110,000 cycles) showed excellent agreement. The allow-
able represents a value determined from a specified probability of survival with a
95% confidence in the assertion. Survival probabilities are 0.99 for the "A' allow-
able and 0.90 for the "B" allowable.

The above indicates that equivalent information could be obtained with a large
saving in time while reducing fatigue machine failures. Figures 10 and 11 show the
censored data results.

Metallurgical Evaluation and Other Mechanical Properties

A metallurgical evaluation was conducted on two failed M60 torsion bars, one
produced by General Dynamics (GD) and the other by Machine Products Company (MPU).
The parameters addressed in this evaluation are:

o chemical analysis
o light microscopy
o retained austenite
o hardness traverse-spline and body
o tensile properties - room temperature (RT)
o Charpy impact transition data longitudinal (LR) orientation
o Charpy impact - (RT and -400 C), transverse (TR) orientation
o fracture toughness (RT and -400 C) (LR and TR orientations)

Table 2 contains a near-complete chemical analysis for the GD material plus a

carbon and sulfur content for the MPU product. As can be seen, the GD bar conforms
to the specifications for AISI 8660, which is also shown in Table 2. It also meets
phosphorus and sulfur requirements (0.040% each) in MIL-5-45387BAT.
Table 3 shows the results of a Rockwell C hardness (HRC) traverse in both a
spline and body location for a GD material and the spline location for the MPU bar.
It can be stated that both companies meet the hardness requirements of HRC 47-51 per
MIL-5-45387BAT.

Tensile data from the GD bar are shown in Table 4. Duplicate tests were con-
ducted and the data are consistent.

Longitudinal and transverse Charpy V-notch energy data are tabulated in

Table 5. Longitudinal tests were conducted over the temperature range of +240 0 C to
-400 C and transverse tests at +220 C and -400 C. A comparison of the data at room
temperature and -400 C shows that the longitudinal energy values are superior to
those of the transverse orientation. Values are 7.7 and 5.5 ft-lb and 2.7 and
2.9 ft-lb, respectively.

Fracture toughness data (KQ) are tabulated in Table 6. These values were
obtained using slow bend specimens at room temperature and -400 C in both an LR

................-............................... ... .
 and TR orientation from a GD bar and room temperature TR orientation from an MPU
bar. Room temperature LR values are higher than TR, 43 versus 38.3 ksi lrn. GD TR
data is slightly higher than MPU, 38.3 versus 34.2 ksi1 in. Two short rod tests
were conducted on MPU material with resulting average KQ value of 35.5 ksi1'Tn.

Microstructural and retained austenite studies were also conducted on the GD

material. The microstructure consisted of tempered martensite and 3% or less of
retained austenite.

IV. FAILURE ANALYSIS

In addition to the metallurgical investigations which were performed on failed

torsion bars, fractographic examination was also performed. One of the basic prin-
*ciples of failure analysis is the preservation of the fracture surface in a condi-
tion as close as possible to that at the time of failure. Another principle is the
compilation of historic data concerning the failed component. The first principle
issue was not achievable since the available torsion bars were found in scrap bins
at either the depot or testing grounds resulting in fracture surfaces with some rust
on them. The second issue was resolved by using a statistical model for the random
loading spectrum, this will be discussed later.

Figures 12a through 12d show typical field failures which were obtained from
Anniston Army Depot, APG, Ft. Knox and GD, which exhibited similar failures. They
obviously initiated within the splines at the ends of the bar, and in most cases at
the wheel arm end. One exception is a case observed at APG where the failure was at
the anchor block end. These observations are supported by 6 case studies obtained
from the Measurement and Analysis Division at APG9 and GD.

Light microscopy and scanning electron microscopy (SEN) were used in the examina-
tion of the fracture surfaces. The failed bar in Figure 12a was cut open and examined.
The arrow shows the origin of the fracture. The arrow on Figure 13a shows the surface
of the serration which experienced surface plastic flow at the initiation site.
This indicates high stresses at the inner end of the splines. There is a slight
indication of fatigue marking in Figures 13b and 13c, but not as large as usual.
The SEM examination (Figure 13d), however, did not show the usual striation beach
marks associated with fatigue crack growth. This indicated that once fatigue crack
initiation occurred, the propagation stage was very short. Further away from the
initiation site, the fracture surface showed signs of corrosion and hydrogen
embrittlement. It should be noted here that the fracture surface was old and cvuld
have been contaminated by the environment after failure.

V. STRESS ANALYSIS

The stress analysis of the splines was performed to evaluate various probable
modes of failure. For the spline teeth, the possible modes of cracking are longi-
tudinal, inclined, and transverse, with respect to the axis. Such cracks (Figure 14)
are associated with various stress states, where the predominant ones are torsional

5. Aberdeen Proving Ground Report, Case Nos.: 78-M-4 - Comparison Test of M60AI Rise Yanks; 79-M-9 - Comparison Test of M6OAJ
Rise Tankv 8OM-34 - Failure Analysis of M6OA3, 82-M457 - Andysis of Torsion Bar Spring Failures; and 82-M-32 - Failure Analyis of
Nv M60AJ Torsion Bar SpringL Masurement and Anal. Div., Phys. Test Branch, STEAP-MT-G, Aberdeen Proving Ground, MD
6. Rebuilt Tank Reliability Analysis, General Dynamics, Land System Div. M60, MG-80-01731-006, March 1982.

94

" .. . . . . .. .
shear stresses, Hertzian stresses, and bending stresses. These stress states are
three dimensional in nature and therefore the analysis could be very expensive unless
some simplifications are made. Three types of stress models were chosen to simplify
the problem. The final results were then combined and used in the failure evaluation.

Bearing Stresses on Spline Teeth

Due to the existence of a missing tooth in the spline, it is suspected that the
highest stressed area in bearing would be in that location. The bearing stresses
were evaluated assuming that the load on the missing tooth is distributed equally on
both neighboring teeth (Case 1) or only on the forward tooth (Case 2). The analysis
was carried out using substructuring and superposition of the two load cases.

The stresses produced by the contact forces depend on the number of teeth in
contact, and the distribution of contact force along the teeth. The number of teeth
in contact depends on the tolerances, misalignment and load, while the distribution
of the contact force depends on the teeth shape and moments of inertia of the axle
and block. Figure 15 shows the finite element model used. Figures 16 through 21,
the stresses in ksi resulting from bearing forces are shown.

Torsion of Nonuniform Bar

The enlargement of the ends of the torsion bar leads to stress concentrations
at the fillets. The solution of this problem was carried out using the finite element
method; a heat transfer analogy of the classical Michell torsion problem. Appendix A
gives the details of the thermal-mechanical analogy and the boundary conditions used
in the analysis. The finite element mesh and the results in ksi of this analysis
are shown in Figure 22.
Evaluation of Stress Intensities
Four crack locations were considered in the fracture analysis:

1. Surface crack in the shaft (Mode III)

2. Subsurface crack in the shaft (Mode III)
3. Surface crack in the spline area away from the missing tooth (Modes I and III)
4. Same as 3, but at the missing tooth (Mode I)

The Mode I analysis used the results of the finite element model in Figure 15.
Mode III results in ksi of Cases 1 to 3 were obtained by the solution of a St. Venant
problem (Figure 23). The finite element penalty method was used for the multiply
connected region case of subsurface cracks (see Appendix B). In all the above
fracture analyses the quarter-point crack tip elements were used. 7

Redesign - Reduction of Stresses in Spline

As demonstrated from the failure analysis and from the bearing stress and fracture
analysis, the splines are the most highly stressed region in the torsion bar. A
reduction of the contact stresses can be accomplished through a more uniform distri-%

7. BARSOUM, R. S. On the Use of IsaparametricFinite Elements in Linear Fracture Mechanics. Int. J. Num. Meth. Eng.. v. 10.
1976. p. 25-37.

5
 bution of the stresses by redesigning the splines. Figure 24a shows the bearing
stress distribution from current spline design (uniform teeth). Using tapered teeth
permits intentional redistribution of the contact stresses between the splines and
end blocks. From the deformation of the splines, as determined by the finite element
analyses in section V - Bearing Stresses on Spline Teeth and Torsion of Nonuniform
Bar - it was calculated that a 0.005 in. tapering increases the contact zone
and reduces the maximum bearing stresses by more than 25% as shown in Figure 24b.
Machining of such tapered teeth is accomplished using a gear shaper.

VI. SIMULATION PROCESS AND RELIABILITY EVALUATION DUE TO RANDOM LOADING

Random Loading - Amplitude Displacement Model

There is a lack of load spectrum data for the M60 tank. The only available data
was that of an MI tank test conducted at APG is shown in Figure 25. The test track
contained a series of artificial ramps and bumps (Belgium blocks), to simulate a
specific test course (Figure 26).

