A l t r a I n d u s t r i a l M o t i o n

Ameridrives

Bauer Gear Motor

Bibby Turboflex

Boston Gear
Revisiting Torsional
Delroyd Worm Gear Stiffness of Flexible Disc
Formsprag Clutch
Couplings
Guardian Couplings

Huco

Industrial Clutch

Inertia Dynamics

Kilian

Lamiflex Couplings

Marland Clutch

Matrix

Nuttall Gear

Stieber Clutch

Stromag

Svendborg Brakes

TB Wood’s

Twiflex Limited

Warner Electric

Warner Linear

Wichita Clutch
Revisiting Torsional Stiffness of
Flexible Disc Couplings
Abstract
System torsional analysis is performed by packagers and OEMs to ensure the
reliable operation of assembled systems such as engine-compressor packages. This is
especially critical when the prime mover or the load are reciprocating machines, with
associated high levels of oscillating torque. Unexpectedly high shaft stresses are likely to
occur if the frequency of the oscillating torque or multiples of it coincide with a torsional
critical speed. Failure to rigorously and accurately model this behavior can result in
fatigue of torque transmitting components, with disastrous results (figure 1).
Flexible coupling characteristics are among the parameters used in system torsional
analysis. Coupling manufacturers routinely supply weight, inertia, center of gravity,
and torsional stiffness data for use by the system analyst. During the development of
a new disc coupling product line, discrepancies were noted between historically and
widely used coupling torsional stiffness data vs. the stiffness measured in laboratory
tests. Subsequently, in-depth investigation was undertaken using FEA (Finite Element
Analysis) and multiple lab tests on the proposed new coupling line, as well as on
competitor couplings, in an attempt to more accurately model coupling torsional stiffness
characteristics.
Based on this research, new equations were derived which provide significant
improvement in the accuracy of flexible disc coupling torsional stiffness calculations.

FIG. 1- THE CONSEQUENCES OF A TORSIONAL

VIBRATION PROBLEM
Nomenclature
Units used in this paper are in the lb-inch-second
system.
AGMA American Gear Manufactureres Association
BSE Distance between shaft ends (in.)
cpm cycles per minute
d Deflection at DBC, inches, in a tangential direction
DBC Diameter of Disc Bolt Circle (in.)
Di Disc. ID, taken as adjoining tube OD (in.)
Do Disc. OD, taken as bolt DBC, in.
E Tensile modulus of elasticity, 29 x 106 psi for steel
F Flange thickness (in.)
FEA Finite Element Analysis
FFT Fast Fourier Transform, a method of separating a
vibration signal into its component parts
g Gravitational constant, 386 in/sec2
G Shear modulus of elasticity, 11.5 x 106 psi for steel
hz frequency in cycles per second (hertz) DISC PACK
ID Inside diameter of tube (in.) PT DISC BOLT
LOCKNUT
K, Kt Torsional stiffness (in-lb/radian)
KFLG Torsional stiffness of flange
L Length (in.)
NC Natural frequency, cpm
OD Outside diameter of tube (in.)
BSE
PT Pack thickness (in) (fig. 2)
psi Pounds per square inch
P Density (lb/in3)
R Bolt circle radius (in.)
Ss Shear stress (psi)
T Rated torque, in-lb HUB

TVA Torsional Vibration Analysis SPACER

µ coefficient of friction
FLYWHEEL ADAPTER
W Weight (lb)
WR2 Inertia (lb-in2), technically known as the polar FIG. 2- TYPICAL DISC COUPLING FOR
weight moment of inertia. Must be divided by ‘g’ GAS COMPRESSION SERVICE
to get mass moment of inertia used for torsional
critical speed calculations.
Introduction
Coupling manufacturers have historically calculated mass-elastic data for their
products by breaking the coupling components down into simple tube and disk sections,
then using conventional equations to calculate the properties of the sections. For weight
and inertia, the sections are summed to provide the data for the component or coupling,
and, for torsional stiffness, the reciprocals of the sections are summed to obtain the
reciprocal of the total stiffness. In equation form:

W = W1 + W2 + W3 + …. (1)
WR2 = WR21+ WR22 + WR23 + … (2)
1/K= 1/K1 + 1/K2 + 1/K3 + … (3)

where subscripts 1, 2, 3, … refer to the individual section numbers. Two concepts to note
from the stiffness equation:
a) the total stiffness of a coupling is always less than the stiffness of the softest
section or component in it
b) every section you add causes the total stiffness to decrease

A full description of the relevant calculations is provided in AGMA 9004, Flexible

Couplings- Mass Elastic Properties and Other Characteristics [1].

Most, if not all, coupling manufacturers generally perform their calculations in

accordance with AGMA 9004. However, AGMA 9004 pointedly avoids attempting to
dictate calculation methods for torsional stiffness of flex elements and bolting, leaving
these two areas to the discretion of the manufacturer. All other parts of a coupling are
straightforward to analyze, providing they are cylindrical, so consistent results can be
expected from various manufacturers for all parts other than flex elements, bolting, and
noncylindrical parts.
For relatively long couplings (say 18” or longer BSE), variations in the calculation
of stiffness of flex elements and bolting have a minimal influence on the bottom line.
But for the short couplings commonly used in the gas compression industry, most of the
components of the coupling do have a significant effect on the total- see figure 3. Note
that hub stiffness includes the use of the 1/3 shaft penetration factor which causes the
shaft diameter to have a major impact on hub stiffness. Furthermore the number 1/3 must
be considered a nominal value as the actual value depends on the tightness of the hub
shrink and the accuracy of the key fit.

