293

EFFECT OF LARGE HYDROSTATIC PRESSURES ON THE

TORSIONAL FATIGUE STRENGTH OF TWO STEELS
By D. J. Burns* and J. S . C. Parryt
This paper describes tests in which two alloy steels were subjected to reversed torsion at
atmospheric or high ambient pressure. It is shown that if the specimens are protected,
superimposed fluid pressures increase the fatigue limits of both steels. For instance a
pressure of 25 ton/in2 raises the nominal fatigue limit of a 2+ per cent nickel-chromium-
molybdenum (En 25) steel by about 46 per cent. It is concluded tentatively that a
superimposed hydrostatic pressure improves fatigue strength by introducing a com-
pressive stress on planes of maximum alternating shear stress.

INTRODUCTION dence (4) (5), that even with reversed stress below the
ALTHOUGH the nucleation of fatigue cracks during cycles of primitive yield, some error may occur in the calculation
small plastic strain has been related to the behaviour of of the stress fields in solid torsional fatigue specimens, as
slip bands at free surfaces (I) (2)$,there is still consider- cyclic stress can modify the stress-strain relationship
able uncertainty as to the parts which tensile and shear (dynamic yielding). For example Chodorowski (5) found
stresses play during the nucleation and subsequent pro- that the dynamic limit of proportionality of En 25 steel
pagation of the cracks. A better understanding of this was about 20 per cent lower than the static limit.
aspect of the problem would be invaluable since many In a related series of experiments White, Crossland
engineering components are subjected to complex stresses and Morrison (6)designed a machine to subject specimens
and have to be designed from a knowledge of the uniaxial to cyclically varying direct stress with additional tensile or
behaviour only. compressive mean stress, at atmospheric or high fluid
Since a triaxial stress will be present at the root of any pressure. They found that the semi-range of reversed
stressed crack, irrespective of its size, it is pertinent to direct stress to cause fatigue failure in rubber-covered,
ask what part triaxial stresses play in the fatigue process. solid, En 25 steel specimens was increased from 30.7 to
T o investigate the effect of static triaxial stresses Crossland 33 ton/in2 by an increase in pressure of 20 ton/in2, i.e. an
(3) built a machine to subject a solid specimen to torsional 8 per cent increase. This was a much smaller percentage
fatigue with a superimposed fluid pressure. Crossland found increase in fatigue strength than that found for torsion
that the oil used as the pressure medium had a deleterious by Crossland (3) and it was suggested (7) that dynamic
effect on the fatigue behaviour of his 2+ per cent nickel- yielding had occurred in Crossland's specimens and given
chromium-molybdenum steel specimens (heat-treated to a misleading result for the effect of the superimposed
an ultimate tensile strength of 56 ton/in2; hereafter re- pressure.
ferred to as En 25) irrespective of the ambient pressure.
T o minimize this effect Crossland protected specimens The aims of the present work were:
with a thin rubber coating and found that, when the (1) To expand Crossland's experiments on the effect of
ambient pressure was increased from atmospheric to static fluid pressure on the fatigue behaviour in reversed
20 ton/in2, the semi-range of shear stress at the fatigue torsion of solid specimens of En 25 steel.
limit was raised from 19.2 ton/in2 to 25.4 ton/in2, 1.e. . a (2) T o repeat part of the above programme with thin-
32 per cent increase. walled tubular specimens of the same steel in order to
However, Crossland's results are subject to some un- avoid, and therefore by comparison to assess, the un-
certainties of interpretation. I n particular, there is evi- certainties of interpretation caused by dynamic yielding in
The M S . of this paper was received at the Institution on 5th May
solid specimens.
1964. (3) To repeat some of the above tests on nitrided and/or
* Lecturer in Mechanical Engineering, Imperial College of Science rubber-covered solid specimens of a 3 per cent chromium
and Technology. steel, heat-treated to an ultimate tensile strength of
f Lecturer in Mechanical Engineering, University of Bristol.
?: References are given in the Appendix. 54 ton/in2, and hereafter referred to as En 40B. It has
JOURNAL MECHANICAL ENGINEERING SCIENCE Vol 6 No 3 1964
2 94 1). J. BURNS AND J. S . C . PARRY

been shown that nitrided fatigue specimens usually fail environment and of anisotropy and inhomogeneity in the
below the surface and it has been suggested (8) that a test materials.
nitrided fatigue specimen fails when the stress at the
The programme of fatigue tests carried out on the alloy
interface between nitrided and unnitrided material
steels is given in Table 1. The number of specimens tested
reaches the fatigue limit of the unnitrided material. It
at each state and the relevant fatigue curve figure numbers
occurred to the authors that, if this suggestion was cor-
are also given in this table.
rect, nitriding could be used as an alternative to rubber
coating for protecting torsional fatigue specimens when
testing with a superimposed fluid pressure. A comparison APPARATUS F O R HIGH-PRESSURE TESTS
of tests at atmospheric and high pressure on nitrided The fatigue machine used for these experiments is a
and/or rubber-coated specimens of a suitable steel would modified form of the one designed by Crossland (3).
then give some measure of the effectiveness of the rubber Figs 1 and 2 show sectional elevation and photographic
coating protection technique and at the same time con- views of the modified machine.
firm that the high-pressure effect revealed in the afore- Basically it consists of a torque input shaft ( S and L in
mentioned En 25 tests was common to other alloy steels. Fig. l), which is oscillated torsionally through a con-
(4) It was also necessary to carry out numerous static necting rod (A in Fig. 2), by a variable throw crank
torsion and tension tests and fatigue tests in reversed (B in Fig. 2). This torque input shaft enters the main
torsion at atmospheric pressure to assess the effects of high-pressure vessel (T in Fig. 1) through a Morrison

Table 1. Test programme

- ~~~ ~~ ~~

Material Type of torsion Specimen surfaces Position of Environment Ambient dumber of Fig. No.
specimen were : specimen axis pressure, ipecimens
in relation to tonha tested
longitudinal axis of
original bar
11 6
6 Kot shown
19 6 and 7
10 7

13 7

10 6

1-Rubber -covered-)
I iLi1 - -I -15
-20
10
8
5
9

9
6

E n 2 5 --- I 1-25 8 9
I-Longitudinal -Oil -~20 11 6

-Thin-walled I-
tubes
I
--Unprotected --Transverse ---
c
,-Air
--p-20
Atmospheric

Atmospheric
7

9
4
8

8
--Rubber-covered -Transverse __ ,--I0 14 9
-oil -20 9 8
I

'-25 6 9
-Air ~-
Atmospheric 11 12
<-Rub ber-covered --Transverse ___
i-oil 20
i 6 12
~ ,-Air ~-
Atmospheric 7 13
En 40R ----Solid ------]--Nitrided ----Transverse ___ I
'-Oil ------2O 6 13
I Nitrided and
'-rubber-covered -Transverse __- Oil --20 6 13

* Atmospheric pressure. j- Oil used in intensifier.

