https://ntrs.nasa.gov/search.jsp?

R=19840017961 2019-09-21T05:02:39+00:00Z
N 8 4 - 2 6 0 29

NASA Technical Memorandum 83608

Factors That Affect the Fatigue Strength

of Power Transmission Shafting and
Their Impact on Design

Stuart H. Loewenthal
Lewis Research Center
Cleveland, Ohio

Prepared for the

Fourth International Power Transmission and Gearing Conference
sponsored by the American Society of Mechanical Engineers
Cambridge, Massachusetts, October 8-12, 1984

NASA
 FACTORS THAT AFFECT THE FATIGUE STRENGTH OF POWER TRANSMISSION SHAFTING

AND THEIR IMPACT ON DESIGN

Stuart H. Loewenthal
National Aeronautics and Space Administration
Lewis Research Center
Cleveland, Ohio

ABSTRACT
A long-standing objective In the design of power transmission shafting Is
to eliminate excess shaft material without compromising operational relia-
bility. A shaft design method Is presented which accounts for variable ampll-
tude loading histories and their Influence on limited life designs. The
effects of combined bending and torslonal loading are considered along with a
number of application factors known to Influence the fatigue strength of shaft-
Ing materials. Among the factors examined are surface condition, size, stress
concentration, residual stress and corrosion fatigue.

INTRODUCTION
The reliable design of power transmitting shafts 1s predicated on several
major elements. First, the fatigue (stress-life) characteristics of the given
shaft 1n Its expected service environment must be established. This can be
accomplished from full-scale component fatigue test data or approximated, using
test specimen data. Some of the Influencing factors to be considered are the
surface condition of the shaft, the presence of residual stress or points of
stress concentration and certain environmental factors such as temperature or a
corrosive atmosphere. Secondly, the expected load-time history of the shaft
must be obtained or assumed from field service data and then properly simulated
analytically. The effects of variable amplitude loading, mean stress and load
sequence are potential Important factors to Include 1n a description of the
loading history. Finally, a reliable mathematical model 1s needed which ra-
tionally considers both the fatigue characteristics of the shaft and Its load-
Ing history to arrive at the proper shaft diameter for the required service
life and reliability. One last step 1s to check shaft rigidity and critical
speed requirements, since these and other nonstrength factors can occasionally
dictate an Increase 1n shaft diameter. This 1s often the case for lightweight,
high speed machinery.
While the above considerations have often been addressed In fatigue anal-
ysis of structural members [1 to 5], their application to the design of powers
transmission shafting has only been partially accomplished. Traditional shaft
design methods [6 and 7] do consider the effects of combined stress loading,
usually through the distortion energy theory of failure, but rarely take Into
account the effects of variable amplitude loading, mean stresses or limited
life design. More recent approaches [8 and 9] adapt traditional methods to
computer-aided design procedures but still neglect some of these other Import-
ant factors.
The principal objective of this present Investigation 1s to develop a more
complete approach to shaft design from a strength standpoint. The proposed
method will emphasize the Influence of these aforementioned operating variables
on shaft diameter and life. Other application factors such as surface con-
dition, stress concentration and size will also be addressed.

NOMENCLATURE
b slope of the S-N curve on log-log coordinates or fatigue strength
exponent (taken as positive value)
d shaft diameter m (1n.)
dp relative diameter, defined 1n Eq. (23)
FS factor of safety
K^ theoretical stress concentration factor
k-factor product of fatigue life modifying factors, defined 1n Eq. (7)
ka surface factor
kj, size factor
kc reliability factor
kg- temperature factor
ke fatigue stress concentration factor
kf press-fitted collar factor
kg residual stress factor
kn corrosion factor
k^ miscellaneous effects factor
LR relative life, defined 1n Eq. (12)
M bending moment, N-H (1n.-lb)
NL total shaft life 1n cycles
Nf number of cycles to failure at of
NI number of cycles to failure under load 1
n shaft speed, rpm
n^ number of loading cycles under load 1
q notch sensitivity
T torque, N-m (1n.-lb)
o bending stress, N/m2 (lb/1n2)
oef effective nominal stress, N/m2 (lb/1n.2)
of corrected bending fatigue limit of shaft, N/m2 (lb/1n.2)
of bending or tensile fatigue limit of 2polished,2 unnotched test
specimen without mean stress, N/m (lb/1n. )
ofm bending or tensile fatigue limit 2of polished, unnotched test
specimen with mean stress, N/m (lb/1n.2)
of true cyclic fracture strength or fatigue strength coefficient,
N/m2
ou ultimate tensile strength, N/m2 (lb/1n.2)
oy yield strength, N/m2 (lb/1n.2)
T shear stress, N/m2 (lb/1n.2)
TU ultimate shear strength, N/m2 (lb/1n.2)
Ty yield shear strength, N/m2 (lb/1n.2)

Subscripts:
a alternat1ng=(max-m1n)/2
m mean=(max+m1n)/2

max maximum
m1n minimum
r fully-reversing

FATIGUE FAILURE
Ductile machine elements subjected to repeated fluctuating stresses above
their endurance strength but below their yield strength will eventually fall
from fatigue. The Insidious nature of fatigue Is that 1t occurs without visual
warning at bulk operating stresses below plastic deformation. Shafts sized to
avoid fatigue will usually be strong enough to avoid elastic failure, unless
severe transient or shock overloads occur.
Failure from fatigue 1s statistical 1n nature Inasmuch as the fatigue life
of a particular specimen cannot be precisely predicted but rather the likeli-
hood of failure based on a large population of specimens. For a group of
specimens or parts made to the same specification the key fatigue variables
would be the effective operating stress, the number of stress cycles and volume
of material under stress. Since the effective stresses are usually the highest
at points along the surface where discontinuities occur, such as keyways,
splines, and fillets, these are the points from which fatigue cracks are most
likely to emanate. However, each volume of material under stress carries with
1t a finite probability of failure. The product of these elemental probabili-
ties (the "weakest link" criterion) yields the likelihood of failure for the
entire part for a given number of loading cycles.
At present there 1s no unified, statistical failure theory to predict
shafting fatigue. However, reasonable accurate life estimates can be derived
from general design equations coupled with bench-type fatigue data and material
static properties. Fatigue test data are often obtained on either a rotatlng-
beam tester under the conditions of reversed bending or on an axial fatigue
tester. The data generated from these machines are usually plotted 1n the form
of stress-life (S-N) diagrams. On these diagrams the bending stress at which
the6 specimens did not fall after some high number of stress cycles, usually
10 to 107 cycles for steel, 1s commonly referred to as the fatigue limit,
of. For mild steels 1t 1s the stress at which the S-N curve becomes nearly
horizontal. This seems to Imply that operating stresses below the fatigue
limit will lead to "Infinite" service life. However, this Is misleading since
no part can have a 100 percent probability of survival. In fact, fatigue limit
values determined from S-N diagrams normally represent the mean value of the
failure distribution due to test data scatter. Statistical corrections must be
applied for designs requiring high reliabilities as will be discussed.
Furthermore, many high strength steels, nonferrous materials and even mild
steel 1n a corrosive environment do not exhibit a distinct fatigue limit [1].
In view of this, 1t 1s best to consider that the fatigue limit represents a
point of very long life (>10& cycles).