The angular amplitude distributions of three bars from these tests is shown in
Figure 27. The positive and negative angular displacements of the bars as a function
of tank travel are shown in Figure 27a. The zero angular rotation is referenced to
the static deflection of the torsion bar and hence does not give zero stress in the
bar. Figure 27b shows the amplitude distributions in a manner describing percent
time less than a specific value (positive sign (+)) and percent time greater than a
specific value [negative sign (-)I, e.g., 25% leve: 'quals a -75% level. The positive
peak represents maximum angular displacement uvry.:er load; the negative peak is the max-
imum rewind angular measurement. The range of angular rotation is defined as follows:

Ae - 6 + 16-1
where e - maximum negative angular displacement, and (1)

e - displacement from Figure 27.

The Beta distribution provided the best representation of the skewed amplituzde
distribution. The dampening effects that occurred under load resulting from a stop
used in preventing further angular twist of the bar produces highly skewed discrete
cumulative probability values. The Beta function is defined as:

f(Ao) - r(P)r(Q) (AO) )Q(2)

O<AO<l P,Q>O

The P and Q values are selected in a manner that provides the best probability
density function (PDF) for representing the data. Figure 28 describes a typical
distribution and Table 7 shows the excellent fit between predicted (Beta representa-
tion) and actual test results. As shown in the fatigue analysis, section VI. Fatigue
Life - S/N Curve Analysis, angles less than 200 represent stresses sufficiently low
that infinite torsion bar life could be expected, therefore, a good representation
below this angle is not essential.

.. _. ..
.!Z , ° . . - °. . . .. . . .. . . .
_.-.*-°**o**, o.... .. .. ° . . . ....
. - . . . .
... ; ,, i
wal,= ii i /iil,1i •. , **.-.-....... _ ••,•.....•
 Figure 29 shows a typical load spectrum result of an M60 A3 torsion bar. The
results were obtained for a DADS* simulation process performed at TACOM. A stochastic
model was introduced in order to simulate the Lar response. The spectrum is repre-
sented by positive and negative angular displacements versus time, as shown. Included
in the figure are tabulated absolute amplitude values using Equation 1.

* The uniform probability density function was selected to represent the amplitude
distributions. The selection process involved evaluating an informative quantile (IQ)
plot8 of the amplitude data. In Figure 30a, the IQ plot of data tabulated in
Figure 29 is shown. The straight line represents an exact uniform distribution with
the disjointed line representing actual data. Figure 30b is a schematic of a typical
distribution model for DADS simulation data.

*Fatigue Crack Growth Law for Estimating Torsion btr Life

Initial efforts in applying the Monte Carlo method for determining reliability
versus cycles to failure of the torsion bar involved using the crack propagation laws.
The da/dN data requires extensive testing, time, and material. For this reason,
and due to the short life expended in crack propagation because of the operating
stress levels, it was decided to rely on published results. As shown in Figure 31,
the percentage of life spent in initiating and propagating a crack depends on the
stress level and material toughness. The da/dN relationships for materials metal-
lurgically similar to the specified AISI 8660 material were obtained from References
9 through 11, and are shown in Figure 32. The dry air results made available by
*" Barsom I0 provided the most representative estimates of crack growth versus stress
*. intensity (AK) since the torsion bar is protected from the environment. It is to be
noted that the mean stress effect was ignored in this analysis. From the basic
da/dN relationship, N cycles to failure as a function of crack growth, angular dis-
placement, and the geometry of the region where the crack initiates in the bar nay be
obtained from the following relationships:
af
"N
da
N a i 0.66 x 10-8 6K2 .2 5 (3)

where AK= A.A8/-a

with Aj's (see below) representing the stress intensity multiplier associated with
each configuration of the cracks discussed in the previous sections. The AG is the
range of angle of twist.

A1 = 4.91 (missing spline tooth)

A2 = 3.29 (other spline regions)
- A 3 = 3.26 (torsion bar body).

- *DADS is a vehicle dynamics computer simulation program used by TACOM's concepts lab. It should be noted that the authors are not
convinced that the M60 suspension system is accurately represented in the DADS program. Therefore. the results in Table 9 should
-" not be taken as absolute values, but as demonstrations of the methodology.
8. PARZEN, E. Entropy Interpretation of Tests for Normalit, by Shapiro-Wilk Statistics Presented at the Twenty-Eighth Conference on
_ the Design of Experiments in Army Research, Development and Testing, Monteray, CA, 20-22 October 1982.
9. Damage Tolerant Design Handbook. Metals and Ceramics Information Center, Battelle Columbus Laboratories, MCIC-HB-01, 1975,
p. 8.2-D.
10. Damage Tolerant Design Handbook. Metals and Ceramics Information Center, Battelle Columbus Laboratories, MCIC-HB-01, 1975,
p. 8.2-E.
II. BARSOM, J.M. Transactions of ASME. J. of Eng. for Ind., Series B, v. 93, no. 4, November 1971.

:-_N

-S..
 The range in stress intensity factor AK was calculated by combining Modes I and
III at locations I and 2 on the spline (see section V, Evaluation of Stress Inten-
sities) through the use of the strain energy release rate definition. The fracture
toughness in Mode III was taken to be \1(1 - v2)/(1 + v) multiplied by the Mode I
toughness.

Note, a percent reduction in ai's will provide a decrease in the stresses in the
specific region of the torsion bar. The ai and the af parameters are initial and
critical crack size, respectively. The af is obtained from the critical stress inten-
sity value KIC for the material considered,'36 ksi ir . in this case. The angular
displacement of a bar A9 can also be represented by the equvalent stress range as

AT = rG (AO)/L where
r - radius of shaft,
G - torsional modulus, (4)
A6 - angular rotation range, and
L = length of torsion bar.

The Monte Carlo Process

Crack Propagation Analysis

A schematic of the Monte Carlo process is outlined in Figure 33 for determina-

tion of frequency of occurrence versus cycles to failure of the torsion bar using the
crack propagation law. An assumed normal distribution is used to represent variabil-
ity in the Aj, ai, and af parameters. A coefficient of variation (CV) defined as:

CV = SD (5)
mean
establishes the standard deviation (SD) for the corresponding known mean value (e.g.,
a i for initial crack size). Coefficient of variation values of 5, 10, and 15 percent
were considered in developing the distributions in order to examine the effects of
variability (inherent errors in measurements, flaw size assumption or the stress
analysis) in the parameters. By selecting the above CV's, a sensitivity analysis
can be developed, thereby providing a method for recognizing the importance of the
* parameters as related to the cyles to failure. The 60 Beta distribution, as shown
in Figure 33, has been previously defined in Equation 2.

The random numbers used in the Monte Carlo process are obtained from solving
for X in

x
(6)
f fl dx - R

where R is a uniform random number and fi corresponds to the desired type of fre-
quency distribution for the parameter. A probability density function for the N
cycles to failure can be obtained by randomly selecting discrete sets of numbers
from the ai, af, Aj and AG distributions and substituting them into Equation 3.
Note, there should be an equal amount of random numbers for each parameter to have
the proper amount of numbers for the N distribution.

.. . 2 " ". Y.
"" . . "."-" .. ".". "."- " . •- - - £ , . . "'' '. ° %
 Fatigue Life - S/N Curve Analysis
Torsional bar life expectancy was obtained using the Monte Carlo process applied
to the SIN curve relationship. The above procedure provided a method for obtaining
lifetime estimates of the bar by combining the effects of crack initiation and
propagation. A description of the S/N curve is given in Figure 34, where the base
line data was obtained from a literature survey for material metallurgically similar
to the torison bar material. The survey provided a set of SIN curves for torsional
fatigue, shown below, for best representing the current materials used in the bar.

* Logl0 N =B + 0.068 AG

where B =7.70 for base line data.

The slope value of 0.068 was essentially the same for all curves in the set.
The adjustment in B from 7.70 to 8.06 is made on the basis of M60 torsion bar quality
assurance tests at a single tie value performed at the Scranton manufacturing facility
* (see Figure 6). A single load equivalent to a 420 angular displacement was applied
* during the quality assurance torsional fatigue test. Use of the mean value and the
cycles to failure in Figure 6 provided a more accurate estimate of B. The curves
representing a range of 10 and 20 percent reduction in bar stress are also shown
in Figure 34.