10C R ?i A L M E C H A N 1 C AL E N G I N E E R I N G S C I E N C E Vol 4 hro 3 I964
 EFFECT O F LARGE HYDROSTATIC PRESSURES ON THE TORSIONAL FATIGUE STRENGTH OF TWO STEELS 295

TO INTENSIFIER

TO I N T E N S I F I E R

A Torsion grips H Retaining plug P Copper washer

B Retaining plug J Mirrors Q Thrust race
C Specimen K Retaining nut R Morrison seal
D Torque tube L Part of torque shaft S Part of torque shaft
E Mirror shaft M Retaining plug T Specimen pressure vessel
F High-pressure joint N Taper joint in boss U Pedestals
G Worm gear 0 Counter pressure chamber
Fig. 1. Sectional arrangement of high-pressure assembly

veal ( 9 ) and connects through a special grip (3) with the both ends of it; this requires two pressure cylinders
specimen. These grips are designed so that only a pure (T and 0 in Fig. l), back to back. For convenience of
couple is applied. The specimen is connected through a machine assembly, the torque input shaft is in two parts,
second grip with a torque tube (D in Fig. I), which is S and L, which are connected by a taper joint, N, and
fixed rigidly at its other end to the main pressure vessel. behave as a solid member when pressure is applied to both
The angle through which the free end of this torque tube cylinders.
rotates is measured optically with mirrors, J. Since the The intensifier used to generate the high pressure for
tube is designed to operate in the elastic range (9) (IQ), the two chambers, T and 0, is capable of producing
this angular measurement is proportional to the torque in pressure up to 46 ton/in2 (11). The modifications to
the specimen. enable it to run automatically at a set pressure have been
I n this modified machine, end thrust on the torque described by Crossland (3). A relay system was installed
input shaft is almost eliminated by applying pressure to for the present experiments to stop the fatigue machine
J 0 C R K A L M E C H A N I C A L ENGINEER I N G S C I E N C E Vol6 No 3 1964
296 D. J. BURNS AND J. S. C . PARRY

(A) Connecting rod. (B) Variable-throw crank. (C) Driving shaft. (D) Contact breaker.
Fig. 2. High-pressure fatigue machine

if the intensifier failed or ran at any pressure other than than the static range because of the clearance in the bear-
the required value. ings but this is allowed for when the eccentric is set.
Recordings of maximum and minimum torque are taken
Experimental technique at intervals during a test. When a specimen starts to break
When the specimen is in position the two halves of the the torque range decreases and the machine is stopped by
main pressure vessel are bolted together and the cranked the observer when the torque range is zero.
connecting rod (A in Fig. 2), is fixed to the boss. The
eccentric is set to an approximate mean position; the MATERIAL AND SPECIMENS
cylinders (T and 0 in Fig. 1) are pressurized and the En 25 steel
eccentric is adjusted to a very small throw. The mean The majority of the work was carried out on two bars of
torque can now be adjusted to zero by rotating the pres- this steel. These were supplied as 2&in diameter 36-in
sure vessels on their supporting pedestals with the worm long bars which had been individually heat-treated by the
wheel and gear, G in Fig. 1. manufacturers; particulars of this heat-treatment and the
The throw of the eccentric is then adjusted to give the material composition are given in Table 2.
desired maximum and minimum torque readings. The
machine is then started; the speed being increased from En 40B steel
zero to 1800 c/min in a few seconds.
It is possible to measure, while the machine is running, This 3 per cent chromium steel was supplied as 2 5 i n
the maximum and minimum torque levels or the torque diameter 13t-in long bars, which had been individually
at any point in a cycle by using a method devised by one heat-treated by the manufacturers; Table 2 gives details
of the authors (J.S.C.P.). A contact breaker D is fitted to and material composition. Only one bar of this steel was
the end of the driving shaft C (Fig. 2), and by rotating the used; some of the specimens were nitrided at 495°C for
casing manually the position in the stress cycle at which 24 hours.
contact is made can be varied. The contact breaker is used
to trigger a stroboscope which illuminates a scale. The Specimens
reflections of this scale, in the fixed and moving mirrors The solid and tubular fatigue specimens are shown in
on the torque tube, are observed with a telescope. Fig. 3. By careful machining it was possible to obtain
Usually the dynamic torque range is slightly larger tubular specimens in which the variation in wall thickness
J O U R N A L M E C H A N I C A L E N G I N E E R I N G SCIENCE Vol6 No 3 1964
 EFFECT OF LARGE HYDROSTATIC PRESSURES ON THE TORSIONAL FATIGUE STRENGTH OF TWO STEELS 297

Table 2
~~

Material Composition percentages

C I Si I M n I S I I? Ni 1 Cr ~ M o

1
~

En 40B 0.25
' 0.21
0.19
0.60
0.45
11 0.034
0.014
0.025
0.015 1
I 2.68
0.22
~ 0.68
3.27
I, 052
0.51
850°C
910°C
4 hours at 645'C
4 hours at 625-C
I
~

.
2 ' / 8 in
a Solid.

0.280in DIA. 0.375 in

I1/4 in RADIUS

k
\ /roe330 in DIA.

W A L L THICKNESS
0009 in 2 0.001
D I A REAMED
AND HONED

b Tubular.
Fig. 3. Torsional fatigue specimens

around the circumference was less than 0.005 in; wall axes parallel (longitudinal) or perpendicular (transverse)
thickness was measured on an instrument designed by to the longitudinal axes of the bars.
Chodorowski (5). The bores of these tubes were honed to No significant difference was observed in the stress-
about 8 pin c.1.a. strain curves, in either torsion or tension, for adjacent
All fatigue specimens were finished externally on a longitudinal and transverse specimens. However, trans-
mechanical polishing machine with successively finer verse torsion specimens failed at a smaller angle of twist
grades of emery paper, using 000 grade paper for a final and transverse tension specimens had a much smaller
longitudinal polish. elongation and reduction in area than the corresponding
After the final polishing all specimens, except those that longitudinal specimens, Table 3.
had been nitrided, were given a stress-relieving heat treat- No significant variations in circumferential properties
ment in a high-vacuum furnace; 1 hour at 600°C for were found from tests on sets of both torsion and tension
En 25 specimens and 1 hour at 575°C far En 40B speci- specimens.
mens. The average values obtained along the two bars are
given in Table 3. The shear stress data listed in Table 3
were obtained from shear stress-strain curves which had
RESULTS: STATIC TESTS been derived from the torque-twist curves using a con-
En 25 steel struction proposed by Nadai (12); typical curves are
Static torsion and tension specimens were cut with their shown in Fig. 4.
l O U R N A L ,MECHANICAL ENGINEERING SCIENCE Vo16 No 3 1964
7
298 D. J. BURNS AND J. S . C. PARRY

Table 3. Aaerage static properties of materials logarithm of the number of cycles to failure, where T is
I I the measured torque, r the minimum outside radius of
Material I En25 ~ En40B the specimen and J the polar second moment of area of the
minimum cross-section.
Tensile properties*:
Limit of proportionality . As yield 37 The nominal shear stress is equal to the true maximum
0.10 per cent proof stress . As yield 43.5 stress in a solid specimen if it behaves elastically but if
Yield stress . . . 49 None yielding occurs the true maximum stress will be lower.
Ultimate stress . 57 53.8
However, in the case of tubular specimens, where the
Reduction in area:
Longitudinal specimens . .1
.1
62 per cent ’ 55 per cent stress variation across the wall thickness is always less than
Transverse specimens .
Shear properties*:
41 per cent
I - 8 per cent, the true maximum stress is very nearly equal
to the nominal stress under all conditions.
Limit of proportionality . As yield 20 The results are from approximately constant strain
0.17 per cent proof stress . As yield 24
Yield stress . 27.5 None experiments, any variation in range of nominal shear stress
Ultimate stress . 38.6 35 during a specimen’s life being indicated on the fatigue
Strain to failure :
Longitudinal specimens. . 3.0 - graph by the length of the vertical line, fatigue limits being
Transverse specimens . 1.4 2.0 determined from the semi-ranges of nominal shear stress
to which the specimens were subjected in the first few
* All stresses in ton/in2. cycles (hereafter referred to as the starting semi-range of
nominal shear stress at the fatigue limit).