FIG. 3- BREAKDOWN OF SHORT DISC COUPLING

TORSIONAL STIFFNESS BY COMPONENT
(8.45” BSE, 15” OD, 6” BORE)
 Consequently, when dealing with short couplings, variations in both manufacturing
tolerances and the manufacturers’ approach to flex element and bolt stiffness calculations
are likely to cause significant variations in the accuracy of the data furnished. Papers by
Murray et al [2] and by Varty et al [3] point out the consequences of inaccurate mass-
elastic data: the accuracy of the analysis is limited to the accuracy of its inputs. Murray
and Feese [9] suggest the use of ±20% on coupling stiffness. Users should consider the
confidence they have in the supplier’s modeling approach.
This study focused specifically on short 8-bolt heavy-duty disc couplings as used in
the gas compression industry. Other disc designs may have different characteristics.

Disc Coupling Features to Note

Flexible disc couplings (figure 2) are of simple construction and have been in
common use for years, but analysis of their torsional characteristics is more complex
than expected.
Disc couplings accommodate misalignment by flexing of the disc material between
two sets of bolts on a common bolt circle, one set being attached to the driving member,
say a hub, and the other to the driven member, such as a spacer piece. In the case of an
8-bolt design commonly used in heavy duty applications, four bolts attach rigidly to the
hub and four to the spacer (figure 6)

FIG. 4A- TORQUE-TRANSMITTING SHEAR BOLT FIG. 4B- TORQUE-TRANSMITTING CANTILEVER

IN TWO SOLID FLANGES (GEAR COUPLING) BOLT IN ONE SOLID FLANGE (DISC COUPLING)

a) Cantilevered Bolts
Bolted flange connections in other flexible coupling types, such as gear couplings,
rely on “through” bolts in shear and/or face friction to transmit torque from one flange
to another (figure 4A). This contrasts with the situation in disc couplings where the bolts
are loaded in a cantilever fashion (figure 4B). In both cases a certain amount of torque
is carried by friction, then when the friction capacity is exceeded, by loading on the bolt
body. Cantilever bolts have higher stress and lower stiffness than through bolts.
b) Unitizing Effect of Washers & Preload
Very high bolt preloads are commonly used in disc couplings, to reduce fretting
and to minimize fatigue in the bolt. The high clamp load from bolt tightening, acting
in concert with the bushing or washers (figure 5) acts to “unitize” the disc and washer
area, tending to make it behave like a solid cylinder with an outside diameter equal to
the washer OD. This to some extent reduces the deflection associated with cantilever
mounting, although at best it is still a cantilever member, but with a larger effective
OD. However, as will be illustrated later, what it also does is transfer increased bending
loads to the adjacent flange compared to the rigid flange joint of figure 4A. The disc
stack construction also appears to have some effect on stiffness, the bushed pack shown
in figure 5 tested stiffer than nonbushed designs, all other factors being equal.

DISC PACK
WASHER OD

S1 TEN
M-
CO ION SIO
N
SS
P RE
H4 H1

PR
N
LOAD

CO SION
SIO

ES
M-
TEN
S4 S - BOLTED TO SPACER S2
PR (DRIVER)

N
H - BOLTED TO HUB
CO SION

SIO
CLAMP
ES

(DRIVEN)
M-

TEN
LOAD

H3 H2
BUSHING TEN M-
SIO CO SION
OR N E S
S3 PR
WASHER

FIG. 5- DETAIL OF BOLTED CONNECTION FIG. 6- TORQUE LOAD DISTRIBUTION IN FLEX DISC

c) Series vs. Parallel Summation of Bolt Stiffness

Again comparing disc to gear couplings, one set of eight flange bolts in a gear
coupling carries the transmitted torque in 8 parallel paths, presenting a bolt stiffness per
flange of 8 times the stiffness of one bolt. In a disc coupling, half of a set of 8 are bolted
to each member, which means that the torque path goes through one set of 4 bolts in
parallel then through another set of 4 in parallel, making the total stiffness of 8 bolts
equal to 2x the stiffness of one bolt, instead of 8x. Then for a spacer type disc coupling,
two flanges are again in series, resulting in a net bolt stiffness for the entire coupling
equal to the stiffness of one single bolt!
FIG. 7- COMPRESSION LEG BUCKLING POINT FIG. 8- NONLINEAR STIFFNESS vs. TORQUE-
vs. COUPLING SIZE 6.75” OD COUPLING LAB TEST

d) Tension & Compression Legs

By virtue of having the disc bolts alternately connected to the flanges of the
driving and the driven members, the disc segments are alternately loaded in tension and
compression as shown in figure 6. Being composed of thin membrane-type material, the
compression legs tend to buckle laterally as torque load increases (figures 19, 20 & 28).
The consequence of this is a nonlinear torsional stiffness, decreasing with increasing
torque load. This effect was clearly visible in the static lab tests (figure 8). The onset
of compression leg buckling (figure 7) occurs earlier on larger couplings because the
material thickness is typically constant on all sizes, but the span length increases as
coupling size increases.
In actual service, the nonlinearity is likely to be somewhat counteracted by the tendency
of the discs to stiffen under centrifugal loads due to speed, and by the tension created by
axial displacement.
As the astute reader will note from examining figures 3 & 8, compression leg buckling
alone is not sufficient to account for the amount of nonlinearity found during testing.

e) Stiffness of Spider Spacer Flanges

Most disc coupling spacers in gas compression service are the “spider”
type shown in figure 9. This is to provide wrench access to the disc bolts
considering the relatively short BSE distances used. The flange stiffness
for this configuration is clearly going to be less than for a solid flange.
Again the modeling approach is difficult and not addressed by AGMA
9004, so variations in manufacturers’ spider geometry and analysis have
inevitably resulted in variations in calculated stiffness.

