METAL DIAPHRAGM COUPLING PERFORMANCE

by
M. M. Calistrat
Manager, Power Transmission Development Section

Research & Development Department

Metal Products Division

Koppers Company, Inc.

Baltimore, Maryland

INTRODUCTION
Michael M. Calistrat, a native of
Rumania, graduated from the University The Metal Diaphragm Coupling is relatively new in
of Bucharest with an M. S. degree in turbo-machinery applications. Although the first recorded use
Mechanical Engineering in 1951. of such a coupling dates back to 1922*, the contoured dia­
He worked for many years designing phragm did not become widely used until the late 1950's.
components for oil drilling rigs, includ­ Diaphragm couplings accommodate the system misalign­
ing rotary tables, hydrostatic transmis­ ment through flexing. Fatigue resistance is the main pertorm­
sions and sludge pumps. ance criterion. The life expectancy of a diaphragm coupling
1n 1964 Mr. Calistrat immigrated to that operates within its design limit is theoretically infinite.
the USA and worked for Crane Carrier Considering that it is a relatively new product, and considering
Co. in Tulsa, Oklahoma, designing truck that machinery users are not accustomed to having products
frames, some for loads exceeding 100 tons. Late in 1966, he that are designed to last forever, Koppers conducted com­
moved to Baltimore, Maryland as a senior project engineer for prehensive laboratory tests at conditions which were much
the Metal Products Division of Koppers Company, Inc., in the more demanding than normal field service. The results of these
Research and Development Department, and after one year, tests are very encouraging.
was promoted to his present position as manager of the Power
Transmission Development Section. Since then, he has worked Figure No. 1 shows a section through a diaphragm coupl­
mainly on flexible couplings, helping in improving the existing ing. The coupling has only five parts: two rigid hubs, one spool
product line and in adding new products. He holds twelve U. S. piece, and two alignment rings. These five parts are solidly
patents. bolted together and misalignment is accommodated through
Since 1975, Mr. Calistrat has been chairman of the Shaft flexing of the two disks of the spool. The spool piece is made up
Couplings and Clutches Subcommittee of ASME. He has of three separate parts: two disks and a spacer tube. These
presented various. papers at ASME and ASLE meetings and parts are welded together through the electron beam (EB)
also authored articles in Hydrocarbon Processing and Mechan­ process. EB welding, a relatively new industrial process, is one
ical Engineering. of the keys to the successful manufacture of diaphragm coupl­
ings.
The heart of these couplings is the flexing disk; it is man­
ufactured from vacuum degassed alloy steel, forged with radial

ABSTRACT *JUNGSTROM Condensing Steam Turbine Driven Locomotive,

The pertormance of a Metal Diaphragm Coupling is to a Sweden.
large extent determined by its endurance limit. The endurance
limit can be estimated by calculations but must be confirmed
through testing. The author cites data obtained from a com­
prehensive test program conducted in the Research and De­
velopment facilities of Koppers Co�pany, Inc. in connection
with the development of an improved version of the Single
Diaphragm Coupling.
Failur e of the coupling's disk, or diaphragm, occurs
through fatigue when it is stressed above its endurance limit;
the mode of failure depends on the angular misalignment and
axial displacement. Full scale tests conducted in the laboratory
under a variety of extreme conditions confirmed the operating
limits and provided a pictorial reference which may be used as
a tool to identifY probable causes of disk failures.
A special test rig was used to obtain S-N (stress versus
number of cycles) curves for diaphragms made of various mate­
rials, for various treatments, surtace finishes, etc. The large
number of tests performed gave a good insight into the influ­
ence of various factors on the endurance limit and on the con­
sistency of the endurance limit from disk to disk. Figure 1. Metal Diaphragm Coupling (one end shown).

117
118 PRO CEE DINGS OF THE FIFTH TURBOMA CHINERY SYMPOSIUM

grain orientation and has a contoured profile machined on high

precision equipment.

WHY CONTOURED PROFILE?