En 40B steel En 25 steel
A typical torque-twist curve for this material is shown in To investigate the effect of material anistropy and in-
Fig. 5. Derived data from tests on torsion and tension homogeneity, fatigue tests in reversed torsion were carried
specimens are given in Table 3. out in air on solid longitudinal and transverse specimens
Two static torsion specimens were nitrided. These from both bars in a machine designed by Chodorowski (5).
torsion results are also plotted in Fig. 5. It has been The results of these tests, Fig. 6, show that the difference
assumed, when calculating the stresses in the nitrided in the fatigue limits for adjacent longitudinal and trans-
specimens, that the core material and the nitrided coating verse specimens and for all specimens from the two bars
have the same elastic constants. was not greater than 3 per cent. Consequently, both longi-
tudinal and transverse specimens were used for the fatigue
RESULTS: FATIGUE TESTS
tests with a superimposed high fluid pressure. A compari-
General r e m a r k s on presentation son of the ‘high-pressure’ fatigue behaviour of longi-
Each set of results is presented as a graph of semi-range tudinal and transverse specimens, Fig. 6, justified this
of nominal shear stress, calculated from Tr/J, against the

0
I
0.5
t
1.0
I
1.5
I
2.0
SHEAR STRAIN (TAN 0 )
I
2.5
-3.0
35

3.5

Fig. 4. Typical static torsion curves for En 25 steel

J 0 U K N A L hl E C H A NI C A I. E N G I N E E R I N G S C I E K C E Vol 6 No 3 1964
 LbFECT OF. LARGE HYDROSTATIC PRESSURES ON TIE TORSIONAL FATIGUE STRENGTH OF TWO STEELS 299

0 0-2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2-2
SHEAR STRAIN (TAN
o Normal. 0 Nitrided.
Fig. 5. Typical static torsion curves for En 40B steel

T o investigate the effect of environment, unprotected trates. An examination of Figs 6 and 8 shows that rubber
solid and tubular specimens were tested in air* or in the coating, which has no intrinsic strength, increases the
intensifier oil at atmospheric or 20 ton/in2 pressure. The strength of specimens tested in air by a small amount;
results of these tests, Figs 6, 7 and 8, confirm Crossland’s this is probably due to the exclusion of oxygen from the
observations that the presence of the intensifier oil at surface. This increase is only a small proportion of the in-
atmospheric or high pressure reduces the fatigue limit of crease in fatigue limit that occurs when the effects of
unprotected specimens although the life in the mortal rubber coating and high pressure are combined.
range is the same. The fall and rise of range of shear stress observed during
Crossland showed with tests at a pressure of 20 ton/in2 most of these tests at high pressure is illustrated in Fig. 10
that the deleterious effect of the high-pressure fluid could which shows the variation in semi-range of shear stress for
be partially, perhaps completely, eliminated by protecting five of the batch of solid specimens tested with a super-
the specimen surfaces with a thin rubber coating. These imposed pressure of 25 ton/in2. For comparison Fig. 10
tests have been checked with solid and tubular specimens, also shows the fatigue history of a solid specimen tested
Figs 6 and 8. in air with a high starting semi-range of shear stress. This
If the rubber coating was completely impervious to the comparison and Fig. 11, which shows similar curves for
high-pressure oil the authors’ results show that a super- six tubular specimens, suggests that the initial fall-off is
imposed pressure of 20 ton/in2 raises the fatigue limit of a characteristic of all specimens which are tested with
these specimens from between 2 1-22 ton/in2 to between high ranges of shear stress, irrespective of the ambient
275-29 ton/in2, i.e. above the static shear yield stress pressure. However, an analysis of fatigue histories shows
for this material. that a superimposed high fluid pressure may modify the
T o investigate further this ‘high-pressure’ effect, rate of fall-off of range of shear stress, increase the fatigue
protected solid and tubular torsional-fatigue specimens life in the mortal range, raise the fatigue limit and in
were tested with superimposed fluid pressures of 10, 15 many instances enable a specimen which becomes less
and 25 ton/in2. These results and the results of the tests stiff initially to regain and in some cases exceed its
at 20 ton/in2 and atmospheric pressure have been used original stiffness.
to prepare Fig. 9, a graph of fatigue limit against ambient If the fall and rise of range of shear stress observed
pressure which shows that the fatigue limits of protected during the high-pressure tests involves strain-softening
solid or tubular torsion specimens increase significantly and strain-hardening processes it could be argued that the
with increase in pressure. This comparison of results for main effect of a hydrostatic pressure is to prolong fatigue
solid and tubular specimens also shows that the error in life for a sufficient time for strain-hardening to occur and
the present method of estimating the shear stress range in that the strain-hardened specimen has a much higher
solid specimens which have yielded is not greater than fatigue strength. To investigate this hypothesis static
15 per cent and probably much smaller. This error is torsion tests were carried out on unbroken solid and
small in relation to the overall effect which Fig. 9 illus- tubular fatigue specimens which had shown the charac-
* All tests in air were at atmospheric pressure. teristic fall and rise of range of shear stress in tests at a
J O U R X A L HECHANICAL ENGINEERIXG SCIENCE Vol6 No 3 I964
300 D. J. BURNS AND J. S. C. PARRY

pressure of 20 ton/in2. These specimens had very high bar of steel. This comparison suggests that the strain-
shear yield values, about 31 ton/in2, which suggests that hardening process is responsible for only a small part of
they had strain-hardened. To further investigate this the improvement in fatigue behaviour which occurs when
hypothesis six unbroken solid specimens, which had been hydrostatic pressure is superimposed on a specimen.
tested at a pressure of 20 ton/in2 and had shown the
characteristic fall and rise of range of shear stress, were En 40B steel
retested, unprotected, in reversed torsional fatigue in air A comparison of atmospheric and 20 ton/in2 pressure
at atmospheric pressure. For convenience the results of tests on rubber-covered specimens, Fig. 12, shows that
these tests are compared in Fig. 6 (retests) with atmospheric the superimposed pressure raised the fatigue limit from
and high-pressure tests on virgin specimens from the same 19 ton/in2 to 26.3 ton/in2. This is in good agreement with

10 4 105 10" 107

CYCLES TO FAILURE
o Unprotected longitudinal solid specimens.
Unprotected transverse solid specimens.
+ Rubber-covered longitudinal solid specimens.
x Rubber-covered transverse solid specimens.
A Retests (solid specimens).
Fig. 6. Torsionalfatigue of En 25 steel specimens
J O U R N A L M E C H A N I C A L ENGINEERING SCIENCE V o l 6 No 3 I964
 EFFECT OF LARGE HYDROSTATIC PRESSURES ON THE TORSIONAL FATIGUE STRENGTH OF TWO STEELS 301

corresponding tests on the En 25 steel where the fatigue

limit was raised from 21 to 29 ton/in2.
A comparison of atmospheric and 20 ton/in2 pressure
tests on nitrided specimens, Fig. 13, shows that the super-
imposed pressure raised the endurance limit (10" cycles)
from 25 to 33 ton/in2. The effect of rubber-covering
nitrided specimens for tests with a superimposed pressure
of 20 ton/in2 is also shown in Fig. 13; it can be seen that
the intensifier oil must have a deleterious effect on the
fatigue behaviour of unprotected nitrided specimens. This
W
is somewhat surprising since all nitrided specimens
5
v) appeared to have fractured below the surface; it appears
14 - _I
that the intensifier oil had penetrated the nitrided surface
13 and reduced the beneficial effect of the high compressive
lo5 106 107 108
CYCLES TO FAILURE surface stresses. Several workers (13) have suggested that
A Tested at atmospheric pressure in air. the fatigue behaviour of nitrided specimens is dependent
o Tested at atmospheric pressure in oil. on both the residual compressive stresses and the stronger
x Tested at 20 ton/inz pressure in oil. skin produced by nitriding.