FIG. 9- SPIDER TYPE SPACER

f) Material Properties
The material properties of course affect the torsional stiffness of the components.
Depending on manufacturing volume and lot size, major coupling components can be
made from cast iron, cast ductile iron, medium carbon steel, or alloy steel. As economic
circumstances change over the years, coupling vendors may and have changed the
material they use in their products. The calculated stiffness of a coupling must be based
on the actual material used to produce it. Spacers have typically migrated from cast iron
to ductile iron or 1045 steel, and hubs from cast iron to 1045. Vendor catalog data have
not always kept up with these changes.
Properties used in the development of this paper are listed in Table 1.

TABLE 1. COUPLING MATERIAL PROPERTIES

Material Density, LB/IN3 E, LB/IN2 G, LB/IN2 Poisson’s Ratio

Steel .283 29,000,000 11,420,000 .27
Cast Iron .260 15,200,000 6,080,000 .25
Ductile Iron .260 23,500,000 9,400,000 .25

g) Effect of Nonlinear Stiffness

Referenced papers by Feese & Hill [9, 13] mention dealing with the effects of
nonlinear coupling stiffness. Den Hartog [4] devotes an entire chapter to nonlinear
behavior and its effect on system analysis. This is beyond the scope of this study, but
in general, nonlinear spring rates have a beneficial effect by reducing the amplitude
of the forced response. For example, API-671 [8] refers to this issue by exempting
couplings with nonlinear axial stiffness from complying with certain axial resonance
requirements. Nonlinear stiffness will result in different response behavior for
increasing vs. decreasing speed changes. According to Den Hartog, when stiffness
decreases with load, the response curve has the skewed shape shown in figure
10. During acceleration the vibration amplitude follows points AFBCD. During
deceleration the response curve follows the path DCEFA. Therefore the amplitudes
at resonance would be expected to be greater on coast-down than on run-up. This is
probably beneficial since torque loads are normally lower on coastdown.
VIBRATION AMPLITUDE

SPEED, CPM

FIG. 10- FORCED RESPONSE CURVE WITH A

NONLINEAR SPRING THAT HAS DECREASING
STIFFNESS VS. LOAD
Fea Modeling of Disc Couplings
Pro-Engineer is used at Ameridrives for modeling and drawing production. Tightly
integrated with Pro-E is Mechanica FEA which analyzes the Pro-E models directly to
obtain stress, deflection, natural frequencies, etc. Torsional stiffness values are obtained
by loading a model with a selected torque, in this case the coupling normal rated torque,
then dividing the torque by the resulting deflection in radians.
Deflections in inches were obtained at the intersection of the flex pack centerline
and the disc bolt circle diameter, in the tangential direction. FEA torsional stiffness
values were obtained from
Kt = T x R ... (4)
d

TABLE 2. FEA VALIDATION CHECKS

Stiffness Shear
Dimensions Calculation WR2
Section in-lb/rad Weight (lb.) Stess
(in.) Method LB-IN2
x 106 PSI
Steel 5.00 OD Conventional 31.20 21.19 111.4 3049
Spacer x 4.13 ID
Tube x 12.00 long FEA 31.20 21.19 111.4 3072
Steel 6.75 DBC Conventional 1514 1.775 16.67 1721
Hub x 5.44 ID
Flange x .50 thick FEA 1512 1.775 16.67 1754

With any FEA analysis, care must be taken to establish the proper application of
loads and constraints. It is very easy to get erroneous FEA results, and the nature of the
FEA output tends to lend credence to the work whether justified or not.
Another advantage of 3D modeling is that precise weight and inertia values are
provided as a ‘free’ by-product of the modeling work. This is especially useful for non-
uniform parts such as the spider spacers.
Validation calculations were performed on simple tube and flange sections, and
compared to traditional stiffness and inertia equations as follows:

KTUBE = π x G x (OD4 - ID4) ... (5) WR2 = W x (OD2 - ID2) ... (8)
32 x LTube 8

KFLG = π x G x Do2 x Dj2 x F ... (6) SsTUBE = 16 x T x OD ... (9)

(Do2- Di2) π x (OD4- ID4)

W = P x π x OD2 - ID2) x L ... (7) SsFLG = 2xT ... (10)

4 π x d2 x F
 Validation comparisons in table 2 indicate good correlation between FEA and
conventional equations.
Considerations For FEA Modeling (Figures 11 - 14)

a) Cantilever Bolt Loads

Of prime importance in developing the models for spacers, hubs, and adapters
was to properly represent the cantilever bolt loads. Early models with loads applied
inside the bolt holes were discarded in favor of adding to the flange face four pins with
a diameter equal to the washer OD, and a length equal to half the flex pack thickness.
Torque loads were applied to the faces of these pins.
A note here to the reader, figures 11-14 show the models as they appear in
Mechanica. The yellow arrows are shown in the axial direction which is misleading-
the FEA loading is tangential but it appears as shown. There is a “preview” button that
shows it properly on the screen but doesn’t allow printing.