* The S/N curve Monte Carlo process is similar to the previously outlined method
* for da/dN relationship. The primary difference involves using models for B and A9
from Figures 34 and 27, respectively. A schematic of the basic SIN representation
is shown in Figure 35a. The simulation of SIN curve variability is shown for a
specific value in Figure 35b. Figure 35b also describes the probability density
function (PDF) for B. A random selection of a discrete set of numbers from A9 and B
distributions is then applied to Equation 7 in order to obtain the Logl 0 N value.
The process is repeated until all values from the two distributions are selected.
This process will then provide a PDF to represent Logj0 N. Again, in the SIN
evaluation, the mean stress effect was ignored for torsional ljoading.12

* Torsion Bar System Reliability

* By assuming a tank with m torsion bar systems, the following procedures would be
applied in order to establish the reliability of the system. After the reliability
for each torsion bar failure is calculated, the next task is to determine the overall
suspension system reliability.

If all of these torsion bar failures are independent of each other then the
system reliability RS would be given by

* m
RS Hi Ri (8)

where Ri =reliability for torsion bar i, i 1, 2...m, and ms total number of

torsion bars in the suspension system.

*12. rORREST, P. G. Fatigue of Metals. Persamon Pres, New York, 1962, p. 103.

- - % **~
 If these torsion bar failures are not independent of each other, then the system

reliability would be given by

RS - R1 X R2/R X...X R lal

l ..l (9)

where

Rm/R-I/.../R 1

is the reliability of torsion bar m, given the reliabilities of m - 1 ..., 1

bars.

It should be noted here that the reliability of the individual bars Ri is

different due to the difference in load spectrum it experiences.

Remaining Life - Reliability of Operation After Specific Number of Cycles

When the torsion bars of a rebuilt M60 tank are reused in service after a Depot
inspection, one would assume that a specific number of cycles on the bars could be
found. Currently, no information is recorded and, hence, the following is a hypo-
thetical evaluation. The reliability of operating an additional number of cycles
when a specified number of cycles of operation has been completed is obtained in the
following manner. Initially, it is assumed that a specified distribution function
such as f(N) is known, for example, the distribution of Logl0 N from the Monte Carlo
method previously described. The reliability R(nl, n) is a conditional probability
requiring the probability of operating for (nI + n) cycles when n1 cycles have been
completed. That is,

S.0

y
f f(N)dN
R(n 1 + n) n1 + n
R(nln) 1) f0f(N)dN (10)

where n is the additional mission in cycles after nI cycles of operation. the number
N. (nI , n) of components (torsion bars) that will survive additional n cycles is
given by

N. (nI n) Ns(nl) • R(nI n) (11)

where N,(n I ) - number of components starting the mission of n additional cycles.

Results of the Reliability Study

The proper number of simulations for the Monte Carlo method depended on the
models under consideration. For example, 5000 and 3000 were required for the da/dN
and S/N curve models, respectively. Using a convergence rate criterion for the
calculated one percent valueb (Ps (see Figure 36) and recognition of the third
and fourth moment stability of the LoglON distribution provided an excellent method
for determining required number of simulations. Differences in percentile values

10
j26!
 for coefficient of variation (CV) of 10 and 15 percent were minimum for da/dN
simulation of parameters. The 10 percent value was used for all da/dN calculations.

The torsion bar reliability results from the da/dN relationship are given in
Figure 37. The current design results were obtained from Equation 3, with A2 =3.29,
* which represents the stress state in the spline region. They indicated a relative
* limited lifetime range of 14 to 500 miles, with probability of survival (PS) values of
0.99 and 0.01, respectively. An appropriate increase in af (final crack length) from
Equation 3 represents the 40% increase in the KIC value. This represents an improve-
* ment in the material's capability with respect to acceptance of larger flaw sizes
prior to failure. The slight improvement in the bars capability indicates that an
improvement in material will not significantly improve bar performance. The 25% and
50% reduction in KI (stress intensity) in Figure 37 is obtained from reducing A2 in
Equation 3 by the respective percentages. These reductions represent improvements in
the design of the spline section of the bar as shown in Figure 24. The KII1 failure
in the shaft represent situations where failure occurs in the shaft rather than in
the spline region.

The maximum life of 70 miles at 25 mph achieved from a 50 percent improvement

in spline design with a 0.99 probability of survival indicates that there is very
limited life of the bar after crack initiation. Table 8 describes minimum life
* estimates (99 percent survivability) for crack, e.g., propagation of the torsion
bar with respect to various tank velocities and the design improvements. Tank
* travel at 5 mph (lowest speed) with a 50% reduction in K1 value shows propagation
life expectancy of only 341 miles at 0.99 probability of survival.

The failure probability obtained from the S/N curve - Monte Carlo application
* is shown in Figure 38. The resultant exponential form is consistent with that
expected from the SIN model in the analysis.

A graphical display of PS versus miles to failure is shown in Figure 39 for the

* 25 mph tank velocity. The life expectancy of the bar is much greater than that
obtained from the da/dN analysis. The minumum life estimate (0.99 PS) of 292 miles
is 21 times greater than the 14 miles determined from the da/dN results for crack
propagation. This result indicates that most of bar life occurs prior to crack
initiation. Therefore, the torsion bar should be manufactured in such a manner that
* flaws are minimized. The current shot peening used in the manufacture of the bar
indicated recognition of this fact by the manufacturer. The bar reliability estimate
- obtained after an assumed 741 miles of tank travel are shown in Figure 39. The in-
crease of the probability of survival from 0.90 to 0.99 if the bar survives the
initial 741 miles does not provide a sufficient gain to warrant reusing bars since
the minimum increase in expected life is reduced very rapidly. The results
- (Table 9) for a 20 percent reduction in design stress gives 865 miles for a PS of
* 0.99. This is a considerable improvement when compared to that of 292 miles for the
* current design. The results from velocity ranging from 5 mph to 25 mph in increments
* of 5 mph are given in Table 9 with respect to current 10 and 20 percent improvements
in design. Reducing velocities of tank operation obviously improves reliability of
the torsion bar.

It should be noted that the field test data used in the comparison with the
* reliability calculation refers to the failure of the first (or front) torsion bar.
The front wheel arms have a different suspension systemthan the rest of the wheel
arms, and hence the spectrum of loading on the torsion bars is quite different. They

%1
 are subjected to higher loads and have a higher incidence of failures. The same
reliability calculations can be performed on the other bars in order to calculate
the reliability of the system (Equations 8 and 9).

Examination of current design mileage capability of the bar for 20 and 25 mph
indicates a range from 276 to 292 miles. These results agree with the 262-mile
minimum life obtained from Aberdeen Proving Ground (APG) test results, Table 10
(taken from Reference 13 of bar failure from 3-mile test course), which are repre-
sented in the probability distribution of Figure 40. This course and tank velocity
were similar to those used in obtaining the load spectrum results. The excellent
agreement between the predicted and actual life expectancy of the bar indicates the
desirability of the Monte Carlo process for modelling variability of loads spectrum
(design stress) and SIN curve (material capability) results. Table 10 gives the
failure results of tests run at APG. The failure life of the torsion bars are in
agreement with the above predictions.

Although excellent agreement has been obtained, the authors would have preferred
representing the spectrum load consistent with an individual peak-to-peak angular
displacement. The simplification applied using the negative peak as base and repre-
senting the displacement relative to this value was a good approximation to the
available individual displacements'. This approximation would provide a slightly
conservative estimate in the reliability values. Using the ASTM recommnended pratice
of representing the lower of 3 times the standard deviation band of the S/N curve
as a measure of material fatigue loading capability combined with maximum angular
* displacement (46 degrees) for 25 mph, a minimum life estimate of 112 miles resulted
for the bar. Selecting this number as a design value could result in an overly
conservative estimate, since chance that this maximum displacement could occur to-
gether with the SIN curve as the actual lower band, described above, is extremely
small.

A minimum life of 575 miles was obtained by using the maximum displacement
* value with the original S/N curve where B = 8.06. This result is obviously wrong
* since in the limited samples of 23-bar failures, two of them failed at mileage less
than 400 miles (see Figure 40).

Results from the application of the Monte Carlo process for DADS simulation
data is shown in Table 11. The PS percentages and their corresponding miles of tank
* travel are tabulated for road arms IL, 2L, and 6L at 5, 18, and 25 miles per hour.
The APG12 course (18 mph) was the most severe according to the TACOM representative.
The results from Table 11 reflect this, since the 0.99 percent PS of 14 miles was
the lowest obtained from any of the computed life estimates at that PS value.*

Since this course introduces unusually large angular displacements of the bar
IL (see Figure 29) a'nd requirements are only one failure in a hundred, this is not
an unrealistic estimate of bar life. It is obvious from examination of Table 11
that road arm 6L will survive much longer than either IL or 2L. Traveling at 5 mph
in such a severe course will also increase bar life considerably.

*See footnote on pare 7.