Fig. 7. Torsional fatigue of unprotected solid specimens Although a comparison of the results for rubber-
of En 25 steel covered unnitrided and nitrided specimens suggests that

- loe 10
7
' 1OS
CYCLES TO FAILUR E

0 Unprotected in air at atmospheric pressure. o Unprotected in oil at 20 ton/in2 pressure.

X Protected in air at atmospheric pressure. -t Protected in oil at 20 ton/inz pressure.
Fig. 8. Torsional fatigue of thin-walled tubes of En 25 steel
JOURNAL MECHANICAL ENGINEERING SCIENCE Vo16 No 3 1964
302 D. J. BURNS AND J. S. C . PARRY

a protective rubber coating eliminates the deleterious The fatigue cracks in solid specimens tested with a
effect of the intensifier fluid it must be emphasized that semi-range of shear stress less than 25 ton/in2 in air usually
the semi-ranges of nominal shear stress under discussion started in one of the planes of maximum shear stress of
are well above the limit of proportionality in shear for the gauge section and then propagated in a helical fashion
this steel, which is approximately 20 ton/in2. I n other (Fig. 14a). Occasionally the crack propagated completely
words the actual stress distribution in the specimens, in the transverse plane of maximum shear stress. Solid
particularly when nitrided, cannot be accurately deter- specimens, unprotected or rubber-covered, tested at
mined. high pressure at semi-ranges of stress less than 25 ton/inz
It is interesting to note that all nitrided and/or rubber- usually failed in the same way.
covered En 40B specimens tested at high stresses showed Solid specimens tested at high pressure at ranges of
the characteristic fall and rise of range of shear stress first shear stress greater than 26 ton/in2 usually had several
observed in the tests on En 25 steel. cracks in the longitudinal planes of maximum shear stress
which had joined, by forking at the ends (Fig. 14.b). This
Mode of failure of torsional fatigue specimens and the fact that a solid specimen tested in air at atmo-
Typical failures of En 25 specimens are shown in Fig. 14, spheric pressure with a nominal semi-range of shear stress
Plates 1 and 2. of 30-9 ton/in2 showed the same longitudinal cracking,
Fig. 14c, suggests that the mode of failure of solid speci-
mens is a function of the range of shear stress and not of
the ambient pressure.
Tubular specimens always failed in the transverse
plane of maximum shear stress (Fig. 14d), irrespective of
the range of shear stress and ambient pressure.
Solid specimens of En 40B and En 25 steels tested
under similar conditions failed in the same way.

CORRELATION OF P R E S E N T AND
PREVIOUS WORK
One of the reasons for undertaking the present work was
to investigate why Crossland (3) found that a pressure of
20 ton/in2 increased the fatigue limit in reversed torsion
of En 25 steel solid specimens by 32 per cent whereas
White (7) found that a similar pressure only increased
the fatigue limit in reversed direct stress by 8 per cent.
c) Solid rubber-covered specimens. White suggested that this difference might be the result
Thin-walled tubes (rubber-covered). of the incorrect assumption of a linear stress distribution
Eig. 9. Effect of a superimposed fluid pressure on the in Crossland's solid torsion specimens. The present fatigue
torsional fatigue behaviour of En 25 steel tests in reversed torsion on solid and tubular specimens

100 10' 102 105

7 REPRESENTS COMPLETE FAILURE CYCLES
106 707

Fig. 10. Reversed torsional fatigue at v e ~ yhigh stresses: tests on solid specimens of En 25 steel
JOURNAL MECHANICAL ENGINEERING SCIENCE
I'ol 6 K O 3 1964
 EFFECT O F LARGE HYDROSTATIC PRESSURES ON T H E TORSIONAL FATIGUE STRENGTH O F TWO STEELS Plate I

u Solid specimen tested in air at atmospheric pressure, scmi-range of shear stress < 24 ton/ina.

b Rubber-covered solid specimen tested in oil at 20 ton/in2 pressure, semi-range of shear stress > 26 ton/in2.

Fig. 14. Fatigue failures of En 25 torsion specimens

JOURNAL MECHANICAL ENGINEERING SCIENCE V o l 6 No 3 1964

Plate 2 D. J. BURNS AND J. S. C . PARRY

c Solid specimen tested in air at atmospheric pressure, at semi-range of shear stress of 31 ton/inz.

d Tubular specimen. Type of failure obtained under all conditions.

Fig. 14-corttirtued

JOURNAL MECHANICAL ENGINEERING SCIENCE Vof 6 No 3 1964

 EFFECT O F LARGE HYDROSTATIC PRESSURES ON THE TORSIONAL FATIGUE STRENGTH O F TWO STEELS 303

show that only a small part of the difference can be ex- adjust the applied strain to give a constant stress ampli-
plained in this way. tude. When discussing his high-pressure reversed-stress
T h e difference between Crossland’s results and those experiments White mentions the anomalous behaviour of
of the authors’ can be explained by comparing the experi- one specimen tested with a direct-stress range of f36-5
mental techniques. If Crossland detected a fall-off in the ton/in2 which ‘yielded’ after 142 100 cycles and showed a
range of nominal shear stress, i.e. torque, he stopped the reduction in area of 5 per cent. White also examined
machine and increased the strain to give the original broken specimens tested at similar ranges of stress and
range of nominal shear stress, whereas the present results found that they had clearly yielded before cracking. This
are from approximately constant-strain experiments. and a careful examination of earlier work by O’Connor
It is pertinent to ask if the strain-softening and strain-
hardening which led to these differences were also
occurring undetected during White’s high-pressure direct-
stress experiments since his machine would automatically

104 105 106 107 108 CYCLES TO FAILURE

CYCLES TO FAILURE
x Rubber-covered specimens at 20 ton/in2 pressure.
x Rubber-covered solid specimens at 20 ton/in2 pressure. A Unprotected specimens at 20 ton/in2 pressure.
0 Rubber-covered solid specimens at atmospheric pressure. 0 Unprotected specimens at atmospheric pressure.
Fig. 12. Torsional fatigue of En 4023 steel Fig. 13. Torsional fatigue of nitrided En 40B specimens
JOURNAL MECHANICAL ENGINEERING SCIENCE Vol 6 No 3 1964
304 D. J. BURNS A N D J. S. C. PARRY

and Morrison (14)suggests that it would not be possible sectioning and optical and electron microscopy, to study
using White’s machine and En 25 specimens to increase the effect of superimposing a static tensile end load on
the fatigue limit much above f37 ton/in”, i.e. a 20 per copper specimens subjected to reversed torsional fatigue
cent increase, without the occurrence of yielding failures. stresses. They show that the cyclic straining gives rise to
It is interesting to note that Mackenzie and Benham (15) abnormally distorted slip zones that are potential fissures
have carried out reversed axial strain cycling at atmo- and that a superimposed tension accelerates the cracking
spheric pressure on EN 25 steel specimens, which had of these zones and reduces fatigue life considerably. These
been heat-treated to give 70 ton/in2 ultimate tensile observations suggest that a superimposed hydrostatic
strength, and have observed a sharp fall in cyclical stress stress, i.e. compressive normal stress, might increase a
range during the first 20 cycles at constant strain ranges material’s fatigue strength by stopping or reducing the
between 9.6 per cent and 1.34 per cent. These authors rate of cracking in abnormally distorted slip zones.