b) Spacers (Figure 11)

Some point on an FEA model must be anchored to ‘ground’ in all directions to
prevent rigid body motion where the model drifts off into space. Anchoring one end of
the spacer led to difficulties because the constraints undesirably restricted deflections
due to loads. Making a much more elaborate model was one option, but good results
were obtained by modeling half the spacer and constraining the cut plane to zero
displacement, which duplicates the condition of symmetric torsional windup about this
imaginary plane in actual practice. This also had the advantage of quicker FEA runs. A
factor of 2 is accordingly required in the denominator of equation #4.
Another consideration is the restriction imposed on the pin deflection by the
bending stiffness of the flex pack. This acts to increase the apparent stiffness of the
assembly slightly. Rather then complicate the models for minimal benefit, a factor of
1.05 is added to the numerator of equation 4 to cover this.

c) Hub (Figure 12)

The hub was modeled from the flange face up to the plane defined by the 1/3 shaft
penetration method, i.e., with a length equal to 2/3 of the total length through the bore.
The hub end defined by the 1/3 penetration plane was fixed at zero displacement. The
deflection of the shaft portion was calculated manually and added to the FEA results.

FIG. 11- HALF SPACER FEA MODELS FIG. 12- HUB FEA MODEL
d) Adapters (Figure 13)
The interface bolt holes where the adapter bolts to the flywheel were constrained to
zero displacement axially in the bolt head spotface, and rotationally about the part axis.
Torque loads were applied through four pins like the spacers. Where the adapter bolts
up against the flywheel, bending deflection of the flange perpendicular to the flywheel is
limited to moving away from the flange only.

e) Individual Discs (Figure 14)

A pin with a total length equal to the disk pack thickness was modeled projecting
out from each side of the disc equally. On one side of the disc half the pin ends were
fixed; on the other side of the disc the ends of half the pin ends were loaded with the
rated torque per disc. The ends of the loaded pins were constrained to stay in their
original plane, to prevent adding cantilever deflection that was already considered in the
connected pieces.

g) Disc Pack Assembly (not shown)

In most cases, assembled packs became too complex to run successfully, even with
symmetry simplifications, but one successful pack model was developed. Modeling
arrangement was the same as for the individual disc.

FIG. 13- ADAPTER FEA MODEL FIG. 14- DISC FEA MODELS

Lab Testing
Several coupling configurations were tested on Ameridrives’ large static test
stand with a capacity of 600,000 in-lb (figure 15). This consists of a large torque arm
and shaft carried by four heavy roller bearings. When unloaded the arm and shaft can
be rotated with one finger indicating low friction losses. Friction loss using =.05 was
included in the stiffness calculations. Force is applied by a hydraulic cylinder and read
by a calibrated load cell and digital panel meter. Deflections are read manually from dial
indicators placed tangentially on the coupling with their bases on or connected to the
fixed stand.
Some General Notes on Lab Testing
1. Couplings were aligned within ±.010” parallel, angular, and axial.
2. Disc bolts were coated with silver Neverseez and torqued to manufacturer’s
recommended values.
3. After installation of each coupling and prior to taking readings, the coupling was
torqued to 110% of the manufacturer’s rated normal capacity and released, then the
indicators were zeroed. This removed whatever clearances could be removed at that
torque level.
4. Indicators #1 and #4 (figure 16), placed at each end of the coupling, reading on the
web or on the back side of the flange adapters, gave total coupling stiffness and
were considered the most reliable. Other indicators used in various spots to try to
determine the stiffness of each component, gave erratic results.
5. In some cases the indicators are subject to cosine error if the dial and mounting
arrangement prevent a true perpendicular mounting. A correction factor of cos
(angle) was included where appropriate.

#4

#2
#3

#1

FIG. 15- STATIC TEST STAND FIG. 16- INSTALLED COUPLING READY FOR TEST.
INDICATORS #1-4 READ DEFLECTION

Stiffness Check by NC Measurement

The large torque arm has a lot of rotational inertia, and being connected to ground
(the test stand) by a spring (coupling and torque arm shaft), it should exhibit a single
strong torsional vibration mode. An accelerometer was mounted vertically near the end
of the torque arm and connected to a portable FFT analyzer. Striking the arm vertically
with a heavy dead-blow mallet produced a strong signal at 28.2 hz as shown in figure
17.
Parameters required to calculate this natural frequency:

Coupling as tested (round flg. spacer)

K = 24.1 x 106 in-lb/rad
WR2 = 772 lb-in2

Test Stand Shaft

Shafting K = 17.6 x 106 in-lb/rad
Shafting WR2 = 1555 lb-in2
Torque arm WR2 = 122,515 lb-in2

Total stiffness = 10.2 x 106 in-lb/rad

Total WR2 = 124,456 lb-in2

For a single mass system

Nc = 187.7 x ( K / WR2 )1/2 cpm ... (11)

For the system as tested

Nc = 1699 cpm or 28.3 hz

This is in good agreement with the measured frequency of 28.2 hz. Note
that this cannot be considered a precise verification, due to the square root
relationship, and the “series” stiffness addition, a 20% change in stiffness
only results in a 5% change in the calculated frequency.