13. Aberdeen Proving Ground Report, Case No. 82-1-37, MT-5376, Aberdeen Proving Ground Test Branch.

12

% %. %~% .!~ ****~. %

 VII. NONDESTRUCTIVE TESTING CONSIDERATIONS
Review of torsion bar field failure reports and examination of a number of
fractured bars clearly illustrated that fatigue failures predominately initiate in
the spline ends rather than in the reduced cross-section central gage length. Based
on these data, it was initially considered that a periodic magnetic particle inspec-
tion for fatigue cracks in the spline surfaces would provide a technique for pre-
dicting premature failures. However, results of the subsequent fracture mechanics
analysis reported herein have identified the critical crack size as 0.019 inch.
Furthermore, the analysis shows that an existing fatigue crack, 0.005 inch deep, will
propagate to critical size within only a few miles. Therefore, if a fatigue crack
equal to or less than 0.019 inch exists, failure is imminent.

Figures 41a and 41b taken from Reference 14, show that a fatigue crack in steel
must be greater than 0.030 inch for a 90% probability of detection at 95% confidence
using the conventional magnetic particle inspection method. Significantly smaller
cracks are indeed detectable using the technique, but probability of detection will
be low in a production environment, as illustrated in Figures 41a and 41b. Inasmuch
as the torsion bar must be removed from the vehicle to conduct an inspection at some
arbitrary time, detection of a crack less than 0.019 inch would be only by chance.
It was therefore concluded that, in the case of the M60 torsion bar, even the most
sophisticated nondestructive crack detection scheme will not provide an effective
life prediction technique. Once a fatigue crack initiates and can be detected,
service life of the torsion bar has essentially ended.

VIII. CONCLUSIONS AND RECOMENDATIONS

1. A methodology for obtaining reliability of the M60 tank torsion bar sub-
* jected to cyclic random loads has been developed where probability of survival is
*represented as a function of miles of tank travel.
2. As discussed in Reference 15, the developed methodology can be applied to
other structures with cyclic random loads.

3. The use of the method appears justified from recognition of the excellent
agreement between predicted reliability estimates and those obtained from the
actual bar life (miles to failure) experienced during the tank operation.

4. Determination of minimum bar life was 21 times greater from the application
of the S/N curve model than that of the assumed da/dN model. This indicates that
most of the bar life is expended in crack initiation.

5. Application of deterministic procedures [use of lower three SD bound for S/N

curve (ASTM method) and mean S/N curve] provide over and under design value estimates
respectively while the Monte Carlo method outlined in the text accurately describes
acceptable design values.
'.5

14. RUMMEL, W., et al. Detection of Tghtly COsed Flaws by Nondestnictive Testihg Method In Steel and Tfmielum. Martin Marietta.
Denver, CO, Finl Report MCR-76-476, NAS 9-14653, September 1976.
15. BARSOUM, R. S., et al. M60.Torson Bar RelbbUilty. Army Symposium on Sold Mechanics, Newport, RI, October 1984.

e de

" ''...''€ ' C. .. 'v" "-::. v".'." ;.'. 131. . -". .-.. '- .... ".'.'. €' ~ '
 6. Calculations show that redesigning of the bars to reduce the stresses in-
creases the life considerably. A slight improvement of 20% is much more effective
in increasing life than requiring a higher toughness in the material.

7. Explore the use of the torsion bar presetting operation at the depot as
a proof test for reused bars, and follow it by 100% magnetic flux inspection. Also,
explore reshot peening of the whole bar, or replacement of the pinion and anchor
blocks at the depot.

8. Recommend revisions to MIL-S-45387A by requiring the use of aircraft

quality material and including requirements for grain size, cleanliness, and Charpy
V-notch impact energy for both room temperature and -400 C. Such would
longer life.

9. For new bars, it is recommended that materials other than AISI 8660, such
as AISI 4350 steel be considered. Also, induction hardening followed by shot
peening should be explored. References 16 and 17 cover this issue to some extent.

IX. TECHNICAL GAPS

In addition to the above conclusions and recommendations the study uncovered

the following "Technical Gaps" which have to be resolved in future studies of this
type.

I. Actual load spectra - the load spectra on actual test courses and other
field ipplications is not available. The design of new or improved suspension
syste -ill require this knowledge.

2. Data from various manufacturers - torsion fatigue data was only available
from GD. Other manufacturers and potential suppliers of torsion bars had no data.
" In addition, the form of the data generated needs to be modified based on modern
statistical methods as discussed in his report.

3. In general, there is a lack of fatigue and fatigue crack growth data for
materials used in torsion bars. In addition, environmental, finishing, and mean
stress effects on fatigue life are very difficult to obtain.

4. The effect of shot peening and hardening on fracture and fatigue life needs
an extensive study.

5. Currently, there are no NDE methods for detection of small cracks. Also,
NDE methods for field inspections are needed.

6. In order that bars which exhaust their fatigue life can be retired, a method
of mileage record keeping for every bar is needed (this could be expensive). Micro-
processors or mechanical angular displacement accumulators would be helpful for such
a task.

16. GIOON, D., DOERMAN, G., and GORUM, A. Ha StreA Toon r Delbpmeont fr Army Applcaone. Contmct
DAAG46-76C.0074, FMC Corp., San Jos, CA, FWin Report, AMMRC TR 79-5, January 1979.
17. High Strength Steel Tonwon Barx TACOM, R7D Tesh. Report No. 12746, 30 May 1983.

14

e'N
 7. Maintenance procedures need to be revaluated in lieu of failure life of the
component. In addition, the interaction between various components should be con-
sidered in their retirement.

8. Introduction of the censoring process should be made in order to reduce the

amount of fatigue cycles necessary for QA of the bars. This will reduce the cost
and time in the testing process.

X. ACKNOWLEDGMENTS
The author would like to acknowledge the following people: BG C. B. Donovan
(PM-M60 at time of study), R. DeGrout (TACOH-PM-H60), A. Cummings and COL J. Aljets
(APG), H. Briggs (GD), G. Boerman (FMC), F. Bibb and J. Bennett (Anniston),
J. A. Robinson and R. A. McKinnon (APG-HTD) and ILT G. Jackson and R. Musseau
(TACOM-Concepts Lab).

', 15/

>-.-

*.%P P
.

.4 *-
"<*:*:d 44~. dP~.. ~. V~'<8

'.%
.1

-..-..- , ..
. .

Figure 1. M60 tank.

EGPI $ACEI
LEG2ND I2 PLuG
13 SCREW
. DSCRPMOd 14 LOCKWASH.E
G AGf 16o v et
Is OVE,
2 NOUSING SEMBLY 17 IN
3 PIN a GASKET
4 NOuSING 19 ANCMO
PLRUG
SEAL 2I
scuiWE
LOCKWASENS
"7 GASKI 22 RUG
I LOCKWASHM

-. 0I TORSION SAR

* Kl

I..° . '
IN I.

~~Figure 2. Numbers 1 through 6 roadwheel support housing, torsion bar, and related parts.

21 7 1
i
,..: %;' ;.,Fgure.2 1 throug 6i
Numbers=.3"
,i~ii~ii i e Iod
supr huig trio bar an reae partIs.

-- ".."
 IN -M
"" M %2 0 TI N 9- MV---. " M . * - ~ . ..

-.

r - IS
pea sw" so-m ~s weysI

pf -LN

0.

Figure 3. Torsion bar and tooth detail.

Figure 4. Torsion bar endurance test fixture (shown with scatter shield removed).

18
 ,li 0

.. w~..- u'

03
".tem

,Ono -i0l

i 2 - .I
PI
: L09 Base 10 Cycles to Failure

. I.malm Figure 5.m M60

n i torsion
n bar fatigue test results (1978).

"" 19 ,
- 1.aS
P S 2~-.a~ a -' - - - - .-. . a ~ ~ . - . A -30

04 0
3c4-

P-0 -4
0 un

r4-

0)
41 '

20 .
 SlUESS
M5h £WOVALUATION P0
,0
- ______ _____ _____

6.3- // '

p -
a
0

0.l-
L
U
| A"
Ir e°4 . . ... . .

Norm I

*'*-~ '"-
---------
Red,'.'

4.6 4.1 5.0 5. 5.4 5.6

Log Base 10 cycles to Failure

Figure 7. M60 torsion bar fatigue test results (1980).