APPROACH
Traditional shaft analysis generally considers that the nominal loads act-
Ing on the shaft are essentially of constant amplitude and that the shaft life
1s to exceed 10& or 10? cycles [6]. Sometimes shock or overload factors
are applied. However, most shafts 1n service are generally exposed to a spec-
trum of service loads. Occasionally, shafts are designed for lives that are
less than 106 cycles for purposes of economy. Both of these requirements
complicate the method of analysis and Increase the uncertainty of the predic-
tion. Under these conditions, prototype component fatigue testing under sim-
ulated loading becomes even more Important.
Short life design. - Local yielding of notches, fillets, and other points
of stress concentration are to be expected for shafts designed for short
service lives, less than about 1000 cycles. Since fatigue cracks Inevitably
originate at these discontinuities, the plastic fatigue behavior of the mate-
rial dictates Us service life. Most materials have been observed to either
cycllcly harden or soften, depending upon Its Initial state, when subjected to
cyclic plastic strain. Therefore, the cyclic fatigue properties of the mate-
rial, which can be significantly different than Its static or monotonlc
strength properties, need to be considered 1n the analysis. For short, low
cycle life designs, the plastic notch strain analysis, discussed 1n detail 1n
[3, 5, and 10] 1s considered to be the most accurate design approach. This
method, used widely 1n the automotive Industry, predicts the time to crack
formation based on an experimentally determined relationship between local
plastic and elastic strain and the number of reversals to failure.
Intermediate and long life designs. - For Intermediate and long life
designs both total strain-life and nominal stress-life (S-N curve) methods
have been successfully applied [3 and 10]. Although both methods provide
reasonable fatigue life predictions, only the nominal stress-life method will
be outlined here.
Obviously, the key to accurate fatigue life prediction 1s obtaining a good
definition of stress-life, S-N, characteristics of the shaft material. Mean
bending and/or torslonal stress effects should be taken Into account 1f pre-
sent. Furthermore, a good definition of the loading history Is also required.
Even when these requirements are met, the accuracy of the prediction 1s approx-
imate with today's state-of-knowledge. As an example, an extensive cumulative
fatigue damage test program was conducted by the SAE to assess the validity of
various fatigue life prediction methods [10]. Numerous simple geometry,
notched steel plate specimens were fatigue tested 1n unaxlal tension. Tests
were conducted under constant amplitude loading and also under a variable
amplitude loading that closely simulated the service loading history. The test
specimens' material fatigue properties and the actual force-time history were
very well defined. Under these well controlled conditions, predicted mean life
from the best available method was within a factor of 3 (1/3 to 3 times) of the
true experimental value for about 80 percent of the test specimens while some
of the other methods were considerably less accurate [10]. Under less Ideal
conditions, such as when the loading history and material properties are not as
well known or when a multlaxlal stress state 1s Imposed, a predictive accuracy
within a factor of 10 of the true fatigue life would not be unacceptable with
today's state-of-knowledge.

S-N CURVE
In order to determine the proper shaft size for a given number of stress
cycles under a variable amplitude loading situation It 1s necessary to con-
struct an S-N curve for the shaft under the proper mean loading condition. If
an experimentally determined S-N curve for the shaft 1s available then, of
course. It Is to be used. However, 1f actual test data Is not available, 1t Is
still possible to generate a reasonable estimate of the S-N characteristics of
the shaft as shown In Fig. 1. In F1g. 1, a straight line connects the fatigue
strength coefficient of at 1 cycle with the shaft's corrected fatigue limit
of at 106 stress cycles((or 107 cycles If applicable) on log-log coordinates
[3]. The coefficient of 1s the true stress (considering necking) required
to cause fracture on the first applied bending stress reversal. It Is normally
greater than the nominal tensile strength of the material ou.
This method assumes that the fracture strength of the shaft 1s not appre-
ciably affected by the presence of any mean bending or torslonal stresses or
the presence of a notch. The reason for this 1s that 1n a bending or torslonal
strength test, the outer fibers fracture first. Any Initial mean or residual
stress or notch effect will be lost to local yielding as the load 1s applied.
This Is not the case for an axial strength test, since the whole cross section
of the specimen rather than the outer fibers must carry the mean load [3].
Values for of are not commonly available 1n the open literature.
Table 1 [11 and 12] lists representative values of of and of along with
other strength properties for several steel compositions. For steels not
listed 1n Table 1 with hardnesses less than approximately 500 BHN, reference
(5) recommends the following rough approximation:

o « o + 345 HPa (1)

or

of * o * 50 000 ps1
where ou = ultimate tensile strength.
The parameter b appearing 1n Table 1 1s commonly referred to as the
fatigue strength exponent [11]. It 1s the slope of the S-N line .on log-log
coordinates, taken as a positive value here, where b = log (of/of)/6 for a
fatigue limit based on Nf = 106 cycles or

b = log (<^/of)/7 for Nf = 107 cycles (2)

Thus, 1f of and of are known or approximated, slope b can be found
and an S-N curve can be constructed from the relation:

of (3)
f
"1 \"1/
3
where 10 < N.1 < Nff and where o,a 1s the alternating failure stress corre-
1
spending to N^ cycles to failure and of 1s the fatigue limit strength
corresponding to Nf cycles to failure.
As shown 1n F1g. 2, Eq. (3) together with the simple approximation for
of given 1n Eq. (1) provides a reasonably good correlation with reversed
bending fatigue data of different strength steels appearing 1n [13]. The well
known approximation that the fatigue limit of 1s about half of the tensile
strength ou seems to hold reasonably well for all the steel test data
appearing 1n F1g. 2, except for that 1n the 0 to 483 HPa tensile strength
range. The reason for this descrepancy 1s not clear. It does, however Illus-
trate the Importance of obtaining actual fatigue life properties rather than
relying on simple approximations. Furthermore the high degree of scatter of
the test data 1n F1g. 2 1s not uncommon 1n fatigue testing. The S-N curve
represents the mean or average strength characteristics of a population of
components. Working stress levels must be reduced to assure higher reliabili-
ties than this 50 percent survival rate as will be discussed next.