FIG. 17- FFT ANALYZER READING FOR TORSIONAL RAP TEST WITH ROUND FLANGE SPACER
 FIGURE 18- SAMPLE OF STIFFNESS TEST RESULTS

Analysis of Lab Data

Over 100 stiffness tests were conducted. Some were identical repeats to gauge the
repeatability of the test, in others indicators were repositioned, couplings were removed
and reassembled, or bolts were loosened & retightened.
Given the small deflections involved, an error of only .001” can have a significant
effect on the results. While the indicators can be read to .001” or less, considerable
variations in test results were found, which seemed to be due to indicator setup and
coupling assembly.
Figure 18 is one sample of the test runs. With data scatter and nonlinearities
evident, a method was needed to reduce the data to usable numbers. Going back to
basics and referring to figure 21, torsional stiffness is defined as unit of load (torque)
per unit of deflection (radians).

Kt = rise / run = in-lb / radian ... (12)

When the test data is entered and graphed in Excel® with radians on the X-axis
and torque on the Y-axis, a curve-fit equation can be obtained which expresses the
torque as a function of deflection. Taking the derivative of this equation gives the
instantaneous slope at any point, which is the stiffness at that point.

FIG. 19- COUPLING AT 267% OF RATED TORQUE FIG. 20- COUPLING AT 500% OF RATED TORQUE
FIG. 21- WAYS TO DEFINE NONLINEAR STIFFNESS

For the example in figure 21 where y = torque and x = displacement, and with a
factor of 1000 to get the units to be in-lb/rad x 106

y = 2.2x3 - 19.7x2 + 72.8x + 18.3 ... (13)

Kt = dy/dx = 6.6 x2 - 39.5x + 72.8 ... (14)

Considering the graph lines one at a time:

Line ‘A’ is a smooth curve fit to the test data.

Line ‘B1’ has endpoints passing through (0,0) and through the data point corresponding
to the deflection at rated torque. Kt = 39

Line ‘B2’ has the same slope (stiffness) as B1, placed tangent to the curve. Kt = 39

Line ‘C’ is the slope of the curve at rated torque (min. stiffness). Kt = 13.8

Line ‘D’ is the slope of the curve at zero torque (max. stiffness). Kt = 72.8

So the stiffness of this coupling could be defined as being anything from 14 to

73! This dismaying fact probably has much to do with the variations found in coupling
stiffness calculations. Ultimately the stiffness may have to be defined by a curve or
equation, much like axial stiffness is for disc couplings, rather than a single easy-to-
handle number.
For the purposes of this study, coupling stiffness is defined by the slope of line
B2, which from figure 21 corresponds to a torque level of 63% of rated torque. This
simplifies the calculations and provides a number that seems appropriate for the torque
levels that might be expected during torsional resonance.
Further Notes on Lab Tests:
1. The first few readings are always erratic especially on larger coupling sizes.
Stiffness results under about 20% of rated torque can only be inferred from an
averaged trend line. When the compression leg buckling point falls in this area, the
data become even more erratic.
2. At torque levels exceeding 100% of rating, minor slippage of disc material under
the bolt heads can occur, made evident by residual gaps between discs that were not
present initially. Therefore most of the testing was limited to 100% of rated torque.
3. Attempts to read individual spacer and flex pack stiffnesses were not consistent due
to flange bending and twisting deflections, so they were not included in the results.

TABLE 3- SUMMARY OF TORSIONAL STIFFNESS LAB TESTS & FEA CALCULATIONS

Measured Average
Coupling Spacer Stiffness Measured Calculated
Test No. Brand Spacer Design Material Based On Stiffness Stiffness
Deflection At IN-LB/RAD x 106
100% Torque IN-LB/RAD x 10
6

31-1 A Spider Cast Iron 9.02

31-2 A Spider Cast Iron 8.36
9.19 10.3
31-3 A Spider Cast Iron 9.53
31-4 A Spider Cast Iron 9.84
312-1 B Spider Cast Iron 9.54
312-2 B Spider Cast Iron 7.97
312-3 B Spider Cast Iron 8.12 8.15 8.90
312-4 B Spider Cast Iron 8.29
312-5 B Spider Cast Iron 6.84
40-1 C Round Flange Steel 24.8
40-2 C Round Flange Steel 25.3
40-3 C Round Flange Steel 26.1
40-14 C Round Flange Steel 21.8
23.2 24.9
40-15 C Round Flange Steel 22.2
40-16 C Round Flange Steel 22.2
40-17 C Round Flange Steel 21.6
40-18 C Round Flange Steel 21.8
40-4 C Spider 3.01 LG Steel 22.0
40-5 C Spider 3.01 LG Steel 22.2 22.4 22.3
40-6 C Spider 3.01 LG Steel 22.8
40-7 C Scalloped Steel 22.8
40-8 C Scalloped Steel 22.9 23.1 24.3
40-9 C Scalloped Steel 23.7
40-10 C Spider 3.58 LG Ductile Iron 21.8
40-11 C Spider 3.58 LG Ductile Iron 21.6
20.9 18.7
40-12 C Spider 3.58 LG Ductile Iron 22.2
40-13 C Spider 3.58 LG Ductile Iron 18.1
40-19 C Spider 3.01 LG Ductile Iron 20.8
19.9 20.8
40-20 C Spider 3.01 LG Ductile Iron 19.0
Lab tests run on larger, stiffer couplings were inconclusive due to wildly fluctuating data. More work on
this is planned.
ANALYSIS OF FEA DATA
Spacer FEA
The spider spacer flanges deflect significantly in a bending and twisting direction
in response to the cantilever bolt loads (Figure 22). The thin light blue lines represent the
undeflected shape of the arm. The deflection is exaggerated for clarification.
The deflection visible in Figure 22 includes a twisting component about a radial
axis and a bending component about the coupling’s rotational axis. The magnitude of
the twisting deflection, previously not considered in torsional stiffness calculations, is
greater than that of the bending magnitude. Figure 23 shows a breakdown of the sources
of torsional deflection of this spacer. The twisting deflection is also present to a lesser
degree in round flange spacers.arrangement was the same as for the individual disc.