21
 itI 'I '

.5-". .- ..
.5g

a.-a. a

-- -
LI II N.IIII..I-
iG I4)II
U,5- 1
__ __ _ _ _ _ _ _ _ _ _ __22

"
".o , ,.'.¢2 .'.
g.-'.x¢;.:¢
.%;e,;JS'.2,.',;.."..'.-.,q
Q .. , ,'2e'€ €.'": ,-... ' -. ---,...-. , '"- " , "," "',""-"-.4.-
4

*
644 64.4
.54 .54
-J
. - ___ *
4 ~ - 5-4 ' - ~4
4 *54~1W .54 ~0
x.ai5414 z.a..5-414
4 *eo *oo
~Z -~3Z
,~d 5~
0

- - ~
,~oo~n -- o~
~r-r-

0
4-
r-.
Is

iljlrlv:
4-
ID, U

II-I 0

____
* S
14
____
____________
____
____________
____ - .5.4
* 0
I2.
o 4-.U
4.'

0
I
- C
o
0
4-

0
VI

_____ _____ * 0 010)

5 S bO 06

"4

4 5
4 9

0 S S
* 5 0

'4

23
 Z, ~ V- -o

um- C 0

unw upq~

VI~
r-4 o r

O4W

I ILI
rI

1'

cu 0
W4.

244
 -aI % If-

biI W W1 4w
ei
Ql

- ~ ~~ -- - - -P4 > o >

V-4 4

tU 0
>11
u

% %!

vi --
 (C)

(a)

(di)

(b) (d2)

* Figure 12. Typical field failures: (a-c) failure at wheel arm end, and (d) failure at anchor end.

26
 4r

(b)

(a)

,'.-

Y.4

Y.5

(C) (d)

Si2 ,
-Figure
o';v 5 :~
13. (a) Plastic flow of teeth; (b) initiation site; (c):*enlargement
;:.S ;. c;.; S '- -. of initiation site; and (d) SEM - 000K
'* .... **.*..*..,..*
corrosion.

a . *.. SSS
5,Longitudinal AlaIcie

CIO Transverse

Figure 14. Modes of failure of teeth.

P P"

28

S 2.
 -~~~n - - - - -- . 1-i -n-.

A101

* /0.

1* 4.0

...........
 u .
Figure 8 n.. s e.

-AO-.

Figure 18. Contact stresses: maximum principle stress contours Case 1.

,- ,.4

30

. .. -:-
"- _.p
.... ,
"4 ae2
Fiur 1. otat trs s maxmu tesc nor
,rnil
.
, q•
- C

"
.° ..""i 30

*,o C°.
 31 Ow

-t o0 -

-to M

Figure 20. Contact stresses: shear stress contours Case 1.

.- t..
-.. 4"m A,- . . ....

.
........
.
Figure 21. Contact stresses: sher stress ontours Case 2.

31:

."."."4.". ". ;, "..','.. '

" " ". ' ' f". ";"" "'
-"."7. " ' :",".".,' ". ".".."> ." "."."
 .......................................

4... .........

. ....
. ... . . . .. . . .2 0

-0.85E+5

... . ..
........
..

S.........

-0.45E+5

(C)

Figure 23. Torsional stresses Mode 1ll; (a) cracked spline, (b) subsurface crack, and (cWfinite element
model for subsurface crack.

32
 (b)

w .

Figure 24. Stress distribution on spline teeth (effect of tapering).

4,4

Figure 25. M1 test.

. 33
 r r"

.4 i~ g
6
---- : J C.= - . L-
U I- ___r:P CID EE, M ML7v-tDn lsC1 -wwa ~x1-mC
2 ~LZ ~IT.' 2~ L-Tzn i-2J2~~?Z~Z~i EI7~ C2J .92

';,c1 ~ ~JS~~E~~ ~iiUEI ~JFVlr2 ~lGJ~Li~C~ ~5

is' - %7---

SCALE.

- - -m-------- 2r.0-R$,

-- 34
 04 *° . I

w + 4 %

0 a

" 5 "

.- ,,'

04C4 ft.

.I In

305
.-- -.. .. .... . .. .. ... .. . A- ...... . .-. .- .,.t - ,- - ,
-. . |a - -- - . -, ~ ---

eson

eta Distribution

Sf (Y) - (r(P)NiQ))

OQ l, pQ0O
Probability
Density y - (0-0s) /0

assOS - Smallest Angle

eL - Largest Angle
*too

40

e... ~ ~ -'''I
~ ~ F l I I I I I

8 (degrees)

Figure 28. Spectrum load representation (run no. 48 - speed 25 mph).

R fGOA3-APGZ2 COURSE-18MPH-ROADAR IL ROTATION

A YNAX * 0.40993
D 0.7S YIIN a -0.65729
I
A
N 0.5s
S

-0.25 15.•00 20.00O

Angular Absolute Displacement

in degrees A 0
-0.50
5.79 35.98
7.66 39.65
-O. ? 7.16 41.78
9.20 44.21
-1 * 14.72 49.78
22.37 62.01
30.02

Figure 29.

36

-'-
'p

,4w p
,,., * y % . . %.rf, . :.. ,.: ,K., .'.
%t,' . ... .. ... ,. . -,,'
,,, . .. .,,.. *' ,, . . ," ,.,,,'
. .....
.. , .,,. .. ,
 I 0

vi

aN

so so 30 4@ £0 ~U so se"

* h

R
E
Q r
U
A
N "3 "

C
y
• Ii =... *. i .*.5, * ."0,

(b) A 9AQZ

4 Figure 30. (a) I.Q. plot (18 mph/i L) .and (b) uniform distribution.

37S
 CYCLE UP' PERIODA

LOt

(.l a *P/A

P, LOAD

-- PLASTIC
a ZONE
1
STAGE

AD/Do i d)

'Of

- TERMINAL
INSTABI LITY
No STAGE

CRACK STAGE
LENGTH I
'a' Nt
IC) P. LOAD 0 CYCLES

-- INITIATION LIVE

-. 38

5-%
 Yield Stress 220 9S1
Distilled Water
5 53
-0- 3. S O AK 3

Y 9-220Ks'L

Lab Air
da -8 1.71
- 9.1-,O1 AX
aa

(in/cycle)
16- Dry Air (Beres,.)
" 10" /f Frequ-ecy. TS Indep.
6 0 2"
-1 .6" 1 ' Figure 32. Determination of cycles

to failure (LEFM appr.).

• ~~10
"6
! ;
10 20 30 60
Ax (151 In)

;af

J ai
",,
.3

t~~~ F
MIWl IIIIIZS

Figure 33.

but

4* 39
.4o

q ' " ,r "

PX
"0 ' ""'d """ ""
 Cycles to Failure Relation
11091 N -B + .066 0

N Current Design,

8 8.06 (Adjusted from Base Line Data

to agree with Scranton's Torsional

40

8.40 (Represents 10%

/Reduction in Bar Stress)

30 7.70)
Base Line Data (B
(4340-230 KSI Yield 8 8 74 (200 Re-
* Based on Smooth specimen Data Mean
N
'%
~Stress
~~Romu - 0 (it
Temp., Lab Air
duction in Bar
Stress
Stes

20 Shot Peened Surface)

los
0o 106 10 10

Cycles EN) to Failure

Figure 34. Torsional fatigue life.

Torsionl ue Life

Cycles to Failure Relation

tglOf- 3+ .068as
so -8.06
""a

" 40

- \
'p Sumulation of a from S/H Curve !Uureuatatiom
20

.25

. 104 10 10 10
(a) Cycles (N) to Failure
.15
Probability
Desity
.10

.. 05
. SIN curve L1tioN a + 06e

.. 00
4
7. 1 7.6 7.8 8.0 8 8.2 6.4 8.6

Figure 35.
40

• - -.-._,,,. ,... '.'.'...'...'.'..'.'.-..'..,,'....'.."." . - .',. ,'.''.'v,.:'.. ° '" .. r,. .. ,- , ,N,, ,,.,.,..

;,
•~~~~~~~~~~ ,:,:e,..,:,:.;',,, ;,-'i .... ".. .. .l '. !)
 Monte Carlo Ps Error Measure

Ps,.
V

I I.
500 1000 500 2000 250
NUMBER OF SIMULATIONS
Figure 36. Additional criteria: convergence of 3rd and 4th moments.
'

1.0-

- K III Failure in Shaft

.8
25% (K) Reduction (Tapered)
In Spline Stress (crack in general tooth)
-.7

. 50% (KI Reduction (Tapered-Tolerance)

(K1 )In Spline Stress
0
.5
-4

.4- K Current Design

.3 40% Incredse in KIC Value

.2

3
101 102 10
Miles (Run #48 - 25 mph)

Figure 37. Torsion bar reliability - probability of survival versus miles - da/dN relationship.

41
 .4
04

'4

0lFigure 38. Miles to failure - no. 1

" torsion bar (miles at 25 mph).