prescnt evidence which suggests that the cyclical strain
softening had produced a reduction of up to 8 per cent in
ultimate tensile strength. CONCLUSIONS
A superimposed hydrostatic pressure has a considerable
effect on the torsional fatigue behaviour of En 25 and
GENERAL D I S C U S S I O N O F T H E E F F E C T En 40B steels providing the specimen surfaces are pro-
OF A S U P E R I M P O S E D HYDROSTATIC tected from the high-pressure fluid. Both materials show
P R E S S U R E ON F A T I G U E BEMAVIOUR a 30-40 per cent increase in fatigue limit for an increase
Crossland ( 3 ) has shown with static torsion tests that the in ambient pressure of 20 ton/in2.
superposition of a fluid pressure has very little effect on Experiments on one of the materials, En 25, show:
the yield and shape of the stress-strain curve for En 25 (1) that protecting the surfaces of specimens with a
steel, although it does increase the strain to failure. For rubber-coating has only a small effect on their fatigue
instance a pressure of 20 ton/in2 increases the strain to behaviour at atmospheric pressure; and
failure from 2.0 to 3.5. It is very unlikely that there is any (2) that the error in the present method of estimating
direct connection between the increase in strain to failure the shear stress range in solid fatigue specimens which
and the increase in fatigue strength produced by a yield is not greater than 15 per cent and probably much
superimposed pressure; it will be remembered that smaller.
adjacent longitudinal and transverse En 25 specimens,
tested in air at atmospheric pressure, give essentially the It is concluded, tentatively, that a superimposed hydro-
same fatigue results, Fig. 6, but give very different strains static pressure improves fatigue behaviour by introducing
to failure, 3.0 and 1-4respectively, if tested statically. a compressive stress on the planes of maximum alter-
A more likely explanation for the beneficial effect of a nating shear stress and thereby stops or reduces the
superimposed hydrostatic pressure is that it introduces a rate of cracking in abnormally distorted slip zones.
compressive normal stress on the planes of maximum
alternating shear stress. The importance of this normal ACKNOWLEDGEMENTS
stress in fatigue was first clearly emphasized by Gough
(16)when he carried out combined bending and torsional The authors wish to thank Professor J. L. M. Morrison,
fatigue tests on solid specimens of a ‘65-ton’ nickel- C.B.E., D.Sc. (Member of Council), for his valued advice
chromium-molybdenum steel to show the effect of and encouragement throughout the work, and also
supcrimposed static bending or torsion stresses. He found Professor B. Crossland, D. Sc. (Member), for his
that the superposition of a static torsion stress, which guidance during the first year of the work. Mr W. H.
introduced no normal stress on the planes of maximum Dearden, M.Sc., is thanked for his advice on metallurgical
alternating shear stress but produced maximum shear problems and for preparing the photographs. Much of the
stresses above or just below the yield point, reduced the material used was presented to the University of Bristol
torsional fatigue limit by 8 per cent whereas the super- by the English Steel Corporation Ltd. The whole of the
position of a static bending stress, which introduced work was carried out in the laboratories of the Department
a normal stress of 34-5 ton/in2 on a plane of maximum of Mechanical Engineering in the University of Bristol,
alternating shear stress but only produced a maximum and the assistance given by the workshop staff and especi-
shear stress of about two-thirds the yield stress, reduced ally Mr B. J. Millet is gratefully acknowledged. The authors
the torsional fatigue limit by 23 per cent. Stulen and also thank the Department of Scientific and Industrial
Cummings ( 1 7 ) ~and others (18) (19) have analysed Research for the grants which enabled the investigation to
experimental work on combined-stress fatigue and have be conducted.
suggested that failure is associated with alternating slip APPENDIX
on certain critical shear planes and that the shear stress
REFERENCES
governing failure is a function of the normal stress on the
critical shear planes. (I) WOOD,W. A,, COUSLAND, S. McK. and SARGANT, K. R.
‘Systematic microstructural changes peculiar to fatigue
More recently Wood and Bendler ( 2 0 ) have used taper- deformation’, Acm Metallurgica 1963 11 (No. 7, July).
J 0U R T i H L M E C H A N I C A L E N G 1K E E R I N G S C I E N C E Vol6 No 3 I964
 EFFECT OF LARGE HYDROSTATIC PRESSURES ON T H E TORSIONAL FATIGUE STRENGTH OF TWO STEELS 305

IZ! GBHN, G, R. ‘The mechanism of fatigue’, Materials (13) FRITH, P. H. ‘Effect of nitriding on the fatigue properties
Research and Standards 1963 3 (No. 2, February). of a Cr-Mo steel’,J. Iron Steel Imt. 1948 159 (No. l l ) ,
(3) CROSSLAND, B. ‘Effect of large hydrostatic pressures on the 385.
torsional fatigue strength of an alloy steel’, Proc. Int. (14) O’CONNOR,H.C. and MORRISON, J. L. M. ‘The effect of
COT$.Fatigue of Metals 1956, 138 (Instn Mech. Engrs, mean stress on the push-pull fatigue properties of an
London). alloy steel’, Proc. Int. Conf. Fatigue of Metals 1956, 102
4 ) FORREST, P. G. and TAPSELL, J. H. ‘Some experiments on (Instn Mech. Engrs, London).
the alternating stress fatigue of a mild steel and an (15)MACKENZIE, C. T. and BENHAM,P. P. ‘Cyclical strain-
aluminium alloy at clevated temperatures’, Proc. Instn softening of a heat-treated steel’, Engineer, Lond. 1962
mech. Engrs, Lond. 1954 168, 763. (28th December).
;5’ CHODOROWSKI, W. T. ‘Fatigue strength in shear of an alloy (16) GOUGH,H. J. ‘Engineering steels under combined cyclic
steel’, Proc. Int. Conf. Fatigue of Metals 1956, 122 (Instn and static stresses’, Proc. Instn mech. Engrs, Lond. 1949
Mech. Engrs, London). 160, 417.
(17) STULEN,F.B.and CUMMINGS, H. N. ‘A failure criterion for
(6; WHITE,D. J., CJOSSLAND,B. and MORRISON, J. L. M. niulti-axial fatigue stresses’, Proc. Amer. SOC. Test.
‘Effect of hydrostatic pressure on the direct-stress Muter. 1954, 822.
fatigue strength of an alloy steel’, J . mech. Engng Sci. (18)FINDLEY, W.N. ‘Theories relating to fatigue of materials
1959 1 (No. l), 39. under combination of stress’, Proc. Int. Union theor.
(7) WHITE,D. J. Ph.D. thesis, University of Bristol, 1958. appl. Mech. Colloquium on Fatigue, Stockholm, May 1955,
(8) CAZAED,R. Fatigue of metals 1953, 184 (Chapman and 35 (Springer-Verlag, Berlin).
Hall, London). (19)FORSYTH, I?. J. E. ‘A two-stage process of fatigue crack
(9)CROSSLAND, B. ‘The effect of fluid pressure on the shear growth‘, Proc. Crack Propagation Symp. , Cranfield,
properties of metals’, Proc. Instn mech. Engrs, Lond. 1954 1961, 77.
168, 935. (20) WOOD,W. A. and BENDLER, H. M. ‘Effect of superimposed
(10) CROSSLAND, B. and HILL,R. ‘On the plastic behaviour of static tension on the fatigue process in copper subjected
thick tubes under combined torsion and internal pres- to alternating torsion’, Trans. Met. Soc. Amer. Inst. min.
sure’, J . Mech. Phys. Solids 1953 2, 27. Engrs 1962 224 (No. 1, February), 18.
IrI? CROSSLAND, B. and BONES,J. A. ‘The ultimate strength of (21) MORRISON, J. L. M. ‘A three-ton testing machine’, Engineer,
thick-walled cylinders subjected to internal pressure’, Lond. 1934 157, 626.