FATIGUE LIFE MODIFYING FACTORS

It should be stressed that the fatigue limit of value to be used 1n
constructing the S-N curve 1n F1g. 1 1s that for the shaft to be designed and
not that of the test specimen material. The fatigue of the shaft 1s almost
always different from fatigue limit of the highly polished, notch-free fatigue
test specimen, listed 1n material property tables such as 1n Table 1. A number
of service factors that are known to affect fatigue strength have been Identi-
fied. These factors can be used to modify the uncorrected fatigue limit of the
test specimen, of*, as follows:

°f 'kakbkckdkekfkgkhk1°f
where
af corrected bending fatigue limit of shaft
of bending or tensile fatigue limit of polished, unnotched test specimen
without mean stress
ka surface factor
((5 size factor
kc reliability factor
k<j temperature factor

ke fatigue stress concentration factor

kf press-fitted collar factor
kg residual stress factor
kn corrosion factor
kj miscellaneous effects factor
Design data for factors ka through ke are relatively available 1n
the open literature [1, 2, and 6] and thus they will be only briefly discussed
here. However, factors kf through k^, although lesser known and docu-
mented, are often quite Important to shafting fatigue and therefore will
receive greater attention. A more thorough examination of all these factors
can be found 1n [14 and 15].
Surface factor. ka. - Since the surface of the shaft 1s the most likely
place for fatigue cracks to start. Its surface finish and any Irregularities
such as oxide and scale defects or surface decarburlzatlons can have a major
Impact on fatigue life. Typical values of ka range from about 0.9 for
turned, ground and polished shafts of low tensile strength (400 HPa) to as low
as 0.1 for high strength, forged shafts with significant surface defects [2 and
14].
Size factor, kjj. - Large shafts tend to have lower fatigue strength then
small shafts. This 1s primarily due to the great volume of material under
stress and the attendant greater Hkelyhood of encountering a potential fatigue
Initiating defect 1n the material's mlcrostructure. Also the metallurgical
structure of large parts tends to be coarser and less uniform than small parts.
Since the diameter of fatigue specimens tend to be small, typically 8 mm 1n
diameter, a strength reduction factor should be applied for larger shafts.
Values of k^ typical range from about 0.9 for 500 mm diameter shafts to
approximately 0.65 for shafts 250 mm 1n diameter [14 and 15].
Reliability factor. kc. - As previously discussed, published fatigue
limit data usually represent an average value of the endurance strength of the
sample of test specimens. In the absence of specific test data, the failure
distribution of steel specimens 1s often assumed to follow a Normal or Gaussian
distribution with a standard deviation of about 8 percent of the mean. Thus
for a 90 percent nominal reliability, kc 1s approximately 0.9 and for a 99
percent reliability kc Is approximately 0.8 [14 and15],
 Temperature factor, kj. - The fatigue limit of carbon and alloy steel
1s relatively unaffected by operating temperatures between approximately -70°
to 300° C. At lower temperatures the bending fatigue strength of steel
Increases while at temperatures above about 400° C, some steels begin to loose
strength [14 and 15].
Fatigue stress concentration factor, ke. - Experience has shown that
shafts almost always fall at a notch, hole, keyway, shoulder or other dis-
continuity where the effective stresses have been amplified. Fatigue data
Indicate that low strength steels, due to their ductility, are far less sen-
sitive to the effects of a stress raiser than high strength steels. This 1s
reflected by the notch sensitivity parameter q which 1s used to modify the
theoretical (static) stress concentration factor K^ as follows:

e Uq(Kt - 1)
Reference [16] 1s an excellent source of design values for both K^.
and q.
Press-fitted collar factor, kf. - A common method of attaching gears,
bearings, couplings, pulleys, and wheels to shafts and axles 1s through the use
of an Interference fit. The change 1n section creates a point of stress con-
centration at the face of the collar. This stress concentration coupled with
the fretting action of the collar as the shaft flexes 1s responsible for many
shaft failures 1n service. A limited amount of fatigue test data have been
generated for steel shafts having press-fitted, plain (without grooves or
tapers) collars 1n pure bending. Based on this data from several sources,
typical fatigue life reductions range from about 50 to 70 percent [16 and 17].
Therefore, approximate range of press-fitted collar factors:
kf * 0.3 to 0.5
Larger shafts having diameters greater than about 75 mm (3 1n.) tend to
have kf values less than 0.4 when the collars are loaded. Smaller shafts
with unloaded collars tend to have kf values greater than 0.4. The effect
of Interference pressure over a wide range between collar and shaft has been
found to be small, except for very light fits (less than about 28 MPa or 4000
ps1) which reduces the penalty to fatigue strength [16]. Surface treatments
producing favorable compresslve residual stresses and hardening processes such
as cold rolling, peenlng, Induction or flame hardening can often fully restore
fatigue strength (kf = 1) [1]. Stepping the shaft seat with a generaous
shoulder fillet radius or providing stress relieving grooves on the bore of
the collar can also provide substantial strength Improvements.
Residual stress factor. kg. - The Introduction of residual stress
through various mechanical or thermal processes can have significant harmful
or beneficial effects on fatigue strength. Residual stresses have the same
effect on fatigue strength as mean stresses of the same kind and magnitude.
Thus residual tensile stresses behave as static tensile loads that reduce
strength while residual compresslve stresses behave as static compresslve
stresses which are beneficial to fatigue strength. Table 2 lists many of the
most common manufacturing processes and the type of residual stress they are
likely to produce. The extent that the residual tensile stresses from these
processes reduce or benefit fatigue strength 1s dependent on several factors
Including the severity of the loading cycle and the yield strength of the
material 1n question. Since the maximum residual stress (either compresslve
or tensile) that can be produced 1n a part can be no greater than the yield
strength of the material minus the applied stress, harder, higher strength
8
materials can benefit more or be harmed more by residual stresses [3 and 18].
This coupled with an Increase 1n notch sensitivity makes 1t Important to stress
relieve welded parts made from stronger steels and Increases the need to cold
work critical areas. For low cycle fatigue applications 1t usually does not
pay to shot peen or cold roll mild steel parts with relatively low yield
strengths since much of the beneficial residual compresslve stress can be
"washed-out" with the first applications of a large stress.
Cold working of parts or the other means listed 1n Table 2 to Instill
residual compresslve stress 1s most often applied to minimize or eliminate the
damaging effect of a notch, fillet, or other defect producing high stress
concentration or residual tensile stresses. This Is clearly Illustrated 1n
F1g. 3 where shot peenlng the notched region of the test specimen has almost
entirely eliminated the notch effect. Cold working processes not only generate
favorable compresslve stresses but also work harden the surface of the part
leading to Increased fatigue strength. Typical design Information and data on
the effects of cold working and many of the other residual stress can be found
1n [1, 3, and 18].
Corrosion fatigue factor. kn. The formation of pits and crevices on
the surface of shafts due to corrosion, particularly under stress, can cause a
major loss 1n fatigue strength. Exposed shafts on outdoor and marine equip-
ment as well as those 1n contact with corrosive chemicals are particularly
vulnerable. Corrosion fatigue cracks can even be generated 1n stainless steel
parts where there may be no visible signs of rusting [1]. Furthermore, designs
strictly based on the fatigue limit may be Inadequate for lives much beyond
10^ or 107 cycles 1n a corrosive environment. Metals fatigue tested even
1n a mildly corrosive liquid like fresh water rarely show a distinct fatigue
limit [1]. For example, the S-N curve for mild carbon steel tested 1n a salt
water spray shows a very steep downward slope, even beyond TO8 cycles.
Corrosion fatigue strength has also been found to decrease with an Increase 1n
the rate of cycling so both the cycling rate and number of stress cycles should
be specified when quoting fatigue strengths of metals 1n a corrosive environ-
ment. Reference [1] contains a wealth of Information on the corrosive fatigue
strength of metals. Typically, the bending fatigue strength of chromium steels
at 10' cycles range from about 60 to 80 percent of the air tested fatigue
limit when tested 1n a salt water spray [1]. Surface treatments such a galvan-
izing, sherardlzlng, zinc or cadmium plating, surface rolling or n1tr1d1ng can
normally restore the fatigue strength of carbon steels tested 1n fresh water or
salt spray to approximately 60 to 90 percent of the normal fatigue limit 1n
air [1].
Miscellaneous effects factor, k^. - Since fatigue failures nearly
always occur at or near the surface of the shaft, where the stresses are the
greatest, surface condition strongly Influences fatigue life. A number of
factors that are often overlooked but are known to affect the fatigue strength
of a part are listed below:
(1) fretting corrosion
(2) thermal cycle fatigue
(3) electro-chemical environment
(4) radiation
(5) shock or vibration loading
(6) ultra-high speed cycling
(7) welding
(8) surface decarburlzatlon
 Although only limited quantitative data has been published for these
factors [1 and 2], they should, nonetheless, be considered and accounted for 1f
applicable.
S-N prediction. - Figure 3 Illustrates the effects that the above fatigue
life modifying factors (k-factor) have on the stress-life relation of Eq. (4).
A comparison was made with rotating beam fatigue data generated 1n [19] for
smooth, notched, and notched, shot-peened steel specimens having a tensile
strength ou of 897 MPa.
From the approximation given 1n Eq. (1), of at 1 cycle was estimated
to be 1241 MPa. The fatigue limit of the test specimens of at 106 cycles
was estimated to be 0.5 ou or 449 MPa. In the case of the smooth, polished
test specimen all of the k-factor = 1 , so the upper line appearing 1n F1g. 3
can be drawn.
In the case of the notched specimen having the geometry shown 1n F1g. 3,
Kt = 1.76 and q = 0.79 according to (16). From Eq. (5), the fatigue
stress concentration factor, ke = 0.63 and the fatigue limit of the notched
specimen = 0.63 (449) or 283 MPa as shown In F1g. 3.
It 1s Instructive to note from F1g. 3, that the compresslve residual
stress and work hardening provided by shot peenlng virtually eliminated the
detrimental notch effect almost entirely. Secondly, the slope of the S-N curve
1s steeper, that 1s b 1s larger, for the notched shaft. Since shaft life 1s
Inversely proportional to stress raised to the 1/b power, where 1/b »13.6 for
the smooth shaft versus 1/bw 9.3 for the notch shaft, the notched shaft's
life, although lower, 1s less sensitive to stress amplitude changes than that
of the smooth shaft. In fact, slope b Increases with a decrease 1n k-factor
or a decrease 1n tensile strength. This 1s shown In F1g. 4 where b Is plot-
ted from the following approximation derived from Eqs. (1), (2), and (4):