.9 6.4

1.0

3I g7 of remainin
uProbability
741 miles hdve
ftertolif
completed.
P
l

.6
>been Current

,, 0esin \r

", ,.1[-
~201 Improvement ill
:"1 Desiqn (Stress
.4A Reuc tion )

..
• Figure 39. Probebility of sifvivft miles
-hIicrereutse'
re741 4,

beencompeted

'm 0

20 mrvmin r
 II I

V4 J
4

* ,0

vfi

t Io

wk1

P-4

* i

* C,

* . %, , -, " . % - ,, . , . ,. :Z ', . ,, . , , . , . " , , . , ' , . . , , _. . Y , . , • , - . - . 'r , 2 e , , ., , ' . . . , , . . F , , . , , ' , , , -

...i ...i .. k. h i l i ~ ,l l m~i i ~l
m ~ ii, lnl d Ii I -I - I

43
 00gooo spoo sso o-
.gz SI a
..
.910'

.410
,.

.90
.,,0
.o .

o ol o
a - 6

-'TI 0.10 .1a

SPC M STM.V4

.3 . IPcC,,, . $1W. ,.146

OAG.
PAIICL TEST. SEG. NO. I
co-IDINCE LEW 95.0

.M0-."

0 t
p p p I P I p p

* o .0U . .. US .... ..00 ....

M .

.%.90 *.. . O

.00 ..

OL
"

.0

- 0.

• S.

:E
,0
-PTH
"SPCIteN.
I M I
STEL 413.0
P"G. PARTICLE TEST. SIO. NO. ,
C€WICENCC L£VC.s 05.6

0 .020 .0'.0 .060 .060 .00 .ILO

*Figure 41. Crack detection probability of the magnetic particle inspection method for steel specimens plotted by
actual crack depth at 96% confidence.

44

..-.,...... ., .
. .. ,,, - . , . .*. . . !:'
,,
 Table 1. SCRANTON FATIGUE TEST RESULTS

Date Heat Code Total Cycles Location of Break

77 JAB 45,072 Test stopped - repair machine
77 JAF 245,157 Failure
77 JAK 284,560 Failure

78 JAE R 108,744 Small end radius

78 JOG R 137,917 Test stopped - repair machine
78 JBG R 246,867 Test stopped - repair machine
78 JBI 181,344 24" from large end
78 JBI 310,785 38" from large end
78 JBI 195,863 18" from large end
78 JB 362,016 18" from large end
78 JBK 270,778 19" from large end
78 JAO 113,341 Failure
78 JAT 248,105 Test stopped - repair machine
78 JBA 105,821 Test stopped - repair machine
78 JB8 273,434 Test stopped - repair machine
78 JBD 357,023 Test stopped - repair machine
78 JBC 207,889 Failure
78 JBG 186,849 Failure

79 JBL 239,396 Small end radius

79 JBI 147,773 Small end radius
79 JBL (2) 271,990 18" from large end
79 JBR 394,758 Machine failure
79 JBS 221,500 Small end radius
79 JBP 137,765 Small end radius
79 JBT 248,754 12" from small end
79 JBL (2) 201,169 Small end radius
79 JBL (2) 331,606 46" from large end
79 J8z 144,300 Small end radius
79 JBV 125,955 24" from large end
79 JCB 196,560 Small end radius
79 JBL (2) 123,325 Small end radius
79 JBL (2) 219,715 Small end radius
79 3CC 99,439 Small end radius
79 JCD 230,325 6" from large end
79 JL (2) 190,373 16" from large end
79 JBL (2) 156,315 Small end radius
79 JBL (2) 56,361 Small end radius
79 JBL (2) 142,939 26" from large end
79 JBL (2) 131,890 Small end radius
79 JCF 119,840 Small end radius
79 JCG 281,225 21" from large end
79 JCI 149,815 Small end radius
79 JBX 88,624 Small end radius
79 JBX 67,784 Large end radius
79 JCD 114,986 40* from large end
79 JCD 107,273 42" from large end
79 JCJ 190,044 8" from large end
79 JCK 190,050 Machine failure
79 JCT 192,170 Small end radius
79 JCS 133,677 Test stopped - repair machine
79 JCU 254,788 Small end radius
79 JCP 85,815 Test stopped - repair machine
79 JCO 45,437 Test stopped per L. R, suggestion
79 3CW 48,125 Test stopped per L. R, suggestion

80 JCX 108,158 Small end radius

80 JCO 260,376 Large end radius
80 3CZ 325,230 Machine failure
80 JCW 265,255 Machine failure
80 iCY 128,628 14" from large end
80 JCX 246,789 14" from small end
80 JCW 264,318 14" from large end
80 JHF 84,728 Test stopped - repair machine
80 JHG 77,608 Test stopped - repair machine
80 JHE 151,301 12" from small end
80 JBX 183,740 50" from large end
Ii

45

.. *,*.**i bb S~,~4".
-%
 Table 1. Continued

Date Heat Code Total Cycles Location of Break

80 JH. 140.401 Test stopped per L. R. suggestion
80 JHK 52,492 Test stopped per L. R. suggestion
80 J8X 107,417 62" from large end
80 JHP 222.374 Test stopped per L. R. suggestion
80 JCX 92,028 Test stopped per L. R. suggestion
80 JHP 46,249 Test stopped per L. R. suggestion
80 JHP 135,232 6" from large end
80 JNP 80,517 Large end radius
80 JP 214.090 48" from large end
80 JKP 167,191 Test stopped per L. R. suggestion
80 JKR 45,092 Test stopped per L. R. suggestion
80 JKS 298,271 60" from large end
80 JKD 123,265 Small end radius
80 JGX 75,850 34" from large end
80 JBY 56,600 32" from large end L
81 JLI 136.800 Test stopped per L. R. suggestion
81 JLG 92,700 Test stopped per L. R. suggestion
81 JLE 87,250 Test stopped per L. R. suggestion
81 JLF 231,750 Small end radius
81 JLH 78,440 Test stopped per L. R. suggestion
81 JLW 52.840 Test stopped per L. R. suggestion
81 JLX 125,140 Test stopped per L. R. suggestion
81 JLY 88,500 Test stopped per L. R. suggestion
81 JMD 180,320 Test stopped per L. R. suggestion
81 3MM 115,225 54" from large end
81 JND 269.811 Test stopped per L. R. suggestion
81 JNA 217,368 Machine failure
81 Jip 267,951 Test stopped per L. R. suggestion
81 JNC 88,192 Test stopped per L. R. suggestion
81 JNT 85,715 Test stopped per L. R. suggestion
81 JLT 95,703 Test stopped per L. Rt.suggestion
81 JLI 80,570 8" from large end
81 JLI 175,235 Test stopped per L. R. suggestion
81 JLI 146,881 Test stopped per L. R. suggestion
81 JLI 127,877 30" from large end
81 JLI 244,732 18" from large end
81 JLI 193,882 Machine failure
81 JLI 227,500 42" from large end
81 JLI 288,725 Ttst stopped per L. R. suggestion
81 JON 223,629 Small end radius
81 JOP 443,417 30" from small end
81 JKT 100,607 60" from large end
81 JLT 49,422 60" from large end
81 JKT 75,488 61" from large end
81 JKT 71,488 Small end radius
81 JLT 41,802 70" from large end
81 JLR 46" from large end
82 JOR 212,084 Large end radius
82 JPB 350,769 46" from large end
82 JOP 301,591 24" from large end
82 JP 303,336 Small end radius
82 JPC 584,366 18" from small end
82 JOS 268,040 Machine failure
82 JOS 228,710 20" from large end
82 JPC 358,269 Machine failure
82 JNL 77,990 48" from large end
82 JNL 57,409 44" from large end
Date Bar No. New Cycles To Failure Location of Break

82 1 185,000 Machine broke

82 2 148,000 Machine broke
82 3 544,000 Test stopped
82 4 447,696 Failed
82 5 240,348 Failed (was broken in MI machine)
82 6 362,771 Machine broke
82 7 367,307 Test stopped
All New Spline Design

A failed torsion bar from Ft. Knox spline failure?

M60-A3. Will be sent to Hickey.