Engineering, Lond. 1955 (21st January, 28th January). (22) MORRISON, J. L. M. ‘Yield of mild steel, with particular
[IZ) NADAI,A. Theory of flow and fracture of solids Vol. 1, 1950 reference to the effect of size of specimen’, Proc. Znstn
ihlcGraw-Hill, London and New York). mech. Engrs, Lond. 1940 142, 193.

J 0IJR N A I. M E CII AN I C A L E N G I N E E R I N G SCIENCE VoI 6 No 3 1964

306

COMMUNICATIONS
The following Comniunications relate t o the paper ‘Tool the magnitude of the normal stress or is it connected with
forces and tool-chip friction in orthogonal machining’, by the relative magnitude of plastic and elastic stresses in the
P. W. WALLACE and G. BOOTHROYD, which was chip ?
published iii 1964, uol. 6, No. 1, p. 74. Finally, I should like to ask the authors if orthogonal
cutting, especially in the absence of lubricants, produces
Dr D. Tabor, F.R.S. (Cambridge)-The authors have the geometrically perfect chip shown in Figs 2 and 3, and
combined a sensible and modest analysis with direct ex- the inset in Fig. 14. Some time ago the late Professor
perimental observation and their conclusions constitute, I Ehrenreich at the Haifa Institute of Technology showed
believe, a very real contribution to our understanding of me some pictures of chips obtained in orthogonal cutting
the forces involved in orthogonal machining. which suggested there was appreciable spreading of the
I should, however, like to raise a few points not of chip in a direction at right angles to the flow direction of
criticism but of enquiry. I n discussing the mechanism of the chip. The chip was broader than the cut. This explains
friction over the stick region the authors construct a why a fine machine-finish is so hard to obtain since the
model (Fig, 16a) in which a slider is constrained so that chip itself may damage the surface; and why the parting
‘no combined-stress yield criterion applies to the slider process is so noisy an affair. But it is particularly important
material’. This may be a perfectly valid approximation in suggesting that, in the chip material near the tool edge,
to the conditions in the chip but it is a crucial one. It there is plastic flow in three dimensions rather than in two.
seems to me that as far as the chip is concerned this would Do the authors think this might have any appreciable
be true only if shearing occurs over an extremely thin effect on their analysis ?
surface layer. I t may, of course, be that even if combined
stresses are to be taken into account the yield criterion D r H. Takeyama (Tokyo, Japan)-This paper con-
may well be of the form: tributes much to research on metal cutting, giving valuable
crm2+arm2= k2 and accurate information on the stresses on the tool face.
so that if CI is a large number (Tabor*) the yield is essenti- The assumption that the force P is independent of
ally determined by the tangential stress T ~ .I would ap- depth of cut is still very important. According to my own
preciate comments by the authors on this. test result, the singular force acting at the cutting edge is
The second point concerns the striking photograph a function of depth of cut. The fact that the lines in
shown in Fig. 14. The distinction between the sticking Fig. 10 approach a common intercept independent of l/t,
and sliding regions is clearly shown. However, I am a little does not give positive grounds for the assumption. More-
puzzled as to how further flow occurs. If one were to view over, it seems to be daring to extend the lines straight as
the same surface a fraction of a second later one could depth of cut decreases to zero in Fig. 15, because there is
easily envisage how the sticking region could move into usually a marked deviation from linearity between force
the sliding region; but how would fresh metal move into and depth of cut in this region.
the place of the sticking region without blurring all the
surface features shown? Do the authors imply that the
sticking position remains for an appreciable time as a AUTHORS’ REPLY
fixed piece of metal whilst shearing occurs a short distance Mr P. W. Wallace and D r G. Boothroyd-We should
behind it within the material of the chip itself? If so, this like to thank Dr Tabor and Dr Takeyama for their in-
presumably would be associated with some built-up edge teresting and very helpful comments.
mechanisms. 1 would appreciate enlightment on this. I n We agree with Dr Tabor that the assumption that no
addition would the authors express an opinion as to why combined stress criterion applies along the sticking region
sticking occurs over one region and sliding on another ? is crucial. We had attempted an analysis based on the
Do they consider that this is determined primarily by equation given by Dr Tabor and found it gave a steeply
decreasing frictional stress with increase in normal stress
* TABOR, D. ‘Junction growth in metallic friction: the role of
within the stress range shown in Fig. 16. The constant a
combined strcsses and surface contamination’, Proc. roy. SOC.
1959 (Series A) 251, 378. was assumed to be equal to 3, the theoretical value for
J O U R X A L M E C H A N I C A L E N G I N E E R I N G SCIENCE Vo16 No 3 1964
 COMMUNICATIONS 307

a two-dimensional stress condition in the contacting our model, we would suggest that the surface layer may
asperities. However, as Dr Tabor has shown*, the stress be subjected to a hydrostatic stress system whose com-
condition in the asperities may well approximate to the ponents are approximately equal to the normal stress at
hree-dimensional axisymmetric case for which the value the interface, the body of the chip supporting the hydro-
of cc may be 9 or greater. A calculation for comparison with static components in the plane of the chip. Under these
the experimental curve in Fig. 16 shows that w would have conditions frictional deformation would occur under the
to be at least 20 to give an acceptable rate of decrease of action of the frictional stress only.
frictional stress r, with increasing normal stress u, over Regarding the stability and the reasons for the occur-
the range where T , ? ~approaches its maximum value. An rence of the sticking region, we do suggest that the thin
explanation based on Dr Tabor’s equation thus seems layer of chip material adheres to the tool surface for an
possible if a value of u greater than 20 is acceptable. T o appreciable time. The occurrence of such a layer could
determine experimentally which model gives the better possibly be viewed in the following way. In the sliding
representation, it would seem necessary to determine the region, near where the chip and tool separate, the frictional
-rm-ujl, relationship at higher values of urn than those strength of the interface is less than that of the chip
shown in Fig. 16 by reducing the chip-tool contact material behind it. Further down the tool face where the
further. Unfortunately it was not possible to do this since normal stress has increased, the frictional strength of the
further reductions in the contact length caused the chip interface on account of the more intimate contact becomes
to bend over and contact the relieved surface of the rake greater than the strength of the material behind it and
face. seizure occurs. Shearing is thus caused some distance
In proposing our model we were influenced by two from the interface in a layer which may be likened to a
principal considerations : first the deformation in the chip narrow boundary layer. We would suggest that the
was mostly confined to a relatively thin layer adjacent to occurrence of this layer is primarily determined by the
the tool face, Fig. C1; a similar result was found by magnitude of the normal stress. We would agree with
Zorev for rhe cutting of steel at I85 ft/min at a rake angle Dr Tabor that the layer is really a very thin built-up edge;
of 10”. Zorcvj- found that a layer only 0.001 in (less than the literature suggests that such a layer is present under
1 per cent of the chip width) from the tool face had most cutting conditions except possibly for very free-
already reached 75 per cent of the mean chip velocity. cutting materials which contain boundary-lubricating
Secondly the frictional stress showed no sign of decreasing agents. This layer, however, differs from the substantial
as the normal stress increased (Fig. 16) as a combined built-up edge described by Heginbotham and Gogiat in
stress criterion would predict. As a theoretical basis to that it does not significantly alter the effective tool geometry.
In our experiments chip side-spread was small, the
* TABOR, n. (same reference as quoted in Tabor’s communication). chip width being about 3 per cent greater than that of the
t ZOREV,N. N. Contribution to discussion on ‘Temperatures in
orthogonal machining’ by ti. Roothroyd, Proc. Instn mech. $ HEGINBOTHAM, W. B. and GOGIA, S . L. ‘Metal cutting and the
Engrs, Lond. 1963 177, 789. built-up nose’, Proc. Instn mech Eltgrs, Lond. 1961 175, 892.