l / % * 345
b 1o
6- 9o.5o\k-factorfor * 1n HPa

where
k-factor = k k.k k.k k,k k . k . (7)
a b c d e f g h l
It should be pointed out that the presence of a mean stress, either
applied or residual, will cause a change 1n endurance strength and therefore
affect slope b. Mean torlsonal, bending or tensile stresses will decrease
Of and thus Increase b while compresslve stresses will have the opposite
effect. The effect of mean stresses will be discussed later.

VARIABLE AMPLITUDE LOADING

The following 1s a greatly simplified approach to estimate the required

shaft diameter for either a limited or unlimited number of stress cycles under
a variable amplitude loading history. It assumes that the loading history can
be broken Into blocks of constant amplitude loading and that the sum of the
resulting fatigue damage at each block loading equals one at the time of fail-
ure 1n accordance with Palmgren-M1ner linear damage rule. Great care must be
exercised 1n reducing a complex, Irregular loading history Into a series of
constant amplitude events 1n order to preserve the fidelity of the prediction.
Reference [4] discusses the merits of several cycle counting schemes that are
commonly used 1n practice for prediction purposes.

10
 A shortcoming of Miner's rule 1s that It assumes that damage occurs at a
linear rate without regard to the sequence of loading. There 1s ample experi-
mental evidence that a virgin material will have shorter fatigue life, that 1s
Miner's sums less than one, when first exposed to high cyclic stress before low
cyclic stress [1 and 4]. This "overstresslng" 1s thought to create submicro-
scoplc cracks In the material structure that can accelerate the damage rate.
On the other hand, test specimens exposed first to stresses just below the
fatigue limit are often stronger 1n fatigue than when new. This "coaxing" or
"understralnlng" effect which can produce Miner's sums much larger than one 1s
believed due to a beneficial strain aging phenomena. While Miner's sums at
the time failure can range from 0.25 to 4 depending on loading sequence and
magnitude, the experimental range shrinks to approximately 0.6 to 1.6 when the
loading 1s 1n a more random manner [19]. This 1s often acceptable for failure
estimates. More complicated cumulative damage theories have been devised to
account for "sequencing" effects. In fact reference [19] discusses seven
different ones, but none of them have been shown to be completely reliable for
all practical shaft loading histories. In most cases, Miner's rule serves
almost as well and because of Its simplicity 1t 1s still preferred by many.
Assuming that the shaft 1s exposed to a series of 3, 1n this case,
alternating bending moments of constant amplitude M acting for n,, '1 loading
1
cycles, M. for n_ cycles, and M for n_ cycles, then according to
Miner's rule:

N
(8)
l N2 3
where N, 1s the number of cycles to failure at bending moment M ,N Is
l
the cycles to failure at M , etc.
From the straight line on the log-log S-N plot of F1g. 1, 1t 1s clear that

^ M = N
(9)
'f \ 3/

where aa 1s the alternating bending stress at bending moment Ma , N. 1s

1 i f
the number of stress cycles corresponding to the fatigue limit of of the
shaft (usually 10&+10? cycles) and b is the slope of the S-N curve
taken as a positive quantity.
Substituting Eq. (9) back Into Eq. (8), noting aa = 32 Ma Ard3 (solid
1 1
circular shaft) and simplifying yields, the following expression for
calculating shaft diameter, d:

1/b 1/b
d3 _ 32 (FS)
iro. • ? (%) * 5? 1/b (10)

where the factor of safety term (FS) has been Introduced.