46

Le
 Table 2. CHEMICAL COMPOSITION

Weight Percent
Specification AISI 8660
Source/Marking GD/JN MPU #5399 Minimum Maximum
Carbon 0.63 0.54 0.55 0.65
Manganese 0.87 - 0.75 1.00
Phosphorus 0.009 - - 0.04
Sulfur 0.02 0.02 - 0.04
Silicon 0.21 - 0.20 0.35
Nickel 0.58 - 0.04 0.70
Chromium 1.02 - 0.04 0.60
Molybdenum 0.15 - 0.15 0.25
f,

Table 3. ROCKWELL C HARDNESS

Source/Marking GDIJN MPU #5399
Location Spline Traverse Body Traverse Spline
HRC (48.0 - 50.3) (48.5 - 49.9) 50.0

Table 4. TENSILE PROPERTIES OF GD/JN TORSION BAR

0.2% Y.S. U.T.S. Elon. R.A.
Orientation (ksi) (ksi) (%) (%)

Long. 231.4 257.2 8.3 28.4

229.6 258.1 8.7 28.5

47

-Z . ..
 Table 5. ENERGY AND FIBROSITY DATA FOR GD/JN

Longitudinal Orientation (LR)

Testing Temp. Charpy Impact Energy Fibrosity

(o (ft-lb) (%)

-40 (-40 0 F) 5.5 5

+22 (+72 0 F) 7.7 10
+100 (+2120 F) 7.9 15
+160 (+3200 F) 8.8 50
+180 (+3560 F) 11.0 80
+220 (+4280 F) 9.0 100
+240 (+4640 F) 8.5 100

Transverse Orientation (TR)

Testing Temp. Charpy Impact Energy Fibrosity

(0C) (ft-lb) (%)

+22 (+72OF) 3.5 5

+22 (+72 0 F) 1.8 5
-25
-40 (-40 0 F) 2.8 5
-40 (-400F) 3.0 5

Table 6. FRACTURE TOUGHNESS

KQ, ksi VTh.

Source/Marking Orientation Testing Temp. (oc)

RT -40
GD/JN LR 45.7 32.2
40.3 24.7

TR 38.9 32.1
37.7 31.8

MPU #5399 TR 34.5

33.9

48

,° p
 Table 7. SPECTRUM LOAD (PROFILE IV COURSE) -BETA FUNCTION REPRESENTATION

Cumulative Test Results e tato

Probability OA (Degree) R (ree)
Run 40 (5 mph)
0.10 0.14 0.86
0.25 4.4 5.8
0.34 5.0 7.2
0.50 8.6 9.3
*0.66 12.5 11.5
*0.75 14.2 12.7
*0.99 17.0 16.7
Run 42 (10 mph)
0.10 14.0 6.1
*0.25 16.0 19.2
0.34 22.8 21.5
0.50 25.8 24.8
0.66 29.7 27.6
0.75 30.6 29.0
0.99 32.6 32.5
Run 48 (25 mph)
0.10 2.3 10.8
0.25 22.7 26.2
0.34 27.2 29.1
0.50 33.7 33.3
*0.66 39.5 37.1
0.75 41.6 39.3
0.99 46.0 46.9
Cumulative time probabilities of torsional bar angular displacement 0 adjusted to
positive range by6'0 9 + 10-1 where 0- = maximum negative angular displacement.

Table 8. MINIMUM LIFE ESTIMATES (99% SURVIVABILITY)

da/dN CURVE RESULTS
Mileage Expected Mileage Expected Mileage Expected
(Function of (Function of (Function of
Velocity Spline Stress) Spline Stress) Spline Stress)
(mph) Current Design 25% Reduction 50% Reduction
5 71.0 138.0 341.0
10 29.9 51.9 143.0
15 15.2 29.3 72.3
25 14.0 26.9 66.3
2514.2 28.4 70.2

* 49
 . ..m . . . . -.

Table 9. MONTE CARLO RESULTS FOR S/N CURVE MINIMUM LIFE ESTIMATE
(99% PROBABILITY OF SURVIVAL) VERSUS VELOCITY (PI4PH)
Mileage Expected Mileage Expected
Velocity Mileage Expected (10% Design (20% Design
(mph) (Current Design) Improvement) Improvement)
5 6974 9474 12970
10 2000 3138 4420
15 345 b33 1089
20 276 515 860
25 292 557 365
Note: A 99% survivability estimate of 262 miles was obtained from cumulative APG
mileage on vehicles at time of torsion bar failure. Velocity of vehicle during
tests was approximately 15 to 25 mph.

Table 10. TORSION BAR SPRING FAILURES FROM SELECTED APG TESTS, SEPTEMBER 1979-April 1982*

Vehicle TECOM APG Report Number Vehicle Position, % Cumulative APG Mileage
Type Project Number Number Failed Number and Side on Vehicle at Time of Failure

M6OA1E3 1-VC-080-060-083;PV-1 APG-MT-5376 9 1R,1L,6L,2R,4R, 1643, 1693, 1860, 2387, 2387,

5R,IR,1L,IR 2387, 2463, 2471, 2686
;PV-2 3 1L,1L,6R 2286, 3086, 3640
;PV-3 2 4L,4R 3025, 3025 -
;PV-4 3 1L,IR,4L 1127, 2218, 3093
;PV-5 5 IR,6L,5L,IL,6R 1190, 1988, 3306, 4085, 4171
;PV-6 6 1R,SL,5L,1R,1L,IL 546, 2217, 2711, 3303, 3834, 3998
;PV-7 4 1R,IR,1L,6R 3411, 3641, 3641, 3641
M6OAI0 1-VC-080-060-087 APG-MT-5299 2 1R,6R 396, 1596
M6OA3 1-VC-08F-060-033 APG-MT-5299 2 1L,6L 1402 average
M6OA3** 1-VC-087-060-035 APG-MT- 2 1L,1L 421, 1418
M48A5**1 3-WE-100-DIV-008 APG-MT- 3 6L,3R,1R 1516, 2638, 2659
*Taken from Appendix I1 .12'
tPositions 1, 2, and 6 ha e shocks attached to the roadwheels.
*Detailed failure report.5
*Discussed in greater detail in this report.

Table 11. PROBABILITY OF SURVIVAL VERSUS MILES - TACOM SIMULATION COURSE

5 mph
(12
Probability 5 mph (38 Displ) displ) 18 mph 25 mph
of Survival
(%) 11 2L 61 IL 1L 2L 6L 1L 2L 6L

Original 99 1467 * - 2702 14 16 1324 620 1060 2289

Design 95 4859 I 5032 37 45 3107 1478 2594 5441
90 12550 7433 70 89 5032 2471 4157 8777
80 13900 11210 176 238 9502 4701 7827 17160
70 398 590 18350 8354 13930 29280

10% Design 99 1618 * - 3742 32 37 1941 1010 1660 3363

Improvement 95 3527 I 7694 85 101 4617 2424 3951 7981
90 5487 11390 159 191 7290 3935 6245 12580 %
80 9693 18760 359 467 13680 7246 11570 24010
70 27860 792 1082 21980 12490 20130 29280

20% Design 99 2222 - 5107 65 74 2757 1523 2336 4686

Improvement 95 4732 | | 10350 166 193 6321 3618 5623 10590
90 7359 I 15420 301 361 9990 5717 8831 16910
80 12910 25160 652 818 18420 10320 16020 29280

50
 " APPENDIX A. TORSIONAL STRESSES IN A NONUNIFORM BAR

Due to the enlargement of the torsion bars at their ends there could be high
stress concentrations in the filets caused by the applied torsional moments.

Thermal analogy is used in the analysis of torsional stresses of the nonuniform

bars. According to the Michell torsion problem, 1 the angle of twist satisfies the
equation

a (Gr 3-) + 2- (Cr3 a)-O(A-l)

and the shear stresses are given by

ii rO 2-Gr -Cr-
r (A-2)
r6 r
DG zO --

where G is the shear modulus, r is the radius from the central axis and Z is the
distance along the axis. Using thermal analogy of axisymmetric heat condition and
setting,

y - r, z -z, T - ', K -K - Gr2 (A-3)

=
qB rTs

where T is the temperature, Ky and Kz the conductivity coefficients in the Y and Z

directions, and q the heat flux, the problem can be easily solved. By substituting
the above quantities in Equation A-i, results in the familiar heat conduction
-" equation of the axisymmetric body

a aT a aT
(yK : --) +- z )-- -Yq (A-4)

. It is to be noted that we could have used the substitution Ky - Kz - Gr3 , resulting

* in the plane heat conduction problem. However, the error in the analysis would be
much greater due to the combining of the term Gr3 rather than Gr2 in every element.
., The analysis would then be carried out using the standard heat conduction finite
element codes.

-.-

.4

1. TIMOSHENKO, S. and GOODIER, J.N. Theory of Elaticfty. McGraw4g, New York, 1951.

"
_%1 * 5l~9%.

4 k
 APPENDIX B. ST. VENANT TORSIONAL STRESSES OF MULTIPLY CONNECTED REGIONS

The existence of subsurface cracks in the torsion bar due to inclusions and
poor material is a possibility. The evaluation of stress intensity factors for
these cracks requires the solution of a St. Venant torsion problem of a multiply
connected region which is:

2 + -2
ax,2 ay2

The stresses are determined from

. Xz txy
a y, yz -(B-2)
a

with G the shear modulus and 0 the angle of rotation. The boundary conditions are

as 0 (B-3)

on the outer boundary and

ar/Bn ds - 2G@A i (B-4)

- where A i are the areas of the interior holes or crack regions and rj is the boundary.
For the case of a shaft with one or two subsurface cracks or holes,- can be chosen
to be zero on the outer boundary and hence satisfy the boundary condition
Equation B-3. While for other interior boundaries ri, Equation B-4 can be substituted
equivalently with * = Ai. The constants Ai will remain undetermined and can be
found from the s6lution of Equation B-I together with the equilibrium equations for
the whole section.