Rake angle 30”,cutting speed 160 ftirnin, depth of cut 0-017 in.
Fig. C1. Chip side-surface showing frictional deformation
J O U R X AL M E C H A P : I C A L E S (iI N E l i R I N G S C I E K C E Vol 6 h’o 3 1964
308 COMMUNICATIONS

workpiece. Consequently, the cutting action was essen- The following Communication relates to the paper ‘Lateral
tially two-dimensional and the geometrically perfect chips vibration of shafts having radial symmetry and appreciable
shown in Figs 2 and 3 were a valid idealization. There is gyro-action’, by H. F. B L A C K , which was published in
no doubt, however, that at smaller rake angles and lower 1964, vol. 6 , No. I , p. 53.
speeds or with different materials appreciable side-spread
can occur. It appears that under these conditions the side- Mr A. R. Finkelstein (Phoenix, Arizona)-The author
spread is greatest in a thin layer adjacent to the tool face states ‘it is well worth considering an iterative procedure
which seems to have been ‘extruded’ outwards by the for approximating to the natural frequencies as an alterna-
frictional stress system. We feel that further work would tive to setting up and solving the frequency equation’.
be necessary to determine whether the proposed model In a paper* presented to the ASME but not yet
gives an adequate description of chip-tool frictional published, a simple numerical procedure is developed for
behaviour in the presence of such appreciable side-spread. determining the whirl velocity as a function of the rota-
With reference to Dr Takeyama’s comment, we agree tional velocity for flexible multimass rotor systems that
that the assumption regarding the nose force P is import- are difficult to handle by completely analytical procedures.
ant. If the problem is looked at another way we can, The numerical procedure follows the Myklestad methodt
without making any assumptions, say that the experimental and introduces an effective mass moment of inertia for
results show that the tool force system may be divided each disc to account for various ratios of whirl to shaft
into two parts : (1) a constant force P which is independent speed, and the linear flexibility of supports is included.
of depth of cut and the geometry of deformation and (2) a This numerical procedure offers very little complication,
force Q which is dependent on both these factors. Our regardless of the number of degrees of freedom or the
assumption then is that P acts on the tool nose region or, number of supports, and introduces no approximations
more precisely, on the small area of contact between tool to the ‘exact’ solution.
edge and flank and the freshly cut work surface. It is also noted that in 1954 Yamamotot used Eulerian
Dr Takeyama’s view seems to be that, in addition to this angles and Lagrange’s equations in obtaining the same
constant force, a force component dependent on depth of results as the author, and further conducted many experi-
cut acts at the tool edge. We do not think that there is any ments physically demonstrating the non-synchronous
strong evidence for this. If the yield pressure of the work rotor motions and how they may be excited. Yamamoto
material is assumed to be 50 000 lb/in2, the applied normal further analysed systems with the particular physical
force on the surface of the tool edge of radius 0-0002 in properties necessary to generate such motions.
will be only 3.8 lb. The singular force predicted by
Takeyama and Usui’s equation is given by the intercept
at zero contact length in Fig. 6 ; for the maximum depth AUTHOR’S R E P L Y
of cut used this intercept is about 230 lb, giving a stress of
about 3 000 000 lb/in2 on the tool edge. It does not seem Mr H. F. Black-The application of the Myklestad
possible that the tool material could support a stress of method in this field is certainly a welcome and notable
this order. Our opinion, which seems to agree with that advance in computational techniques. In systems with a
expressed by Takeyama and Usui (7), is that the singular large number of rotors the elastic coefficients of the type
force obtained by extrapolation to zero contact length employed in the Myklestad method will be simpler and
indicates the existence of high stresses along the portion possibly fewer in number than the Maxwell coefficients
of the rake face adjacent to the tool edge. If this is so, the employed in the adaptations of Stodola’s method suggested
singuiar force found by Takeyama and Usui’s method in the author’s paper. One advantage that may be claimed
comes within the definition of the rake face force Q and for the Stodola-based method is the fact that it does not
our conclusions on this point would thus seem to be in entail a search for natural frequencies : the procedure
agreement. involved, like that of Myklestad’s method, is repetitive
Dr Takeyama is most probably correct that the rela- but more truly iterative in the sense that it is self-con-
tionship between force and depth of cut deviates from vergent and does not depend on the unguided re-calcula-
linearity at very small depths of cut. However, the plots of tion of a function or functions for a large number of trial
force versus depth of cut at constant Zitl in Figs 9 and 10 frequencies. Whilst criticism can fairly be levelled at the
were drawn to study the force behaviour at the depths of Stodola procedure on the grounds of complication in
cut investigated and not at depths of cut which physically large systems, there is little cause for boggling at the intro-
approached zero. The experimental points showed a duction of initial approximations to ‘exact’ solutions (or
linear relationship between force and depth of cut and the * FINKELSTEIN,
A. R. ‘Myklestad’s method for predicting whirl
linear extrapolation to zero depth of cut simply enabled velocity as a function of rotational velocity for flexible multimass
the experimental result for the depths of cut investigated rotor systems’, presented in 1964 to Applied Mechanics Division
to be expressed mathematically as a constant force plus a of Amer. Soc. mech. Engrs, unpublished.
force proportional to depth of cut. The way in which the t MYKLESTAD, N. 0. ‘Numerical analysis of forced vibrations of
beams’, paper No. 52-FR, 1952, Amer. Soc. mech. Engrs.
forces vary at small depths of cut was therefore not an $ YAMAMOTO, T. ‘On the critical speeds of a shaft’, Mem. Fuc.
important consideration in our work. Engng, Arugoya Unit!., Japan 1954 6 (November, No. 2).
JOVR\-AI MECHANICAL ENGINEERING SCIENCE Vol 6 N o 3 I964
 COMMUNICATIONS 309

rather, shapes). The rate of convergence of subsequent If this similarity between the axial velocities of the water
iterations is very rapid and in smaller systems this may and vortices exists at high Reynolds numbers, their pre-
compensate for a fairly wide range of unguided ‘shots’ sence would be difficult to detect by the flow visualization
by thc Myklestad procedure. methods described.
The author is grateful for the mention of Yamamoto’s It may be of some value to describc my rotary heat-
work in 1954 which had been unknown to him. One of the transfer apparatus,
main purposes of the present discussion was not, however, My first apparatus consisted essentially of two vertical
to establish the equations of motion: these have been concentric annuli. The inner annulus was formed by a
treated by a number of authors as recorded in the reference smooth stainless-steel rotor of 3$ in diameter, 13+ in long,
list; but to show the extent to which these equations and a steel tube of 33 in inside diameter thus giving an
differed from the canonical small vibrations equations, annular gap of in. The outer annular width was 3in.
the adaptations needed to apply principal co-ordinate The cold-side fluid was water, which could be passed
analysis and consequently the limitations of such analysis upwards through the inner annulus at rates up to 20 000
in this context. lb/h. The hot-side fluid was steam, pressures up to
100 Ib/in2 gauge being possible. The rotor speed range
was 3750-6000 revlmin. The temperature of the water
The following Communications relate to the paper ‘Heat at inlet and outlet was measured by mercury-in-glass
transfer from a rotating tube with controlled fluid flow’, by thermometers, the condition of the stcam at inlet being
R . F. PATTENDEN, which was published in 1964, vol. 6, determined by a throttling calorimeter.