Hub FEA
A conventional Kt calculation would divide the hub into three sections as shown in
figure 24, which includes 1/3 shaft penetration based on Ker Wilson’s [5] approach. For
this analysis, section 1 is done by hand and sections 2 & 3 are modeled in FEA.
A deflected hub model is shown in Figure 25. Note the out-of-plane bending
deflection similar to that of the spacer spider arm. Figure 26 breaks down the deflection
of the total hub. This is for a hub with catalog dimensions and a bore diameter of .85x
max bore.

1/3 SHAFT
PENETRATION
SECTION

FIG. 22- DEFLECTION IN SPIDER ARM

FIG. 24- SECTIONS FOR HUB Kt CALCULATIONS

FIG. 23- SOURCES OF DEFLECTION IN FIG. 26- BREAKDOWN OF TOTAL

FIG. 25- DEFLECTED HUB FLANGE SPIDER SPACER OF FIG. 16 HUB STIFFNESS
Flywheel Adapter FEA
Here again out-of-plane bending is evident, affecting the entire part since it is relatively
short (Figure 27). Choice of SAE or HD bolting does have a small but reportable effect
on the stiffness.

Disc FEA
Modeling the discs is fairly straightforward since no cantilever loads are
included. What is unique here is the behavior of the compression legs (Ref. Figures
6, 19, 20, & 28). At some torque load depending on coupling size, buckling of the
compression legs starts to become visually evident. Well before this happens, the
contribution of the compression legs to disc stiffness starts to drop off, resulting in a
decreasing disc stiffness vs. torque load.
By invoking the nonlinear geometry option in Mechanica, the buckling
behavior of one disc can be accurately modeled. With this option the load is broken into
several steps, in this case three, and the software recalculates the deflections based on
the previous step’s deflected shape. Figure 29 shows how the calculated disc stiffness
varies with torque load, and how that variation affects overall coupling stiffness. Lab
test data, however, showed considerably more nonlinear behavior than can be attributed
to compression leg buckling alone.

Disc Pack FEA

Modeling entire disc packs caused memory overflows because of the many
intersections where the solid pin cylinder passed through each thin disc. The software
automatically refines its mesh in such areas causing an inordinate number of equations
to solve. Attempts at simplification by utilizing symmetry and shell elements were
unsuccessful.
One pack model was run successfully (Figure 30), and it displayed some
interesting behavior. The stiffness of this pack with 11 discs was 53% of 11 times the
stiffness per disc. The reasons for this were not clear but may be due to unequal load
sharing within the disc pack.

FIG. 28- NONLINEAR DISC FEA RUN SHOWING

FIG. 27- ADAPTER FLANGE DEFLECTION BUCKLING IN COMPRESSION LEGS.
TENSION LEGS ARE FLAT
Correlation of Test Data & FEA
The results of lab tests and FEA runs are tabulated and compared in Table 3.
Coupling stiffness for purposes of this study is the summation of the following:

(2) flywheel adapters

(2) disc packs
(2) bolt sets
(1) spacer

Classical equations were derived as well as possible to describe each component,

then the discrepancy between equation results and FEA results was assessed for
the range of coupling sizes. Coefficients or curve-fit equations (fudge factors) were
established to bring the two methods into agreement.
The contribution of the bolts to coupling stiffness is very difficult to determine
either analytically or by testing. Therefore the bolts were used as the “everything else”
factor to correlate test results to analytical.
In the case of the spider spacer, considerable time was spent on deriving equations
to calculate bending, twisting, and shear deflections, but ultimately, better correlation
was obtained by just curve-fitting the FEA data and using a basic equation containing the
parameters that describe the part.

What About Legacy Data

Coupling users may ask themselves what to think about existing installations in
light of this study. Certainly there are hundreds if not thousands of installations around
the world operating smoothly and safely on the basis of ‘legacy’ coupling data. On the
other hand, use of the latest calculation techniques gives coupling manufacturers the
chance to improve the accuracy of their mass-elastic data. Overall, this study’s results
suggest that actual coupling torsional stiffness is greater than past calculations indicated.
Centrifugal and axial loads in service are likely to increase this discrepancy.

FIG. 29- TORSIONAL STIFFNESS OF 6.75” OD FIG. 30- FEA DEFLECTION DISTRIBUTION
DISC PACK CALCULATED BY NONLINEAR FEA WITHIN DISC PACK
So Whats The Bottom Line?
To investigate the effect of changes in stiffness on system behavior, Ameridrives
commissioned two system analysis companies to perform studies of ‘typical’ IC engine
- reciprocating compressor systems.

