A Novel Gear Test Rig with Adjustable

Shaft Compliance and Misalignments

Part I: Design

A. Palermo, J. Anthonis, D. Mundo and W. Desmet

Abstract This paper describes the design aspects for a new gear test rig aimed at
adjusting the influence of shaft compliance on gear meshing stiffness, while allowing
the operator to impose gear misalignments. Static and dynamic testing is possible for
the most important gear-related physical quantities: Transmission Error, relative
displacements, tooth root strain, transmitted torque. The discussed test rig has a
mechanical power circulation arrangement, where two sides can be identified. One
side is dedicated to testing a cylindrical gear pair; the other side is needed for retaining
a torque preload in the system by means of a second gear pair. Relative misalignment
can be imposed between the test gears in the five possible degrees of freedom (three
parallel misalignments, two angular misalignments). Shaft compliance can be
adjusted by setting the axial position of the gears before fastening them to the shafts.

 
Keywords Angular Misalignment Gear Test Rig Lead Modifications Parallel 
  
Misalignment Profile Modifications Shaft Stiffness Vibration Isolation

1 Introduction

Experimental dynamic characterisation of mechanical transmissions requires

measurement of vibrations related to motion or forces in response to excitations.
Gear test rigs allow overcoming the limitations related to typically poor

A. Palermo (&)  D. Mundo

Department of Mechanical Engineering, University of Calabria, Ponte Pietro Bucci 87036
Rende, Italy
e-mail: palermo.antonio@gmail.com
J. Anthonis
LMS International, Interleuvenlaan 68 B-3001 Leuven, Belgium
A. Palermo  W. Desmet
Department Mechanical Engineering, K.U. Leuven, Celestijnenlaan 300 B B-3001 Heverlee,
Belgium

G. Dalpiaz et al. (eds.), Advances in Condition Monitoring of Machinery 497

in Non-Stationary Operations, Lecture Notes in Mechanical Engineering,
DOI: 10.1007/978-3-642-39348-8_43, Ó Springer-Verlag Berlin Heidelberg 2014
498 A. Palermo et al.

accessibility of gears in service and offer the chance of heavily instrumenting the
gears to seek validation of numerical gear meshing models.
Considering mechanical power, gear test rigs described in literature can be
divided in two main categories: power absorption and power circulation. In both
cases, electric machines are typically used to power the system thanks to their
flexibility in speed and torque ranges and thanks to their suitability for control units.
For a cylindrical gear system with one input shaft and one output shaft, power
absorption test rigs rely on a motor which provides input power and a generator
which absorbs the output power. Power circulation test rigs are self-balanced
systems and only one motor is required to overcome losses to spin the system.
Design choices are discussed in the next paragraphs for a power circulation test
rig for parallel cylindrical gear pairs.

2 Objectives and Global Design

Objectives for the discussed gear test rig are the static and dynamic measurement of
typical gear-related physical quantities, under imposed conditions of misalignment
and with a given degree of influence of shaft compliance on gear meshing stiffness.
One of the final goals is to experimentally validate the gear multibody element
proposed in [1]. Focus for the experimental campaign is on a gear pair belonging to
families of spur or helical cylindrical gears. Both gear families have different
amounts of profile and lead microgeometry modifications for the tooth surfaces
(Sect. 4). Parallel or angular misalignments, can be imposed along the five degrees
of freedom for relative orientation of the gears (Sect. 5). Shaft torsional and bending
stiffness are by nature placed in series with gear mesh stiffness and affect the total
stiffness of the gearing. The ratio between the shaft stiffnesses and the gear pair
stiffness can be adjusted to impose which of the two elements will dominate the
dynamic response (Sect. 6). All the tests can be performed in a range of speed, for a

Fig. 1 Test rig three-dimensional representation. 1. Test gears; 2. Reaction gears; 3. Bearings
support plates; 4. Flexible couplings; 5. Flywheels; 6. Clutch flange for preload
A Novel Gear Test Rig 499

Table 1 Test rig specifications

Parameter Range Uncertainty
Speed 0–4,500 rpm (0–75 Hz) Measured
Torque 0–500 Nm ± 0.05 %
Angular misalignments 0–2 mrad 0.1 mrad
Parallel misalignments 0–0.3 mm 0.020 mm