In order to obtain the minimum afforce for a given deflec­
tion (maximum flexibility) a beam must be uniformly stressed
throughout. Under normal operating conditions a diaphragm of
a coupling is subjected to uniform and cyclic stresses. The
uniform stresses are generated by torque, centrifugal forces
and axial deflection. The cyclic stresses are induced by the
angular misalignment seen by the disk. A method for calculat­
ing these stresses was published for the first time in 19 48 by
Wolff (1), and was updated by the manufacturers of diaphragm
couplings. The following simple analysis of the stresses in a
disk is merely intended to demonstrate the need for a con­
toured profile.
a. Torque transmission (See Figure No. 2) Figure 4. Stress Distribution Under Axial Deflection.
The tangential force at r; is T/r; (1)
Where T is the torque tance between rim and hub increases) and bending. The stres­
The section area at r; is 21Tr;t; (2) ses imposed by axial deflection (See Figure No. 4) are larger at
the hub than at the rim. As will be shown later, these large
The shear stress is T = T /21Tr;2 t; (3) stresses significantly influence the failure mode of the disks.
For constant shear stress
d. Angular misalignment
r? X t; = Constant or Figure No. 5, reproduced from Wolffs paper, illustrates
how the stresses generated by angular misalignment are influ­
t; = K1 /r? (4) enced by the ratio of outer diameter/inner diameter; the
curves are drawn for a disk with uniform thickness and for a
b. Centrifugal forces
profiled disk. As the ratio b/a ( See Figure No. 4) for most of the
Due to centrifugal action, a section through the disk is also diaphragm couplings is approximately .45, it can be seen that
subjected to a radial force C. Because the disk is solidly bolted the advantage of profiled disks ,for this type of stress, is small.
to a rigid hub, the calculations of the centrifugal force acting on
the disk are rather involved. Standard methods to calculate In conclusion, in order to have a uniform stress when all
centrifugal stresses in a rotating disk (2) show that both the the various forces acting on the disk are at their maximum
tangential and radial stresses increase rapidly with a decrease rated values the disk must have a profiled contour. If this stress
in the radius. For a uniformly stressed disk we can again write exceeds the endurance limit of the material the failure can
occur anywhere between the hub and rim area. This statement
t'; = K2/r;2 (5) is no longer true if only one or more of the forces are at their
maximum values, as will be shown later by the test results.
c. Axial deflection (See Figure No.3) Failures of early disks occurred mainly in the thinnest section.
This type of disk deformation is called "umbrella" by Even Wolff states in his paper: "the disk .. . should be pro­
Wolff, but is more appropriately known as "drum head" be­ gressively thickened over the outermost 15 per cent of the
cause the rim diameter remains constant. It can be seen that annulus to give a rim thickness some 25 per cent greater than
the disk deformation has two effects: radial stretch (the dis- the theoretical."

Figure 2. Section Through a Profiled Disk. Figure 3. Axial Deflection in a Disk.