The other alternative is to use the condition Wa/as -0 on all boundaries,

which also complicates the solution of the differential Equation B-1.

The approach used here is based on the penalty function method.1 In the
* equation, B-1 is cast in a variational problem where

.. ( )2+ -2G0 dv- mi. (B-5)

La[(x ay

subject to the condition af/39 - 0 on all free boundaries.

1. REDDY, 1. N. PnAwky. F-ite Element Medtods hi Medmui= ASME, AMD, v. 51, 1982.

53
U,,

,- " . " ' " " " %" % " '" - %" " _ " " ' '. " ' r ". . '', . :
 Using the penalty function method, Equation B-5 is written in the following form:

= f[(#)2+ (#y) 2- 2G6.] d+lp .i ) 2 ds (B-6)

where Yp is the penalty parameter and is taken as a very large number thus guaran-
teeing Do/as to be very small and close to zero.

The minimization of Equation B-6 gives the heat equations of a two-dimensional

problem with the same analogy as in Appendix A. The penalty members are linear heat
condition elements (beam like) with very large (or infinite) conductivity.

54

,''",'
.. .. ,-..,..r
-.. . .. ." , .- * . .-. - .. .. -. . .. ~.5*
.- -. . ....-...-. -%*,
 DISTRIBUTION LIST

No. of
Copies To
2 Coimmander, Defense Technical Information Center, Cameron Station, Building 5,
5010 Duke Street, Alexandria, VA 20301

1 Commander, U.S. Army Materiel Command (AMC), 5001 Eisenhower Avenue, Alexandria,
VA 22333

Commander, U.S. Army Tank-Automotive Command, Warren, HI 48090

1 ATTN: AMSTA-NS, Dr. H. Dobbs
I AMSTA-RC, Mr. H. Keifesh
I AMCPH-M60, COL J. A. Logan
1 AMCPM-M60, Mr. R. DeGroot
1 AMCPM-MI, Mr. A. Jokubaitis
1 AMCPM-GC-L, Mr. J. Chapin
1 AMCPM-FVS, BG C. B. Donovan
1 AMCPM-FVS-Q, Mr. R. A. Gasparski
1 AMCPM-M9, LTC R. F. Huttner
1 AMCPM-LA, COL B. L. McClain
1 AMSTA-H, Mr. R. L. Swint
I
1 AMSTA-QK, Mr. W. McGinley
" I AMSTA-T, Hr. G. Nevcomb
I AMSTA-Z, COL W. Chandler
I Tech. Management Dis., Mr. M. McCoulough

Commander, Aberdeen Proving Ground, MD 21005

I ATTN: STEAP-MT-G, Mr. J. A. Robinson, Material Testing Directorate
1 STEAP-MT-G, Mr. R. A. McKinnon
1 STEAP-MT-D, Mr. A. Cummings

FMC Corporation, Ordnance Engineering Division, 1105 Colman Avenue, San Jose,
CA 95108
1 ATTN: Mr. A. J. Austin
1 Mr. R. Hare
1 Mr. G. L. Boerman

FMC Corporation, Steel Products Division, 2101 West 10th Street, Box 1030,
Anniston, AL 36201
I ATTN: Mr. J. P. Lewis

General Dynamics, Fort Worth Division, P.O. Box 748, Fort Worth, TX 76101
1 ATTN: Mr. L. M. Smith

Jet Propulsion Laboratory, 4800 Oak Grove Drive, Pasadena, CA 91109

" 1 ATTN: Dr. D. H. Anderson
1 Dr. Levy Reinhart

Bolt-Beranek & Newman, Inc., 10 Moulton Street, Cambridge, MA 02138

1 ATTN: Mr. A. Galaitsis

General Dynamics, Land Systems Division, P.O. Box 1852, Warren, MI 48090
I ATTN: Mr. E. Hirych
p -%

. . ' ' ' .

. p 1 i. . . ..
 No. of
Copies To
Commander, U.S. Army Engineer Waterways Experiment Station, Vicksburgh,
HS 39180
1 ATTN: Mr. C. Nuttall
1 Mr. N. Murphy

Battelle Columbus Laboratories, 505 King Avenue, Columbus, OH 43201

1 ATTN: Library

Director, Army Materials and Mechanics Research Center, Watertown,

MA 02172-0001
2 ATTN: AMXWR-PL
7 Authors

4
S

.4

'

! p.

* U.S- GOVERNMqENT PRINTING OFFICE: 1 S5--S02-244--20112 P

.*): ,6
 U cS A*Q

jpj L. Cc z
I 0

~I
V .1 -, b.
do t..C S 0 aw
*-M g
@ 0 C
4, 0@
LO~ 0*C ~ ue 0:.b.
A 4; @ I 45 'm- L , K"445'

As0C~I L:' C5L=O@'Z

W c
0 i 0 C;
C i ~ -~ 0. C0. 110 5"

ix- u... c z ..Q i;gZ cC .

C a,
- ~ ~
~ LA ~L ~ ~C~.
~ 0. ~ V.~Ci5
0C ~ ~~5I20
40
C, . 0.%C *LA
0-
a).V-4

C-2 0 , ~ C@
0
U S
0. I .Wo *-I '. :1
'UCM'
2 I= '5 0.
zA iU~RVC
LO::
--
F.5.
SAC I0 w X, 3 CCLOwS CL. $_ 0

.gw m*, ft u4 L. t O' MCC

U.0- U'=1 . '
-v .0 & .? Cm 0.~

Cie -I- *.. ccL

OW "-0 ": 0A c- -- V LU F0 v5 10 z 5 u a

aGL C =! C 'c CL
0. C.0 30
w 0 0-
9 w2 '
.-)Z
a -QC C.s
0,C.~
- w~@' O
SAUC.
M
9-0 ~.- tA UC CC S Wt0 LU 0
-CYC
040~SMC6A
'4- L2S
C
0. 1.0-: I C -~-~
*
Cl0-i-
,C0...A .CC
0

C V.:C.C
C~~J C UC0.0@
0.~~- C .C S A' O C:10-

C, m

- CC 0 M& 50-C %-MV

-0 1 - :
3; l)ZE 2:C W
A 0- - @00 SA

'-5-.
~ 0..0.
C .L IC '. 0
.0.. UL6m 40
.
-0
I
*W Os'
C
u5
C-;
W- .2
C lC
.
O 0,2
ual C
VC
Z
s 2f UCs~f u A C CL.;
OsACL -2 'lci
00- 0 4, - 10 a '

±
Q
0.c.

~
C-
C ~CC'.C
~if 'l
. .

2
~
'V
--
.- .
4 I
I ~ A
- 2
ch
5C,
:
C.
5
c
~ . s ~C3-0
0L ~
mf*~S
L. __UC..
-
C )~ C

L V. -
c 5 2c C C "J 0 c1'
- C * . A .0*. C s. .
-~4 ~c.' .c0 I C%

w. -C .VC Au 0: 1. 0C *VC,-~C
O~C
.C0-
L. ~~~~~~c LC 01. ' .-. O 5
0~~~o .5 5' cop P-~ 4A
C W-0A
- SMC

jC
01 a4
o.
OU.SS55 ''
C , C s- LI.
m~ OC L
C C a.. AW
LV. Z0 C.-. SA
CL0C L
A

5..Lt c 0 C IVCA- C CLC

WS' -Z GOP
C '6 0.,CS1'V 4 A~C C

*.!e-. .CC-- ~K -,02 C w N.540 S C.C. if-VS V.?.t

5.- * Ol Os7.-t S
CCO' *C '.
0.0C.0.1
C.. * .; 0C
.c.
%A.
20
'5. C
C0 17 0
- MCA
W40
a-- 0 0 06L0 c .0
C,
C30.. ,~
-R
~
1.
2.2
~ ~ .
Ow
.C .-.OC
u
L
2
C Vv
£2-.
1-. .--
0:
5
.2. - -W
CC..
S 0,
L4
WO
C

C167A00 2: .- C'--C V

C..N . L CC
A V~qt C ICU.i' t. CC 200C,'

a,, M. w .0 0C: UC3 6,4

I _ .01 :1 1C. z
hin - I
,
JA
1 w,-LO , I" I - 0 , A 4L iS .- a.C0
CC2

.Lr L...C0..
gc1a v1-

---- - I- At 'A-------- 22-v

* '.. * 2. 01, 4010 -

Ci 1 I... I .

1* 34 9, * -