TVA Study #1
This was a short preliminary study by Advanced Vibration Solutions, Windermere,
FL, [14] looking at an existing relatively simple system for frequency effects only.
Calculations were run over the range from 100% to 800% of vendor-specified coupling
stiffness.
Driver: Waukesha V-12 engine
Coupling: Brand ‘A’ 15” dia.
Driven: 2-throw piston compressor
Conclusion: results in table 4 indicated no harmful effects on the vibration behavior of
this system when the coupling stiffness was changed over a large range of values.

TABLE 4- RESULTS OF TVA #1

Coupling Torsional Natural Frequency (CPM)

Torsional Mode 1 Mode 2 Mode 3
Stiffness
1x 4,710 12,030 13,030
2x 4,710 12,100 15,500
3x 4,720 12,110 16,660
4x 4,720 12,110 17,320
5x 4,720 12,110 17,750
6x 4,720 12,110 18,040
7x 4,720 12,110 18,260
8x 4,720 12,110 18,420
1 x stiffness = 82.5 x 106, operating speed = 1000 rpm
All harmonics of mode 3 are safely above 1000 cpm

Cplg C1
Cplg C1
Cplg B1

Cplg B1
Cplg A1
Cplg A1

FIGURE 31: SAMPLE PAGE FROM IDC REPORT

TVA Study #2
This was a full forced response study by IDC Technical, Metamora, IL, involving a
variety of couplings and connected equipment [15].

Driver: Caterpillar G3516 V16 engine

Couplings: Brand ‘A1’ 19” OD Kt = 104

Brand ‘B1’ 19” OD Kt = 108
Brand ‘C1’ 18” OD Kt = 120
Brand ‘A2’ 16.75” OD Kt = 73
Brand ‘B2’ 16.75” OD Kt = 83
Brand ‘C2’ 15” OD Kt = 82
Driven: Ariel JGE/4 compressor &
Ariel JGT/4 compressor

Opr Cond: 1000 kW @ 1400 rpm

The report is too lengthy to include here, but Figure 31 is one sample page.
Conclusion: all stresses and amplitudes for all combinations of couplings and
compressors remained within allowable limits regardless of coupling type.

Overall Conclusions
These studies only apply to the couplings and systems actually studied. However,
it can be concluded that, at least in some cases, errors or changes in coupling stiffness
calculations may not have any appreciable effect on torsional behavior.
Conversely, many readers will know personally of instances where coupling
stiffness does have a major impact on vibration levels, and where coupling stiffness or
inertia were intentionally changed to achieve a desired result.
It is recommended that torsional analysts bear in mind the possibility of inaccuracy
in their coupling mass-elastic data, and consider the benefits of performing field
measurements to confirm that actual vibration signatures agree with predicted values
(figure 32).
The author is interested in users’ experiences one way or the other concerning
measured vs. calculated torsional vibration characteristics. See the biography for e-mail
address.

FIG. 32- TORSIONAL LASER VIBROMETER USED

TO MEASURE TORSIONAL VIBRATION AT THE
COUPLING (COURTESY OF BETA MACHINERY)
Appendix - Equations
The following equations were developed to approximate the results of the FEA
and lab testing performed in this study. No attempt is made to rigorously prove their
derivation, they are based on standard accepted equations listed below, with modifiers
to achieve correlation with test results. They are offered as reference only- the user is
responsible for confirming their accuracy for any particular coupling or application. The
equations only apply to the tested 8-bolt heavy duty couplings within the normal range of
dimensions in inches. See pages 1-2 for terms not defined below.

a) Stiffness of Spider Spacer Flange Per End:

KFLG = E x ADF3 x DBC1.91x F2 ... (15)
1.1 x (DBC - OD)
where ADF = arm design factor from table 6.

b) Stiffness of Round Spacer Flange Per End:

KFLG = π x G x DBC2 x OD2 x F x .75 ... (16)
3.8 x (DBC2 - OD2) x (F + .3).29
c) Hub Stiffness

Calculation includes 3 sections per figure 18. Sections 1 & 2 use the standard equation
(5). Section 3 uses the following:
KFLG = π x G x DBC2 x B2 x F ... (17)
(DBC2 - B2) x 8.89

where B = hub barrel diameter (in)

d) Stiffness of One 8-Bolt Disc Pack:

KPack = F2 x DBC1.1 x ((ODDISC x F1) - IDDISC) ) x t x Nd x E x 1.1 x F3 ... (18)
where F1, F2, F3 = factors from table 7, t = disc thickness (in), Nd = number of discs

e) Disc Bolt Set Stiffness for 8-Bolt Pack

KB = π x E x DBC2 x b4 x F3 ... (19)
(PT + b/2)3 x 5.33
where
KB = stiffness of one set of 8 bolts
F3 = factor from Table 8
b = bolt body dia.
PT = pack thickness per fig. 2 (distance from locknut face to flange face, whether
bushed or not)

f) Flywheel Adapter Stiffness

For section 1:
KFLG = π x G x Y2 x DBC2 x Z ... (20)
(Y2- DBC2)
where
Y = Interface bolt pattern DBC (in)
Z = Interface bolt flange thickness (in)

For section 2:
KTUBE = π x G x (DBC4 - Q4) x .65 ... (21)
32 x (V - Z/2)
where
Q = Adapter ID (in)
V = Adapter overall thickness (in)
Z = Interface bolt flange thickness (in)

Total adapter stiffness:

K = .3 / (1/KFLG + 1/KTUBE) ... (22)