No. 2, p. 144. For a given water-side Reynolds number (based on
axial velocity), the ratio of the overall heat-transfer coeffi-
Mr J. E. R. Coney (Graduate)-I have read this paper cient with the rotor turning to that with the rotor stationary
with great interest since I am working in the same field. was up to 1.7. An attempt was made to estimate the heat-
I found the author’s results presented in Fig. 7, re- transfer coefficient at the inner annulus wall-water inter-
lating to film coefficients with no vanes in either annulus face but without success.
interesting and their form as might havc been expected. The success of these preliminary investigations has led
Taylor* treated the case of the outer cylinder moving to the design of an improved apparatus, using water and
with the inner cylinder at rest and found the motion of the steam as before, but with 39 thermocouples and six
fluid continued between them completely stable. pressure tappings at the inner annulus wall, 39 thermo-
Hence, one would expect the heat-transfer coefficient couples at midstream in the inner annulus, and a rotor
on the hot side to be much lower and its rate of increase speed range from 100 to 6000 rev/min. This apparatus
with rotational speed much less than that on the cold side. will be ready in the near future.
It would have been of interest had the author compared
the hot-side heat-transfer coefficient with a calculated Mr A. L. Greenshields (Manchester)-The final
value derived from his equation (1). conclusion that the heat transfer rate is proportional
In the conclusions, the author states that Taylor directly (or nearly so) to the relative velocity has been
vortices could not be confirmed and that the high heat- postulated previously in a paper by Seban and Johnson /I
transfer coefficients may not be due to their presence. in which they propose the following equation for heat
This is a point which has concerned me for some time and transfer from a rotating tube:
I should like to offer the following observations.
Shi-I Pait found that, for air having zero axial velocity _ _-_ -
TI’
‘f
0
(pr)0.356
A L V
in an annulus formed by a stationary outer cylinder and a
rotating inner cylinder, Taylor vortices cxisted at several where Cf is the coefficient of skin friction.
hundred times the critical Taylor number. Kaye and The obtained results which were about 20 per cent
Elgar$ have confirmed this. greater for the individual heat-transfcr coefficient than
Considering the case of axial flow of water in such an those predicted by the above equation.
annulus, Donnelly and Fultzs have shown that for Rey- Another published paper which may be of interest in
nolds numbers (based on the axial velocity) up to 5-7, the this connection is that of Tachibana, Fukui and Mit-
rate of axial motion of the vortices is up to 1.3 times greater SumuraJI which deals more with Taylor numbers and they
than that of the axial velocity of the water. obtained some correlation with the equation:
* TAYLOR, G. I. ‘Stability of a viscous liquid contained between
Nu = 0.21(Ta2Pr)+
two rotating cylinders’, Phil. Trans. 1923 223 (series A), 289. for Taylor numbers above 41, the heat-transfer coefficient
t Par, SHI-I, ‘Turbulent flow between rotating cylinders’, being the overall value across an annulus.
Tech. Nares nut. adv. Comm. Aero., Wash. No. 892, 1943.
$ KAYE,J. and ELGAR, E. C. ‘Modes of adiabatic and diabatic
fluid flow in an annulus with an inner rotating cylinder’, Trans. 1 SEBAN,R. A. and JOHNSON, H. A. National Aeronautics and
Amer. Sac. mech. Engrs 1958 80, 753. Space Administration, Washington, paper No. 4-22-59W.
5 DONNELLY, R. J. and FULTZ, D. ‘Experiments on the stability J1 TACHIBANA, F., FUKUI,S. and MITSUMURA, H. ‘Heat transfer in
of spiral flow between rotating cylinders’, Proc. N a t . Acad. Sd. an annulus with an inner rotating cylinder’, yap. Sac. mech.
1960 46, 1150. Engrs Bull. 1960 3 (No. 9).
J 0U R N A L M E C H A N I C A L EN ti IN E ER I N G SCIENCE Vol6 No 3 1964
3 10 COMMUNICATIONS

Further reference to heat transfer from rotating bodies but, with vanes in the inner annulus, the thermal resist-
in general can be found in a work by Dorfmann*. Chapter ance of the steel tube limited the heat transfer, and we had
7 is of special interest in this connection and the results to change to copper. It is possible that Mr Coney’s
predicted by equation 797 are in approximate agreement stainless tube will have insufficiently high conductance to
with those of Seban and Johnson. be satisfactory .
It would be interesting if Mr Pattenden could comment It would seem that he will not be able to measure the
on the correlation of his results with reference to the temperature of the rotating tube. Without this, the
equations and hypotheses of the above papers. assessment of the film coefficient hardly seems possible.
Thermocouples in the fluid stream, while giving the radial
temperature gradient if they can be relied on, may SO
AUTHOR’S R E P L Y
modify the flow pattern that this temperature, read from
M r R. F. Pattenden-Mr Coney’s remarks are interest- them, may be misleading.
ing. In reply, the calculated values for the heat transfer Mr Greenshields’s contribution is much appreciated.
coefficient in the inner annulus, when no vanes are fitted, The paper by Seban and Johnson considered rather
agree well with those shown on the curve in Fig. 7. I n different conditions since their cylinder rotated in a large
making these calculations the effective velocity is taken (as tank of liquid, and so could not be looked on as forming
in the previous paper) as u = \/(4ua2+w2r2) where u, is the core of an annulus. Thus it seems likely that the flow
the axial velocity, and WT the velocity at the tube surface, patterns would be different. However, my results do
i.e. the circumferential velocity. appear to be in line with these, in that the heat-transfer
Taking constants at their mean values for the tests, this coefficient is nearly proportional to the Reynolds number
gives the film coefficients as 390 c.h.u./h ft2 degC at stand- and to the cube root of the Prandtl number.
still and 510 c.h.u./h ft2 degC at 3000 rev/min of the tube, This approximation applies only to the outer annulus
the ratio being 1.3.This agrees fairly well with Mr Coney’s when no vanes are fitted at all. If vanes are fitted in the
figures assuming that his ratio is between 6000 rev/min inner annulus the heat-transfer coefficient appears to vary
and zero. It is, therefore, reasonable to suppose that the as Re0’89PrS, whereas if no vanes are fitted the index for
stabilizing effect of rotation suppresses turbulence which Re is about 0.75.
might exceed that due to relative motion, and that if The skin friction could not be determined with accuracy,
vortices are formed and persist, their effect on the film since pressure drops were very small, and torque measure-
coefficient is not great. ments, which were made on the previous rig, showed that
In the outer annulus no such stabilization occurs, and by far the most important source of friction was in the
the effect of rotation, whether Taylor vortices are formed seals. It is not possible, therefore, to correlate the results
and persist or not, is greatly to enhance the value of the with those of Seban and Johnson numerically, but it is
film coefficient, Where vanes are fitted to the inner stator unlikely that the coefficients would be the same, since the
the same effect is found, and on o w apparatus the ratio configurations were different.
of heat-transfer coefficients rose from 1.3 to over 10. It found that the Nusselt number was proportional to
The description of Mr Coney’s apparatus is not quite the fourth root of Tu2taking the form of Taylor number
clear, for the inner annulus is given as being formed given by Becker and Kaye, over a fair range. This agrees
between a rotating tube of 3$ in diameter and a fixed tube with Tachibana, Fukui and Mitsumura assuming a
of 39 in diameter. Thus the annulus is outside the rotating constant for the fourth root of the Prandtl number. This
tube, and is what we have referred to as the outer annulus. is a fair approximatron, for although the Prandtl number is
Perhaps the dimensions should be interchanged. not constant, since the mean temperature of the fluids
In our first model we, also, used a steel rotating tube, changes with speed, this does not make a very great
* DORFMANN, L. A. Hydrod-vnamic resistance and the heat loss of difference to the fourth root over the temperature range
rorati?g solids 1963 (Oliver and Boyd;. used.

JOCRYA!. MECHANICAL EXGINEERING SCIENCE Vol6 No 3 1964