11
 Equation (10) can also be rearranged to find the life ML of a shaft of
given diameter for a prescribed operating duty cycle. Multiplying both sides
b
of Eq. (10) by i I \ where N. = total shaft life 1n cycles = n~ n3, etc.
^\
and solving for gives:
/b
ird
i32(FS)
N
L = n 1/b (11)

n n
2 , and nrp
3
where the terms rr- rp- are the fraction of time spent at each
N N
L L
bending load M , M , and M , respectively.
a
l a2 a
3
Effect of duty cycle. - Equation (11) can be used to Illustrate the large
detrimental effects that high loads have on shaft life. Consider the case
where a shaft 1s exposed to two blocks of alternating bending moments, where a
bending moment of amplitude M acts for n /N fraction of the time and M
a, I L a«
acts for the remainder according to the schematic appearing 1n F1g. 5.
Defining relative life L_,
L , to be shaft life when M * M divided by shaft
R a_ a,
life when M M then from Eq. (11) LK 1s found to be:

(12)
=M
1 -
N
Plotting Eq. (12) in F1g. 5 1t is clear that even a 20 percent overload
i«j d -i = 1.2\i acting only 20 percent of the time (n i /N L = 0.8) will cause a
/M
30 percent life reduction for b = 0.16 or a 64 percent life reduction for
b = 0.08 relative to a shaft with only constant amplitude loading. In prac-
tice, the life reduction would be closer to 30 rather than 64 percent since a
b - value of 0.16 1s more representative of a machined, mild steel shaft with
stress concentration while b = 0.08 would be representative of a smooth, notch
free (k-factorw 1), high strength shaft, (see F1g. 4). However, 1n any case,
this example points out that the high fluctuating loads acting on a structural
element, such as a shaft, tend to dictate its service life.

EFFECT OF MEAN STRESSES

The analysis presented 1s predicated on the knowledge of the S-N charac-

teristics of the shaft under the anticipated loading conditions. Modifying
factors have been Identified 1n Eq. (4) to correct specimen fatigue data for

12
certain geometric and envlromental factors that can affect fatigue strength.
The effects of mean stresses will be addressed next.
Since most shafts transmit power and rotate with gear, sprocket or pulley
loads, mean torslonal stresses are Invariably present. Also mean bending
stresses can be developed such as those due to rotating unbalance forces.
These mean stresses cause a reduction In fatigue strength. Residual stresses,
Induced deliberately or unattentlonally (see Table 2) behave like mean stresses
and can either benefit or reduce strength depending on whether they are com-
presslve or tensile [3 and 18].
The effects of mean stresses on long term fatigue strength are sometimes
available 1n the form of experimentally determined constant life diagrams [20].
In these diagrams the amplitude of the fluctuating stress 1s plotted versus
the magnitude of the mean stress at 104, 10$, etc. cycles to failure. Some-
times notched specimen data 1s Included. When specific data 1s unavailable,
mean stress effects are often approximated by certain mathematical failure
relations, such as Soderberg, Gerber and Modified Goodman failure lines [2, 3,
and 6]. When specific test data 1s available then, obviously, this 1s prefer-
red. However, the following discussion outlines how mean stresses can be
reasonably accounted for by knowing only the fatigue limit, yield and ultimate
strengths of the material.
Mean bending stress. - For the case when only bending loads are acting on
the shaft, that 1s zero torque, the loading 1s considered to be "simple" since
only one kind of stress 1s present. For simple loading several failure
relations have been proposed, but the modified Goodman line 1s, perhaps, the
most widely used. It 1s given by:
a o
-J « - * = , ! (13)
°f °u
where oa and om are, respectively, the alternating and mean components of
the simple bending stress, ou 1s the ultimate tensile strength and of 1s
the fatigue limit of the shaft material as determined from specimen fatigue
tests with no mean stress present.
Since oa = of when om = 0 according to Eq. (13), oa can be
Interpretted to be the bending fatigue limit strength 1n the presence of a mean
bending stress, say ofm. The asterUk 1s used to denote test specimen
rather than shaft fatigue properties. Thus the reduction of fatigue strength
with mean strength takes the form:

(14)
°f °u
In other words, the bending fatigue strength of the material decreases
linearly with mean bending stress, becoming zero when the ultimate strength 1s
reached (Immediate fracture failure).
Combined stress. - Host power transmitting shafts are not simply loaded,
but are subjected to combined stresses. The most common situation 1s a com-
bination of reversed bending stress (a rotating shaft with constant moment
loading) and steady or nearly steady torslonal stress. Although a large body
of test data has been generated for the simple stress condition, such as pure
tensile, flexural or torslonal stress, little Information has been published
for the combined bending and torslonal stress condition. However, some cyclic
bending and steady torslonal fatigue test data for alloy steel analyzed 1n [21]
13
 shows a reduction 1n reversed bending fatigue strength with mean torslonal
stress according to the elliptical relation:

(15)

where or 1s the reversed bending stress, that 1s, the alternating bending
stress component with no mean bending stress present, and where Tm and TV
are, respectively, the applied mean shear stress and yield shear strength of
the test specimen. Since or = of at tm = 0, Eq. (15) represents the
reduction of reversed bending fatigue strength with mean torslonal stress or,
1n other words.

(16)

where ofm has been Introduced, as before, to represent the fatigue limit 1n
the presence of a mean stress.
Superimposing the effects of both mean bending and torslonal stresses on
fatigue strength, that 1s combining Eqs. (14 and 16), results 1n

°fm = °f (17)

In the case of solid, circular shafts, the mean stress levels are
32 M m 16 Tm
and T
m = (18)
irdv trdv
and since, for steels.
(19)

then substituting Eqs. (18 and 19) back Into (17) yields:

°fm = °f
V 1 - 77.8 - 10.2

Thus, 1f a mean bending moment Mm and/or mean torslonal load Tm are

(20)

present, the fatigue limit of the specimen ofm can be found from Eq. (20)
and the fatigue limit of the shaft of 1s then:
"f = kakbkckdkekfkgkhk1°fm (21)
The above value of of can then be substituted back Into Eqs. (10 and 11) to
;f1ndshaft diameter and/or life.
14
 The Appendix contains an example to Illustrate how the proposed method Is
to be applied. As a precautionary note, 1t Is good practice to check 1f the
shaft diameter calculated from Eq. (10) 1s sufficiently large to withstand
static failure considering the combination of peak bending moment, Ma + Mm,
and mean transmitted torque Tm. Standard, static strength equations found
1n [2, 6, 14, and 15] and elsewhere can be used.
Unlimited life design. - For the special case of an unlimited life design
of a shaft having a constant amplitude bending moment Ma with both mean
bending and torslonal stresses present, the required shaft diameter d can be
found by substituting Eq. (20) back Into Eq. (10) and setting n-j = Nf and
a =
M . This gives:
1

32(FS)
(22)

Eq. (22) with Mm = 0 1s the basic shaft design equation proposed for
the soon-to-be-issued ASHE Standard 8106. 1 M, Design of Transmission Shafting
[21].

