Downloaded from SAE International by Purdue University, Monday, August 20, 2018

Design and Analysis Procedures

for Shafts and Splines

Paul E. Burke
Warner-Motive Div., Borg-Warner Corp.

Walter Fisher
Roy C. Ingersoll Research Center, Borg-Warner Corp.

Automotive Engineering Congress

Detroit, Mich. 680024
January 8-12,1968
Downloaded from SAE International by Purdue University, Monday, August 20, 2018
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

680024

Design and Analysis Procedures

for Shafts and Splines

Paul E. Burke
Warner-Motive Div., Borg-Warner Corp.

Walter Fisher
Roy C. Ingersoll Research Center, Borg-Warner Corp.

MATERIAL STRENGTH properties for steel are generally

compiled from static tensile and bending fatigue tests. The
data fall within failure bands when nominal stress is plotted
against surface hardness (Fig. 1).
Nominal stress is defined by the conventional stress equa-
tions :

MODE OF LOADING

Allowable loads for various modes of loading, such as

completely reversed, zero to maximum, and plus to plus,

ABSTRACT

A general analytical and test procedure approach is pre- derived from fatigue tests are presented for torsional fatigue
sented for shafts under fatigue loading conditions. Construc- loading. The concept of nominal stress is used in correlat-
tion and interpretation of comprehensive analytical curves ing test and analytical curves, introducing simplicity and
are outlined, comprising the basic loading conditions of perspective in interpreting such complex factors on fatigue
torsion alone, bending alone, combined bending and tor- life as stress concentration, physical properties, and mode
sion, and cumulative damage. General spline design curves of loading.
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

are indicated on the modified Goodman diagrams. Fig. 2 ditions of torsion and bending considered separately. The
illustrates an infinite life Goodman diagram for torsion with endurance limit in terms of nominal stress or loading in
a specific zero to maximum loading condition indicated. terms of bending or torsion for the particular mode of load-
The diagram may be defined in relation to allowable nom-
ing established by the Goodman diagram is plotted at 106
inal stress or torque scales. Similarly, Fig. 3 illustrates a
bending diagram with a completely reversed loading con- stress cycles. The nominal yield stress is taken as the allow-
dition indicated. The two conditions represent common able loading condition at 103 stress cycles. The straight
loading conditions for shafts. line connecting the two points will establish the slope of
the analytical failure curve and approximate the slope of
LOAD VERSUS CYCLES CURVE
the actual test failure curve.
Figs. 4 and 5 illustrate theoretical load versus life curves,
EFFECTS OF STRESS CONCENTRATION
on logarithmic scale coordinates, for the two loading con-

Stress concentrations merely change the slope of the

failure curve, as illustrated by shafts with various fillet radii
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

in Fig. 6, and may be simply accounted for by a reduction NOMINAL STRESS RELATIONSHIPS
in the allowable nominal stress at the endurance limit or
The method of describing stress in terms of nominal stress
106 cycles (1)*. The effects of stress concentration in re- rather than true stress retains the fundamental equation re-
ducing strength vary with life, having a maximum effect
lationships of strength of materials for fatigue as well as
at the endurance limit and a progressively decreasing effect
for static loading. Some of these relationships for shaft
with decreasing life. The stress reduction at 103 cycles is design are shown in Figs. 8-10.
negligible and establishes the graphical pivot point for the
analytical failure curves. A maximum strength reduction

factor of approximately 3 at 106 cycles is reached for the

most severely notched condition, establishing a maximum
slope to the failure curve.

COMBINED BENDING AND TORSION

Combination fatigue loading conditions of bending and

torsion may be described by similar theoretical curves of
the form summarized in Fig. 7. These combined loading
curves and cumulative fatigue damage are covered in later
sections of this paper.