 METAL DIAPHRAGM COUPLING PERFORMANCE 119

mined for various materials, heat treatments, manufacturing

methods, etc.; second, full scale tests of entire couplings were
conducted at gradually increasing loading conditions until fail­
ure occurred.
a. S-N tests
Establishing S-N curves with entire couplings is impracti­
cal for several reasons: considering the large number of test
points required to determine an S-N curve the cost of the test
parts would be prohibitive; the energy transmitted through the
couplings would be wasted; and the overall variations behveen
the many test parts used for one curve would cause a scatter of
the results. It was therefore decided that any single S-N curve
would be established using test samples machined from one
disk. The disk was cut radially into 10° segments, and the sides
were ground in a special jig to insure uniformity from sample to
sample.* A special test rig was built in order to simulate, as
closely as possible, the flexure of a disk rotating under condi­
tions of angular misalignment. The test rig is illustrated in
Figure No. 6.
Although this test rig did not reproduce any of the con­
stant stresses in the disk, the results of the tests helped us in
determining the endurance limit, and enabled us to select the
best variables for the coupling.
o+---+---+
0 0·1 The following graphs illustrate how the endurance limit is
b/a influenced by some of the factors studied.
Figure 5. Stress Under Angular Misalignment. Figure No. 7 represents the results obtained with a disk
having all the parameters chosen for production couplings:
\Vhat was not recognized originally was that progressive material was modified 4335, vacuum degassed, radially forged;
thickening was not necessary for reducing stresses, rather it is the disk's surface finish was very fine, and proper heat treat­
required for reduction of stress concentra tions. Both at the hub ment and shot-peening were applied. Not only is the endur­
and at the rim, fillet radii are used to connect the contoured ance limit very high (75, 000 psi) but the scatter of the test
profile to the thick annuli required for making the connections results is minimal. Figure Nos. 8 and 9 represent the results
to the other components of the coupling. At the hub, the con­ of tests with similar disks having surface finishes 4 and 10 times
toured profile and the fillet radius have the same sign, i. e., coarser than the one selected for production. It can be seen
both tend to thicken the disk with a decrease in r; At the rim, that, as the surface quality worsens, the scatter of the test
however, the contoured profile and the fillet radius have oppo­ results becomes larger, and beyond a certain point an endur­
site signs, and this fact causes an unacceptable stress concen­ ance limit cannot be established.
tration.
Although shot-peening generally improves the endurance
Methods used to obtain the progressive thickening vary limit of a component subjected to cyclic stresses, it was found
somewhat from manufacturer to manua f cturer. Optimum pro­ that, for parts as thin as these profiled disks, shot-peening can
file is obtained through experimentation rather than through sometimes be detrimental. Figure No. lO illustrates this find­
theoretical analysis. ing. Poor results can be caused either by shot that is too large
or by impact velocity that is too high.
EXPERIMENTAL PROGRAM
The experimental program we conducted had t w o phases: *Note that slight variations in the width of the test samples do not
first, S-N (stress versus number of cycles) curves were deter- affect the stresses.

Figure 6. Fatigue Test Rig. Figure 7. S-N Curve for Production Disk.
120 PROCEEDINGS OF THE FIFTH TURBOMACHINERY SYMPOSIUM

Figure 8. S-N Curve for Disk with Medium Grade Swjace Figure 10. S-N Curve for Disk with Improper Shot-Peening.
Finish.

-
; ...
Ct.
..
'Ill
""
..
"'
::#
I> 15
Jill;
... 0

60
-

Figure 9. S-N Curve for Disk with Coarse Surface Finish. Figure 11. S-N Curve for Disk Manufactured Through ECM.

Rather surprising results were obtained with disks man­ In order to stop the test rig shortly after the failures oc­
ufactured by "electro-chemical-milling" (ECM) process and curred, the rig was equipped with sensors that measured tor­
with those which were "electro polished. " Figure No. 11 rep­ que, speed, and vibrations in the gear boxes, and with prox­
resents the S-N graph of such a disk. Although the ECM pro­ imity gauges placed in the vicinity of the coupling tubes. Other
cess yields parts which are geometrically perfect, it was found sensors measured temperatures at various points, oil pressur­
under electron scanning microscope observation that small es, etc. In case of a malfunction, the rig's annunciator automat­
cavities sometimes develop on the surface of the metal . . . ically stopped the rig and indicated which of the sensors
cavities which apparently generate stress concentration. triggered the test interruption. This instrumentation was suc­
The results shown here are only a small part of the data cessful in stopping the rig before major damage was done to the
obtained in our laboratory fatigue tests. They help us to under­ failed disks, and thus enabled us to analyze the failures and
stand some of the problems which prevented successful prod­ learn much about the failure modes.
uction of this type of coupling until sometime in the 1950's. ·when a coupling is used below its ratings, one expects the
b. Full scale testing coupling: to transmit the required torque, to withstand the
operating speed, and to operate for a reasonable length of time
Through careful stress calculation, and by knowing the For a machine element subjected to fatigue, the third condi­
endurance limit of the material selected, a designer can calcu­ tion, time, is usually the most demanding. Considering that a
late the conditions under which fatigue failures of contoured disk operating at high rotating speed accumulates an impress­
diaphragms will occur. By applying a reasonable safety factor, ive number of cycles every day, the only acceptable "reason­
he can then establish ratings. able length of time" is infinity. Fortunately, it is known (3)
But verification of the work, experimental and theoretical, that, for steels, infinite life is obtained if the parts can success­
can be done only through testing the couplings to failure. fully operate for more than ten million cycles. Our tests were
conducted until failure, or until a minimum of 10" cycles were
Testing in a laboratory, as compared with field experi­
accumulated. All the failures occurred before a disk could ac­
ence, has two major advantages: first, it can be done under
cumulate more than four million cycles.
well controlled conditions, and second, the testing can be
stopped shortly after failure occurs. c. Test Procedure
Testing in a laboratory also has disadvantages; it wastes To obtain meaningful results, the tests had to be con­
energy, and the size of couplings that can be tested is limited. ducted at various levels of constant and cyclic stresses. Any of
To offset these disadvantages, testing of contoured diaphragm the three constant stresses (torque, centrifugal force and axial
couplings was performed in a "four-square" rig, described in displacement) could have been varied. Both theoretical
the Appendix. analysis and experimental work showed that the axial dis-
 METAL DIAPHRAGM COUPLING PERFORMANCE 121