DISCUSSION OF RESULTS

As previously discussed, F1g. 5 shows the effects of a duty cycle consist-

Ing of two blocks of cyclic bending moments on relative shaft life. In a
similar manner, Eq. (11) can be used to study the Influence of duty cycle on
relative shaft diameter. This 1s Illustrated 1n F1g. 6. A representative
slope, b, value of 0.13 was selected which approximately corresponds to a
steel shaft of about 690 MPa (100 000 psl) tensile strength with k-factor =
0.5 according to F1g. 4. It 1s clear from F1g. 6 that a relatively small
Increase 1n shaft diameter 1s needed to accommodate a modest overload without
sacrificing shaft life. Furthermore, the relative constancy of this Increase
1n diameter, particularly at the higher overload values. Indicates that shaft
size 1s basically dictated by the highest bending load, even 1f 1t 1s only
present for a relatively small percentage of the duty cycle.
The effect of a mean torslonal load Tm and cycles to failure on
relative shaft diameter Is Illustrated 1n F1g. 7. To normalized this data,
relative shaft diameter has been arbitrarily set equal to 1.0 at 10& cycles
to failure and Tm = 0. These predicted curves were derived by substituting
Eq. (20) back Into Eq. (10) and considering that only a single cyclic bending
moment load of amplitude Ma 1s present. When this 1s done, the following
expression for relative shaft diameter dp results:
1/6
2b
+ 4
7 k-factor (23)

For purposes of Illustration, a representative case was selected where

ou = 690 MPa (100 000 ps1), k-factor = 0.5, b * 0.13, of * 0.5 (ou)
and 0.85

15
 Several general observation can be made about the trends appearing 1n
F1g. 7. First, the savings 1n shaft diameter for a limited life design at
10^ cycles to failure versus that for an unlimited life (fatigue limit)
design at 10& cycles becomes smaller as the transmitted or mean torque 1s
Increased. For example, a 26 percent smaller shaft can be used at Tm = 0
while a diameter reduction of just 12 percent 1s possible at Tm = 3 Ma.
Secondly, the required Increase 1n shaft diameter to accommodate an Increase
1n transmitted torque at constant shaft life Is relatively modest for high
cycle fatigue life designs. For example, an Increase shaft diameter of only 8
percent 1s needed to accommodate a transmitted torque that 1s 3 times the bend-
Ing moment amplitude (Tm = 3 Ma) at 10& cycles. However, at lower
cycles to failure, this Increase 1n diameter with transmitted torque becomes
greater, being about 28 percent at 103 cycles for Tm = 3 Ma.
Finally, the sensitivity of shaft diameter to changes In the required
cycles-to-fa1lure, Nf, although not shown 1n F1g. 7, 1s greater for higher
values of slope b according to Eq. (23). Thus the diameter of highly notched
(low k-factor), low strength steel shafts (see F1g. 4) will exhibit a greater
reduction with a decrease 1n design life than will those of smooth, high
strength shafts. Also, highly notched, high strength shafts (low k-factor,
high ay) will exhibited a smaller Increase 1n diameter with an Increase In
transmitted torque than will smooth, low strength shafts according to Eq. (23).

SUMMARY AND CONCLUSIONS

A shaft design method 1s presented which can be used to estimate the

diameter required to survive a specified number of stress cycles under a vari-
able amplitude loading history. The analysis 1s based on a nominal stress-life
method In which a straight line connects6 the true fracture strength at 1 cycle
to the fatigue limit of the shaft at 10 or 10? cycles-to-fa1lure on log-
log coordinates. A number of fatigue life modifying factors have been Identi-
fied to correct test specimen fatigue strength data for geometric and envlro-
mental conditions which the actual shaft will likely encounter 1n service.
Among such factors are surface condition, size, reliability, temperature,
stress concentration, press-fitted collars, residual stress and corrosion
fatigue. The effects of varU^le amplitude loading were Incorporated Into the
analysis using a Palmgren-M1ner linear damage approach. Mean bending stresses
were accounted for using a Modified Goodman failure relationship. The In-
fluence of a steady transmitted torque was considered through a elliptical
reduction 1n reversed bending fatigue strength with mean torslonal stress
exhibited by previously published fatigue test data.
The method presented was used to determine the effects of certain key
material and operating variables on shaft diameter and fatigue life. The
following results were obtained:
(1) The amplitude of the peak cyclic bending moment from a variable
amplitude loading history, even briefly applied, has a large Influence on
shaft diameter and/or fatigue life.
(2) The sensitivity of shaft fatigue life to bending stress 1s primarily
a function of tensile strength and the value of the fatigue life modifying
factor. For example, life typically varies with stress to about the -14 power
for small, smooth, high strength shafts and to the -5 power for large, rough,
heavily notched, low strength shafts.
/ (3) The sensitivity of shaft diameter to the presence of a mean or steady
transmitted torque 1s relatively small for high cycle fatigue life designs but
steadily Increases as the desired cycle life 1s reduced.
16
 (4) The savings In shaft diameter from a reduction 1n the required number
of cycles to failure Is greater at lower transmitted torque levels. This
savings becomes relatively small for shafts that carry a relatively high
amount of torque.

17
APPENDIX - APPLICATION EXAMPLE
To Illustrate application of the proposed method consider that a shaft 1s
to be designed with safty factor of 2 from SAE 1045 steel, quenched and tem-
pered Q&T (225 BHN, ou = 724 MPa and ay = 634 Mpa from Table ]) for 10°
000 cycles under a steady torque of 3000 N-M and the following variable bend-
Ing moment schedule:
M
aa1• Percent Number of cycles, Fraction of Nf
time
VNf
n
N-M
1
2000 15 15 000 0.015
1500 35 35 000 .035
1000 50 50 000 .050
100 100 000
The fatigue limit of a smooth 1045 steel specimen without mean stress
of 1s listed as 323 MPa at Nf = 106 cycles 1n Table 1. (Note this
1s somewhat smaller than the approximation 0.5 <*u or 362 MPa.)
Start with an Initial shaft diameter guess of d = 0.055 m (a good start-
Ing point 1s to calculate d from Eq. (10) assuming that no mean load Is
present). The effect of the mean torque of 3000 N-M on of can be found
from Eq. (20) as follows:

3000
a*m = 323xl06^/ 1 - 77.8
,0.0553 634xl06

= 313x1O6 N/m2
Let's assume that 1n this example that the product of all the k-factors
described by Eq. (4) 1s equal to 0.4, so the shaft's corrected bending fatigue
limit according to Eq. (21) 1s

of = 0.4 (313xl06)

= 125xl06 N/m2
For this material of 1s given as 1227xl06 N/m2, so the S-N curve
slope 1s
b = log (1227/125)/6
= 0.165 or 1/b = 6.05
Finally, for a FS = 2.0, the required shaft diameter d can be found from
Eq. (10) to be:

,3 =
32 (2.0) [o.015 (2000)6-05 t 0.035 (1500)6'05 + 0.05 (1000)6'05]0.165
l
w 125x10*

= 1.71xlO~4 m3
18
or
d = 0.056 m or 2.2 Inch
It 1s Instructive to note that 1f the calculation were repeated considering
that only the maximum bending moment of 2000 N-H acted 15 percent of the time
and that If the shaft ran unloaded the rest of the time, that 1s

then
d = 0.054 m or 2.1 Inch
The Insignificant reduction 1n shaft diameter from Ignoring the lower loads
clearly Illustrates the dominant effect that peak loads have on fatigue life.
This 1s also apparent from Eq. (3) where life 1s Inversely proportional to the
1/b power of stress amplitude. The exponent 1/b typically ranges from about 5
for heavily notched shafts to about 14 for some polished, unnotched steel test
specimens without mean stresses (see Table 1). Even at a modest 1/b value of
6, 64 times more fatigue damage 1s caused by doubling the alternating bending
moment or bending stress amplitude. This underscores the necessity of paying
close attention to overload conditions In both shaft and structural element
fatigue designs.

19
REFERENCES
1. Forrest, P. G., Fatigue of Metals, Pergamon Press, New York, 1962.
2. Juvlnall, R. C., Engineering Considerations of Stress and Strain, and
Strength. McGraw-Hill, New York, 1967.
3. Fuchs, H. 0. and Stephens, R. I., Metal Fatigue 1n Engineering. Wiley,
New York, 1980.
4. Collins, J. A., Failure of Materials 1n Mechanical Design: Analysis.
Prediction Prevention. John Wiley and Sons, New York, 1981.
5. Graham, J. A., ed. t Fatigue Design Handbook. SAE, New York, 1968.
6. Deutschman, A. 0., Mlchels, W. J., and Wilson, C. E., Machine Design:
Theory and Practice. MacMlllan, Inc., New York, 1975. pp. 331-360.
7. Borchardt, H. A., "Shortcuts for Designing Shafts," Machine Design. Vol.
45, Feb. 8., 1973, pp. 139-141.
8. Bagel, C., "Computer-Aided Fatigue Design of Power Transmission Shafts with
Strength Constraints using a Finite Line Element Technique arid a Proposed
Fatigue Failure Criterion," ASME Paper 79-DET-103, 1979.
9. Maschlno, S. and Kaftanoglu, B., "Computer-Aided Design of Power Shafts,"
ASME Paper No. 83-DET-63, 1983.
10. Wetzel, R. M., ed., Fatigue Under Complex Loading: Analyses and
Experiments. Vol. 6, Society of Automotive Engineers, Warrendale, Pa.,
1977.
11. "SAE Information Report Technical Report on Fatigue Propert1es-SAEJl099,"
SAE Handbook. Pt. 1, Society of Automatlve Engineers, Warrendale, Pa.,
1978, p 4.44.
12. "SAE Information Report: Estimated Mechanical Properties and Mach1nab1l1ty
of Hot Rolled and Cold Drawn Carbon Steel Bars-SAE J414," SAE Handbook.
Pt. 1, Society of Automotive Engineers, Warrendale, Pa., 1978, p. 3.14.
13. Heywood, R. B., Designing Against Fatigue of Metals, Relnhold, New York,
1962, p. 30.
14. Loewenthal, S. H., and Orthweln, W. C., Shafts, Couplings, Keys, etc.,
Mechanical Design and Systems Handbook. H. A. Rothbardt, ed., 2nd edition
McGraw-Hill Book Co., New York, 1984.
15. Loewenthal, S. H., Design of Power Transmitting Shafts. NASA Reference
Publication to be published, 1984.
16. Peterson, R. E., Stress Concentration Factors. Wiley, New York, 1974.
17. "Shafts with Interference-Fit Collars, Part IV: Fatigue Strength of Plain
Shafts," ESDU-68005-Pt. 4, Engineering Sciences Data Unit, London, 1968.
18. Almen, J. 0. and Black, P. H., Residual Stresses and Fatigue 1n Metals.
McGraw-Hill, New York, 1963.
19. Harris, W. J., Metallic Fatigue. Pergamon Press, New York, 1961, p. 48.
20. "Metallic Materials and Elements for Flight Vehicle Structures," M11 -
Handbook 5, Department of Defense, Washington, D.C., 1962.
21. Loewenthal, S. H., "Proposed Design Procedure for Transmission Shafting
Under Fatigue Loading," NASA TM-78927, 1978.

20
 TABLE 1. - REPRESENTATIVE STRENGTH AND FATIGUE PROPERTIES OF SELECTED STEELS

BASED ON TEST SPECIMEN DATA WITHOUT MEAN STRESSES FROM REFERENCES 11 AND 12

1
SAE spec BHn Process b *
description °u °y *f °f
Ult Str. Yield Str, Fat Str Cof, Fat Str, 106 cycles,
ksl (MPa) ksl (MPa) ksl (NPa) Exp ksl (MPa)

1005-1009 125 Cd Sheet 60 (414) 58 (400) 78 (538) 0.073 35 (244)

1005-1009 90 HR Sheet 50 (345) 38 (262) 93 (641) 0.109 29 (202)
1015 80 Normalized 60 (414) 33 (228) 120 (827) 0.11 27 (186)
1018 126 CD Bar 64 (441) 54 (372)
1020 108 HR Plate 64 (441) 38 (262) 130 (896) 0.12 30 (208)
1022 137 CD Bar 69 (476) 58 (400)
1040 170 CD Bar 85 (586) 71 (490)
1040 225 As Forged 90 (621) 50 (345) 223 (1538) 0.14 33 (233)
1045 225 Q&T 105 (724) 92 (634) 178 (1227) 0.095 47 (323)
1045 390 Q&T 195 (1344) 185 (1276) 230 (1586) 0.074 79 (547)
1045 500 Q&T 265 (1827) 245 (1689) 330 (2275) 0.08 104 (715)
1045 595 Q&T 325 (2241) 270 (1862) 395 (2723) 0.081 122 (843)
1050 197 CD Bar 100 (690) 84 (579)
1140 170 CD Bar 88 (607) 74 (510)
1144 305 Drawn at Temp 150 (1034) 148 (1020) 230 (1586) 0.09 66 (454)
1541F 290 Q&T Forging 138 (951) 129 (889) 185 (1276) 0.076 63 (435)
4130 258 QftT 130 (896) 113 (779) 185 (1276) 0.083 59 (404)
4130 365 Q&T 207 (1427) 197 (1358) 246 (1696) 0.081 77 (532)
4140 310 Q&T Drawn at Temp 156 (1076) 140 (965) 265 (1827) 0.08 90 (619)
4142 310 Drawn at Temp 154 (1062) 152 (1048) 210 (1448) 0.10 53 (366)
4142 380 Q&T 205 (1413) 200 (1379) 265 (1827) 0.08 83 (574)
4142 450 Q&T and Deformed 280 (1931) 270 (1862) 305 (2103) 0.09 83 (572)
4340 243 HR Annealed 120 (827) 92 (634) 174 (1200) 0.095 49 (337)
4340 409 Q&T 213 (1469) 199 (1372) 290 (1999) 0.091 80 (550)
4340 350 Q&T 180 (1241) 170 (1172) 240 (1655) 0.076 82 (567)
5160 430 Q&T 242 (1669) 222 (1531) 280 (1931) 0.071 103 (709)