*Numbers in parentheses designate References at end of

paper.
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

SUPERPOSITION OF FAILURE BANDS failure band can be obtained and the load versus load cycles
failure band of any size part of similar geometry in a pro-
Load cycle fatigue bands of similar type parts of different duct line can be predicted, based on a series of fatigue tests
sizes frequently can be correlated on a common basis by of one size.
means of the conventional fundamental nominal stress Eqs.
1-3. FATIGUE LIFE TESTS
Use of nominal stress permits the position of the load
versus cycles failure band to be displaced parallel to itself The analytical approach to fatigue will establish the
on a log-log scale plot and its position is primarily a func- order of magnitude and slope of the failure curve and the
tion of: effects of changes in geometry or material in a production
1. The diameter cubed in torsion. part. Ideally, it should be supplemented by fatigue tests
2. The area in simple tension. and the tests compared to the theoretical curves for inter-
3. The depth of section squared in bending of rectangu- pretation in a critical design (Fig. 11).
lar sections. The objective of fatigue life tests is to define the entire
The result is that a common stress versus stress cycles load versus load cycles failure band. Defining the failure
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

band involves obtaining the width and slope of the band fatigue life tests to check his strength estimates. When
over as large a spread of load and life as possible. This is only one part is available for testing, the suggested proce-
generally from 1000 to 10,000,000 stress cycles. At the dure is to establish an analytical failure curve as outlined
high load end, tests may be limited by yielding of the test above and obtain a failure test point near infinite life as
piece or deflection (amplitude) limitations of the testing a check. The strength reduction effects of stress concentra-
machine. The number of test points required for completely tion are maximum in this life region and the order of magni-
defining the failure band will vary, depending upon the load tude of the slope of the failure band and the strength level
scatter (width of the failure band in the load direction) and of the part can be approximated. Such single point testing
the slope of the band. The greater the load scatter and the can be of value if it can be assumed that the entire failure
shallower the slope, the more the number of test points band is relatively narrow.
necessary to define the band.
When testing with a limited number of parts, it is recom-
FATIGUE LIFE TESTS WITH LIMITED mended that failure points be obtained at various loads,
NUMBER OF PARTS uniformly distributed, to establish the width and slope of
the failure band. The fatigue band is bracketed in by lines
An engineer is usually faced with the problem that only estimated to be essentially 0-100% failures. The failure
one or a few critical prototype parts will be available for band obtained in this way is generally sufficiently accu-
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

rate to evaluate the fatigue life of the entire shaft for most wide scatter in fatigue life of a high quality part with uni-
engineering purposes. form mechanical properties.
2. The slope of the failure band is defined. The slope of
VARIABLE LOAD FATIGUE LIFE TESTING the failure band is a measure of the degree of effective
fatigue stress concentration. When it is required to increase
The reasons for testing at various loads rather than at fatigue strength and life of a part, knowing the slope of the
one particular load are as follows: failure band will indicate the magnitude of gain that can
1. The scatter in relation to both load and life are de- be obtained by relieving stress concentration due to stress
fined. Testing at one load shows only the scatter in life raisers, such as oil holes, fillets, notches, etc.
at that particular load. It does not indicate the effect of 3. Multiple critical failure locations are determined.
variations from the selected test load. The significance of In parts of complex geometry with many areas of stress con-
life scatter should be judged in relation to the band width centrations, the failure locations may change, depending
in the load direction. For example, tests may show exten- upon the loads. Testing at one load level will not reveal
sive and frequently large scatter with respect to life at one all critical failure locations. Each failure region on the test
particular load level and yet have a narrow and well-de- piece may be considered to have its own failure band. A
fined shallow sloping failure band which is simple to apply crossover of failure bands can occur and result in shifting
in a design. This would indicate a normal and expected the location of failure on the part. This is due to differences
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

in size, strength, and stress concentration of the critical EXAMPLES OF FATIGUE LIFE ANALYSIS PROCEDURES ON
areas. It is, therefore, possible to strengthen a critical area PRODUCTION TYPE SHAFTS
of a part without significantly increasing the load carrying
capacity or life of the part as a whole. What can result TRANSVERSE HOLES - Torsional fatigue test failures
is merely changing the failure location on the part. at a transverse hole and a spline are shown in Fig. 12. Fail-
ure of a design modification with four transverse holes in
line is shown in Fig. 13.
Fig. 14 illustrates a modified Goodman diagram for steel
in torsion of Rc 60 hardness. Allowable nominal stresses
for specific zero to maximum stress loading conditions are
indicated for fatigue stress concentration (stress reduction
factors)of 1, 2, and 3. A comparison of test failures at the
single transverse hole with the theoretical nominal stress
versus stress cycles curves (Fig. 15) indicates correlation
with a stress reduction factor, KF, of approximately 2. Fa-
tigue tests of four transverse holes in line indicate a far
greater degree of strength reduction than predicted by theory
and emphasize the importance of confirmation by fatigue
tests.
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

rig. 16 shows fatigue test failures at the transverse hole

with various shaft hardnesses. The trend is increased fatigue
strength with increase in hardness for shafts in torsion with
stress raisers.