TABLE 6- ARM DESIGN FACTORS FOR EQ. 15 TABLE 7. DISC PACK FACTORS FOR EQ. 18

Spacer Geometry ADF Round OD Flats On OD

Shape Factor F1
1.00 .92
Bushed Pack Unbushed Pack
Assy Factor F2
.64 + DBC/115 1.4 1.0
Buckling Factor
F3 = 0.5 + 1000/DBC4
F3 (≤ 1.0)

Rexnord CMR .52 + DBC/115 TABLE 8- DISC BOLT FACTOR FOR EQ. 19

Bushed Pack Unbushed Pack

Assy Factor F3
1.0 + 100/DBC2 1.00

Ameridrives GC 1.00
Biography For Sam Steiner
Graduated from Penn State University with a Bachelor of Science degree in
Mechanical Engineering. Employed with Zurn Industries - Mechanical Drives Division
(MDD) in Erie PA from 1977-1989, involved with all phases
of design, testing, manufacture, and sales of rotating flexible
couplings. Active member of the AGMA Flexible Couplings
Committee during this period. Worked as a design engineer
and as department manager in the design and manufacture of
large rubber mixers from 1989-2006, then returned to Zurn
MDD in 2006, now known as Ameridrives Couplings.

E-mail: [email protected] or
[email protected].

This paper was first presented at the Gas Machinery Research

Council Conference, Oct. 3, 2007, Dallas TX.
References
1. AGMA 9004-A99, Flexible Couplings-Mass
Elastic Properties and Other Characteristics,
American Gear Manufacturers Association,
Alexandria, VA, 1999.

2. Murray, Howes, Zacharias, & Chui of Beta

Machinery Analysis, Sensitivity of Torsional
Analyses to Uncertainty in Mass-Elastic
Properties, presented at the ASME International
Pipe Line Conference, Calgary, 1996.

3. Varty, R.V., and Harvey, J.D. of Beta Machinery

Analysis, Torsional Vibration Modeling and
Analysis Continued, GMRC Gas Machinery
Conference, 2003

4. Den Hartog, J.P., Mechanical Vibrations,

McGraw-Hill Book Company, New York, New
York, 1956.

5. Wilson, W. Ker, Practical Solution of Torsional

Vibration Problems Volume 1, Wiley & Sons,
New York, New York, 1956.

6. Neale, M.J., Proceedings of the International

Conference on Flexible Couplings for High
Powers and Speeds, Neale & Associates Ltd.,
Surrey, England, 1977.

7. Mancuso, J.R., Couplings and Joints: Design,

Selection, and Application, Marcel Dekker Inc.,
New York, New York, 1986.

8. API-671, Special-Purpose Couplings for

Petroleum, Chemical, and Gas Industry Services,
American Petroleum Institute, Washington D.C.,
1998
9. Feese, T.D., and Hill, C.H. of Engineering
Dynamics Inc., Guidelines for Improving
Reliability of Reciprocating Machinery by
Avoiding Torsional Vibration Problems, GMRC
Gas Machinery Conference, 2001.

10. Murray, Howes, & Zacharias of Beta Machinery

Analysis, A Systems Approach to Torsional
Analysis, GMRC Gas Machinery Conference,
1995.

11. Varty, R.V., and Harvey, J.D. of Beta Machinery

Analysis, Torsional Vibration: The Value of Field
Verification, GMRC Gas Machinery Conference,
2004.

12. Stephens, T. of Ariel Corp., Torsional Case

Studies on High Speed Separable Reciprocating
Compressors, GMRC Gas Machinery Conference,
2004

13. Feese, T.D., and Hill, C.H. of Engineering

Dynamics Inc., Guidelines for Preventing
Torsional Vibration Problems in Reciprocating
Machinery , GMRC Gas Machinery Conference,
2002

14. Zubritsky, P.D. of Advanced Vibration Solutions,

memo of 8/28/06 to Ameridrives concerning
torsional analysis.

15. Chundi, R.K. of IDC Technical Services,

Torsional Study for Ameridrives International,
report no’s 07-003-01 thru -03, 2007.
 About Altra Industrial Motion

Altra is a leading global designer and manufacturer of quality

power transmission and motion control products utilized on a
wide variety of industrial drivetrain applications. Altra clutches and
brakes, couplings, gearing and PT component product lines are
marketed under the industries most well known manufacturing
brands. Each brand is committed to the guiding principles of
operational excellence, continuous improvement and customer
satisfaction. Highly-engineered Altra solutions are sold in over
70 countries and utilized in a variety of major industrial markets,
including food processing, material handling, packaging machinery,
mining, energy, automotive, primary metals, turf and garden and
many others.

Altra’s leading brands include Ameridrives, Bauer Gear Motor,

Bibby Turboflex, Boston Gear, Delroyd Worm Gear, Formsprag
Clutch, Guardian Couplings, Huco, Industrial Clutch, Inertia
Dynamics, Kilian, Lamiflex Couplings, Marland Clutch, Matrix,
Nuttall Gear, Stieber Clutch, Stromag, Svendborg Brakes, TB
Wood’s, Twiflex, Warner Electric, Warner Linear and Wichita Clutch.

US (Application Assistance)
+1-814-480-5095
www.ameridrives.com

Asia Pacific
For a list of our AP sales offices:
www.AltraMotion.com/ContactUs

P-8456-AC 10/17