placement imposes the largest stress of the three. Also, in field

service, it is more likely that a diaphragm coupling can be
overstressed through improper axial spacing than through
overtorquing or overspeeding. It was hence decided that var­
ious levels of constant stresses would be applied by gradually
increasing the shaft spacing while the torque and speed were
kept constant.
A graph was established, having as coordinates the axial
displacement and the angular misalignment per disk, as shown
in Figure No. 12. It was known from stress analysis that a curve
exists on this graph . . . a curve which separates the infinite life
zone from the failure zone.
The object of our tests was to experimentally determine
this curve. Tests started zero axial displacement and a safe
amount of angular misalignment. Ater f 107 cycles the mis­
alignment was slightly increased and another 107 cycles were
accumulated. The process was repeated until failure of a disk
occurred. The failure defined the point "A" on Figure No. 12.
The failed coupling was replaced with a special gear coupling,
and testing was continued until the second diaphragm coupling
in the "four-square" rig failed. The second failure gave us the
point "A" for the second coupling size.
Al'!gular Misalignment
New diaphragm couplings were then installed in the test
rig. A certain amount of axial displacement was introduced at Figure 12. Typical Failure Diagram.
each coupling, and testing was resumed at a safe amount of
angular misalignment. The procedure described above was re­
peated and points "B" were obtained for the two coupling
sizes. After obtaining a sufficient number of points to define
the failure curve many "repeat" tests were performed to ver­
ifY the consistency that could be obtained with this type of
coupling.
d. Test Results
It was found that: a. As expected, the failure curve is
similar in nature to a typical Goodman diagram; and b. The
repeatability of failure points was excellent, in fact, many of the
tests resulted in the failure of both disks of a single coupling.
Angular Misalignment
Figure No. 13 gives a typical* example of a failure curve
determined experimentally, and compares it with the coupl­ Figure 13. Failure Curce of a Koppers Size 1MDH Coupling
ing's rating. Very similar results were obtained with other size Operating at 1,800 HP and 14,000 RPM.
couplings.
Constant Stress
RATING METHOD
Any product must incorporate a safety factor. A good
safety factor is a compromise between safe operation and a
reasonable cost. Once a safety factor is selected, the ratings of a
contoured diaphragm coupling can be established by tracing a
curve roughly parallel to the failure curve at a distance repre­
senting the safety factor (See Figure No. 1 4).
Such a rating has two disadvantages: a. It can cause seri­
ous misapplications at both extremes, and b. It is relatively
awkward to use. Let us analyze these disadvantages: a. Short
of laboratory conditions, installations having either no axial
displacement or no angular misalignment are highly improba­
ble. To safeguard both the manufacturer and the user, it is
better to limit the maximum axial displacement and the angu­
lar misalignment as shown in Figure No. 15.
b. Such a rating curve is awkward to use because it re­
quires a graph. A simpler method is to specify a maximum axial
displacement and a maximum angular misalignment, as shown
in Figure No. 16. The drawback of that rating method is that
the safety factor varies widely, and in extreme conditions can
be smaller than the one originally selected. Cyclic Stress

*Note: The coupling with the smallest safety factor was chosen. Figure 14. Constant Safety Factor Rating Method.
122 PROCEE DINGS OF THE FIFTH TURBOMACHINERY SYMPOSIUM

oL----�----- lb---------w L-----------

Figure 1 8. Stress Variation Under Three Conditions.