Correlation of the magnitude of strength reduction due

to the transverse hole with various shaft hardnesses is illus-
trated in Fig. 17. Endurance limits are superimposed and
relative strengths are compared with theoretical curves of
various hardnesses. Test points indicate a fatigue stress
concentration factor of approximately 2 independent of
hardness in the hardness range of 30-60 Rc, showing no change
in notch sensitivity with hardness for this particular shaft.
A few failures occurred at the spline near 106 cycles, indi-
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

eating a failure crossover, as mentioned previously, which straight-sided spline. A crossover type of failure as pre-
is illustrated in Fig. 18. viously discussed occurred at the transverse hole (Fig. 36).
Fig. 19 illustrates a shaft of complex geometry with a Torque versus stress cycles to fatigue failure tests are
torsional fatigue test failure at the transverse hole shown plotted in Fig. 37 for the two types of splines. Using the
in Fig-. 20, and photoelastic stress distribution as shown in correlation principle of superposition of nominal torsional
Fig. 21. The fatigue test failures (Fig. 22) are compared shear stress indicates that the fatigue failure bands super-
with theoretical curves of Rc 30 hardness for zero to maxi- impose using the mean diameter of the spline in the nominal
mum torsional loading. The strength reduction factor due stress equation(s) (Fig. 38):
to the transverse hole ranges 1.0-1.5.
SPLINES - A 10 tooth straight-sided spline (Figs. 23 and
24) torsional fatigue test failure is shown in Fig. 25. The
origin of the fatigue failure is at the bearing at the end of
the hub on the shaft spline teeth, indicating the fatigue
stress concentration effects at this point are greater than at
the spline runout. The hub was positioned well back from
the spline runout (Fig. 23).
Figs. 26 and 27 illustrate the relative stress distribution
along the splined shaft. The tooth bending stress reaches
a peak at the end of the hub bearing on the shaft and de-
creases rapidly in either direction along the shaft axis. Tor- From these and similar type spline tests, spline design
sional shear stress in the fillet also declines rapidly in the allowable stress curves are established (Fig. 39). The curves
direction into the hub and falls off slightly along the spline
runout due to increase in section size. The relative stress
distribution across the shaft spline section is indicated in
Figs. 28 and 29.
A hub fatigue failure is shown in Fig. 30 with corre-
sponding photoelastic pictures of relative stress distribution
in the hub (Figs. 31 and 32).
Another spline fatigue test series was conducted with the
32/64 pitch involute splined shaft shown in Fig. 33. Typi-
cal torsional fatigue spline failures are shown in Figs. 34
and 35. Fatigue failures originated at the end of the hub
bearing on the shaft spline as in the case of the 10 tooth
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

10
Downloaded from SAE International by Purdue University, Monday, August 20, 2018

11

are for zero to maximum torsional stress loading for splines

of hardness Rc 60.
It must be emphasized that the nominal stresses are used
as correlation numbers and are not true stresses. A spline
is subjected to extremely complex and interacting stress
conditions, of which the dominant factor appears to be nom-
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

12

inal torsional shear. Other factors influence the position COMBINED BENDING AND TORSION
of the point of failure within this failure band, such as spline
fit, runout geometry, hub position and bearing location, Mohr's circle representation for combined bending ten-
material and heat treat, etc. As an example (Figs. 40 and sile and torsional shear stress (Fig. 42) illustrates graphic-
41), a hardness traverse across the 32/64 pitch spline section ally the elliptical form of the fundamental combined stress
indicates that the deeper case depth of points A and B have equation:
greater fatigue strength than the shallower case of points
C and D, and fail nearer the high side of the failure band.
Downloaded from SAE International by Purdue University, Monday, August 20, 2018

13
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

14

Equivalent combinations of nominal stresses of bending

and torsion in terms of maximum shear stress are shown in
Fig. 43.
A similar elliptical curve may be applied to fatigue
analysis in the form of a constant life ellipse:
Downloaded from SAE International by Purdue University, Monday, August 20, 2018