Note: Values listed are typical. Specific values should be obtained from the steel producer.
Symbols:
CO = cold drawn
HR = hot rolled
Q&T = quenched and tempered

TABLE 2. - MANUFACTURING PROCESSES THAT PRODUCE RESIDUAL STRESSES

Beneficial residual Harmful residual

compress We stress tensile stress

Pre-stresslng or overstraining Cold straightening

Shot or hammer peenlng Grinding or machining
Sand or grit blasting Electro-discharge machining (EON)
Cold surface rolling Welding
Coining Flame cutting
Tumbling Chrome, nickel, or zinc plating
Burnishing
Flame or Induction hardening
Carburlzlng or nltrldlng
 FATIGUE STRENGTH COEFFICIENT

r-Of, CORRECTED
| FATIGUE LIMIT
' OF SHAFT -
CO
CO
_j
\
\
^-N,
b= log(a'f/Of)MogNf

I I I I I I
2 3 4 5
1 10 10 10 10 10 106 107
STRESS CYCLES TO FAILURE, N
Figure 1. - Generalized S-N curve constructed for o'f and Of.
 200
\ \

/
/-TENSILE STRENGTH
241 MPa (35 ksi)

8
x
| 100
UJ
ee.

o L-TENSILE STRENGTH
2 1030 MPa (159 ksi)-
oo 50
UJ

o
i— TENSILE
STRENGTH,
CS
z MPa. (ksi)
Q
LLj
D 0-483(0-70)
OQ O 483-690 (70 - 100)
TEST DATA, REF. 13
A 690-1030 (100 - 150)
O 1030+ (150+)
—- PREDICTED

20 I I I I
2 3
1 10 10 10 104 105 106 107 108 109
CYCLES TO FAILURE
Figure 2. - Predicted and measured effect of tensile strength on
the stress-life characteristics of steels in reversed bending
(from ref. 13).
 o
-2^0,-
•? o £
~J
o •!>"
t z
oo 2; -Si
•
o oe t/>
_
S <u o
=£
« fc «•&
W.J CJ^ Q)
•^ _i ° .•=
t-> VI
s «s
O 3= -^
CM
!«
u£
^ "=
<"
U_ QJ O
c/0 .2*
o u_
o
CM
'3Ayno N-Sjo 3dois
i/j -5^
OJ O
<u -o 1_
CL C
o
x ro ^
«5 ^J g
•o o .5
^| I^S
^ " 2f o E
o °- 2 TO
"S in -5
r- ••- en
o
10 °
t/> .E
^_«
•— o co
O CD CD
(_) t> S
i 2 •"
. TO -
0> — C
l_ CD CD
8
oo
8'
CM
'SS3N1S 9NiaN39 ONI1V10U
 i m
,i
0 t
8
CM . ._. CM
CD i . i ^T CO
t/l C
C* Q)
§/ / /
_ / » • o "£
t — ^••^
1w
^n
^~
Q)
^«
d> r—
S
r-4
to
UJ g
/ ' / ,„
Q>
« S-
E *
T / / / — co
Q

ti „
<^ £ ^i
* - • II
1 i | <t c <-> "-*
avoiONiQNsa / / j O 0 .S n
&" • ^ / ' / 9 Q
>, *> CH
o-Si
— O 5j -S: O ir» OO iI /1 I § =* J= °*
*t 5 P CO^ ^' -* CM CA 1 j 1 h™ *O *ca <~>
oi DO -• ^
~~
/I ;
•
1
1 "^ i - r* m
LU O l/> Q.
Ii 1' 1 i g S "g —
CVI 1 J3
1
^ «r
i ( i «~sl
I J1 III O *^
y313WVIQ JJVHS 3AI1V13U
-18
O
o
a>
V §
o>

o
"Jo
«J
v
° -
r--/v\
en ca
1 1

CM
\ Z \\
\ \
QV01 ONIQN38

\
\
\ \ Of
^5
O
evj

oo
uj
•r
OO
i i
3JH
 'm
Ms
1.1

1.0

o
t
.9

>
5 .8

103 104 105 10C

SHAFT LIFE IN STRESS CYCLES
Figure 7. - Effect of transmitted torque and life on shaft
diameter (notched shaft hawing k-factor = 0.5, tensile
strength = 690 MPa, slope b = 0.13, FS = 1).
 1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.

NASA TM-83608
4. Title and Subtitle 5. Report Date

Factors That Affect the Fatigue Strength of Power 6. Performing Organization Code
Transmission Shafting
505-40-42
7. Authors) 8. Performing Organization Report No.

E-1984
Stuart H. Loewenthal
10. Work Unit No.

9. Performing Organization Name and Address

11. Contract or Grant No.
National Aeronautics and Space Administration
Lewis Research Center
Oeveland, Ohio 44135 13. Type of Report and Period Covered

12. Sponsoring Agency Name and Address

Technical Memorandum
National Aeronautics and Space Administration 14. Sponsoring Agency Code
Washington, D.C. 20546

15. Supplementary Notes

Prepared for the Fourth International Power Transmission and Gearing Conference
sponsored by the American Society of Mechanical Engineers, Cambridge,
Massachusetts, October 8-12, 1984.
16. Abstract

A long-standing objective in the design of power transmission shafting is to

eliminate excess shaft material without compromising operational reliability. A
shaft design method is presented which accounts for variable amplitude loading
histories and their influence on limited life designs. The effects of combined
bending and torsional loading are considered along with a number of application
factors known to influence the fatigue strength of shafting materials. Among the
factors examined are surface condition, size, stress concentration, residual
stress and corrosion fatigue.

7. Key Words (Suggested by Author(s)) 18. Distribution Statement

Shafts. Unclassified - unlimited

Shafting STAR Category 37
Design
Transmission shafting
9. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of pages 22. Price"

Unclassified Unclassified
*For sale by the National Technical Information Service, Springfield, Virginia 22161
National Aeronautics and SPECIAL FOURTH
Space Administration
Washington. D.C.
20546
OflicWl
Penalty for Pfiian Ut». $300 *M*gt and F«« paid
Nation* Aeronautic* vid

NASA451

(WNSA rOSTM ASTER: If Undclhrcritilc (Srction IS*

PcwUI Manual) IV. Nul Return