Cyclic Stress

Figure 1.5. Simplified Rating Method A.

Constant Stress

170,000

Figure 19. Failure Under Zero Axial Displacement Condition.

Another method, which is a compromise of the previous

two, is to replace the constant safety factor portion of the rating
with a straight line, as shown in Figure No. 17.

FAILURE MODES
'
Figure No. 18 represents the stress variations in three
Cyclic Stress
hypothetical cases.
Figure 16. Simplified Rating Method B. \iVhile in the case of reversing stresses any particle of a
disk is subjected alternately to compression and tension, in the
case of high constant stresses any particle is subjected only to
various degrees of tension. It is to be expected that the failure
mode of a disk will be influenced by the amount of stress
reversal. Two distinct modes of failure were found, one at zero
axial displacement and the other at large axial displacement.
Figure No. 19 illustrates the a
f ilure of a disk that had zero
axial displacement. Three aspects characterize this type of fail­
ure: a. The crack line is circular and goes through the thinnest
portion of the disk; b. The crack is relatively smooth; and c.
There is no, or very little, buckling in the disk.
Figure Nos. 20 and 21 illustrate the failure of a disk that
had a large amount of axial displacement. The three aspects
that characterize this type of failure are the opposite of the
previous ones: a. The crack line has a random path going from
the thinnest to the thickest portions of the disk; b. The crack
line is very irregular; and c. There is severe buckling of the
unfailed part of the disk.
The transition of these characteristics from zero axial dis­
placement to large axial displacement is gradual only for as­
pects a. and c; a smooth crack line is obtained only when the
Figure 17. Simplified Rating Method C. axial displacement is zero.
 METAL DIAPHRAGM COUPLING PERFORMANCE 123

.Figure 22. Failure and Buckling Under Large Axial Displace­

ment and Angular Misalignment Condition.
Figure 20. Failure Under Large Axial Displacement and Angu­
lar Misalignment Condition.

Figure 23. "Four-:>quare" Test Rig.

By observing a failed disk one can easily identifY the cause

of failure, whether there was excessive angular misalignment,
axial displacement or torque.

APPENDIX A
The "Four-square" Test Rig
This type of test rig has the advantage that the energy is
Figure 2 1. Double Failure Under Large Axial Displacement recirculated within the rig, hence only the energy lost in fric­
and Angular Misalignment Condition. tion must be supplied externally. Schematically, the rig is
shown in Figure No. 23.
Another aspect of the failure mode is significant. As shown The test rig on which the metal diaphragm couplings were
in Figure No. 22, the crack line propagates more than 270° tested can develop 4,000 HP, and the maximum speed is
before buckling of the disk takes place. As it was anticipated 20,000 RPM. The heat generated through fi·iction is trans­
theoretically, this condition indicates that the torque load ferred by the lubricating oil to a water circuit.
makes only a small contribution to the total stress in the disk.

REFERENCES
CONCLUSIONS
Metal Diaphragm Couplings of the type tested are highly 1. Wolff, P. H. W., ..The Design of Flexible Disk Misalign­
reliable pieces of machinery when operated within their rated ment Couplings"; Proceedings of Institute of Mechanical
Engineering; London, 19 48 (revised 1950).
conditions. Angular misalignments had to be increased at least
three times beyond ratings before failure could be induced. 2. Timoshenko, S., .. Strength of Materials," Part II, page 205;
In its ability to transmit torque the coupling is even safer Van Nostrand Co., 1958.
. . . a disk can transmit the rated torque even after more than 3. Battelle Memorial Institute, .. Prevention of the Failure of
half of it is broken. Metals under Repeated Stress," page 108; Wiley, 19 41.
124 PROCEE Dll'.'GS OF THE FIFTH TUHBOMAC:Hl\iEHY SYIIIPOSIU\1