15

where:

Sx = Nominal bending tensile stress, psi

Sxy = Nominal torsional shear stress, psi

S 1 = Allowable nominal bending tensile stress for

x
the required life in bending alone (intersection
of the ellipse with bending stress scale, zero
torsional shear stress), psi
1
(Sxy ) = Allowable nominal torsional shear stress for the
required life in torsion alone (intersection of
ellipse with torsion stress scale, zero bending
tensile stress), psi

Examples of the application of the constant life ellipse

are illustrated in Fig. 44. Conditions are for steel shafts
of Rc 60 hardness. Curves A, B, C, and D represent allow-
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

16

able combinations of nominal bending and torsion stresses able stress cycles of 1000, 10,000, 100,000 and infinite life
which will result in 1000, 10,000, 100,000, and 1,000,000 under zero to maximum torsional loading. A stress concen-
allowable stress applications for both bending and torsion. tration factor of 2.0 is assumed.
Completely reversed stress loading and no stress raiser is
assumed. CUMULATIVE FATIGUE LIFE DAMAGE
Curves E, F, G, and H represent allowable combinations
of nominal bending and torsion stresses allowing infinite Miner's theory of cumulative damage (2) sums fractions
life completely reversed bending in combination with allow- of life expended at the particular stress levels in relation to
a failure life curve. Equal fractions of life expended defines
a curve parallel to the original failure curve; so, therefore,
Miner' s theory may be reexpressed by the following equa-
tions. Assuming a log-log scale straight line conventional
stress versus stress cycles fatigue failure curve:

Fatigue damage in terms of expended life at a particular

stress above the endurance limit would, according to Miner's
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

17

terion, which adds a new perspective in evaluating test

correlation with Miner's cumulative damage theory.
Expended and remaining life in relation to the original
fatigue failure band is illustrated graphically in Fig. 46 with
the following notations:

ni = Life expended
N = Minimum total life to failure
i
N + Ni = Maximum total life to failure
i
Ni = Fatigue life variation
Ni - ni = Minimum remaining life
Ni + Ni - ni = Maximum remaining life
= Average slope of the failure band

Fatigue damage at any number of stress levels is taken

as the sum of all equivalent expended lives referred to any
one arbitrary stress level (Fig. 47):

Fatigue failure occurs when the total number of equivalent

expended life stress cycles falls within the conventional
S/N failure band at a given stress level.
theory, have an equivalence in expended life at any other Failure criterion:
stress level defined by a line parallel to the original fail-
ure curve. Each point on the line represents the same frac-
tion of life expended and the life expended in terms of stress
cycles can be expressed by the relation (Fie. 45):
Fig. 48 shows apparently erratic correlation between test
failures and predicted failure life (2). However, when cumu-
lative damage expended life is referred to the original fa-
tigue failure test band, the same test results indicate good
correlation (Fig. 49).
A graphical damage summation example selected from
Miner's tests is illustrated in Fig. 50.

REFERENCES

1. P. E. Burke, "Shaft Design for Finite Life." Paper

Miner's damage life is usually referred to a single fail- 359D presented at SAE Summer Meeting, St. Louis, June
ure curve and neglects the normal fatigue scatter variation 1961.
of failure life. The procedure presented in this paper uses 2. Milton A. Miner, "Cumulative Damage in Fatigue."
the original S/N fatigue failure test band as the failure cri- ASME Transactions, Vol. 67 (1945).
This paper is subject to revision. Statements and opinions been edited by SAE for uniform styling and format. Discussion will be printed
advanced in papers or discussion are the author's and are with the paper if it is published in SAE Transactions. For permission to publish
his responsibility, not the Society's; however, the paper has this paper in full or in part, contact the SAE Publications Division and the
authors.

20 page booklet. Printed in U . S . A .

Downloaded from SAE International by Purdue University, Monday, August 20, 2018
 Downloaded from SAE International by Purdue University, Monday, August 20, 2018

ERRATA

PAPER NO. 680024

Page 14, column 2, Eq. (6) should read:

Page 15, column 2, Sx 1 = Allowable nominal bending . . . should read:

Page 17, column 2, Ni + Ni - ni = Maximum remaining life, should read:

Ni + Ni - ni = Maximum remaining life