set of transmitted torque values. The CAD model of the test rig is represented in
Fig. 1, while its specifications are reported in Table 1.
The test rig is mounted on a concrete base suspended on air springs which
fulfils three functions. First, it provides reference pin holes to align the test side
and the reaction side. Second, it provides rigid anchoring islands for the two sides
and isolates the overall system. Third, it separates the two sides to avoid propa-
gation of stray vibrations from the reaction side to the test side; such vibrations are
damped through the concrete. A temperature-controlled pressurized lubrication
system feeds oil to the test and reaction side bearings and gears. Bearings are
lubricated through channels leading to the lubrication groove in their external race.
Gears are lubricated by oil jet. A Plexiglas enclosure collects lubrication oil for the
test gears which is returned to the lubrication circuit. Safety grids, acting also as
safety retention elements for the flywheels, cover the rotating components between
the test and the reaction side.
The quantities to be measured are: gear Transmission Error, speed, torque, tooth
root strain along the axial direction, relative displacements of the gear bodies, shaft
displacements and imposed misalignments. Detailed discussion of the measure-
ments and the related instrumentation is addressed in the second part of this paper.
Speed can be set by means of a 9 kW asynchronous motor controlled by a
frequency inverter. Each torque value is imposed as a preload in the system by
using a torque wrench and a set of weights, executing the following sequence: the
flywheel connected to the clutch is fixed to the base by using a pin; the clutch is
released untightening the fastening screws; the torque wrench is fitted on the clutch
disk and the required weights are hung; the fastening screws are tightened at the
prescribed torque; the weights and the torque wrench are removed. Shafts are
verified at the maximum gear torque of 500 Nm and provide a static safety factor
of at least 4 to yield point, using the Von Mises criterion, considering combined
bending moment and torsion at the worst locations. Materials are a ductile med-
ium-carbon C45 steel (360 MPa yield stress) for the massive parts and a low-alloy
1.2312 steel (820 MPa yield stress) for shafts.
Bearings (Fig. 2) are chosen to maximise stiffness and their rating provides
infinite service life under maximum loading conditions. In particular, high-preci-
sion spherical roller bearings are chosen to support test gears; more details on such
bearings are provided in Par. 5 and 6. For improving turning accuracy, a conical
adapter sleeve is used to adjust inner race interference and bearing clearance. Light
interference fit is used for the external race to improve turning accuracy. Both
bearings are in a locating configuration and require careful tuning of their relative
500 A. Palermo et al.

Fig. 2 Bearings arrangement for one shaft branch of the test rig and average bearing stiffness in
the loading range. 1. High-precision spherical roller bearing (1 9 109 N/m); 2. Y-bearing unit; 3.
Wide-face single-row cylindrical roller bearing (6 9 108 N/m); 4. Double-row tapered roller
bearing (1 9 109 N/m radial, 3 9 108 N/m axial)

distance. Bearings for the reaction side are chosen to support flywheels and
reaction gears. Radial loads from the flywheels are supported by Y-bearing units
and by wide-face single-row cylindrical roller bearings. Y-bearings are chosen
because they do not require precise alignment of their supporting base; cylindrical
roller bearings are chosen with flanges on both races and are placed in a non-
locating configuration allowing light clearance fit on the external race (which is
stationary with respect to load). The same cylindrical bearings support radial loads
originating from reaction gears, together with double-row tapered roller bearings
in a locating configuration. The locating configuration is achieved using a housing
shoulder on one side and a bearing cap with laminated shim on the opposite side.
Tapered rollers allow absorbing thrust originated by the reaction gears.

3 Vibration Isolation Between Test and Reaction Sides

Vibration isolation between the two sides is achieved by decreasing torsional

stiffness and increasing rotary inertia for the reaction side. The overall effect is to
obtain the lowest possible rotational natural frequencies for the reaction side, so
that the reaction gears meshing excitation can be filtered out as soon as possible.
Torsional stiffness is decreased by using flexible couplings and by increasing shaft
lengths. When preloading the system, low torsional stiffness implies high twist
angle (1.5̊ at 500 Nm). This condition ensures that rotational vibrations generated
by the test gears (hundredths of degree) do not influence the torque preload, since
the latter comes from an elastic deformation. Furthermore, flexible couplings
prevent non-torsional loading between the two sides, allowing eccentricity up to
0.28 mm and angular misalignment up to 35 mrad between the two sides.
Analytical equations for a rotational model (Fig. 3) were solved to obtain the
rotational natural frequencies along with their related mode shapes. Four degrees
of freedom were used in total, allowing rotation for each gear.
The term k12 represents the equivalent torsional stiffness for the series between
the test gear pair mesh stiffness (equal to 1:6  106 Nm=rad), the two shafts torsional
stiffness and the two (adjustable, Par. 6) shafts bending stiffness. The term k34
represents the torsional stiffness for the reaction gear pair (equal to
1:8  106 Nm=rad), since supporting shafts there are very short. The terms k13 and
A Novel Gear Test Rig 501

Fig. 3 Illustration of the 4-DOF rotational model of the test rig

Fig. 4 Histogram representation of mode shapes for the 4-DOF rotational model of the test rig

k24 represent the torsional stiffness of the flexible couplings. The terms Ji represent
the moments of inertia around the rotation axis for each gear plus the related shafts.
Moments of inertia for the test side (J1 ¼ 0:0218 Kg  m2 , J2 ¼ 0:0118 Kg  m2 ) are
two orders of magnitude lower than the ones for the reaction side
(J3 ¼ J4 ¼ 0:2 Kg  m2 ). The obtained rotational mode shapes and their natural
frequencies are illustrated in Fig. 4.
The first 0 Hz natural frequency is due to the allowed rotational degree of
freedom. The second mode at 280 Hz involves in-phase rotation for both the test
and the reaction gear pairs with opposite rotations for the two sides. Therefore, this
mode involves minimal mesh deflections in proportion to the deformation of the
flexible couplings. The third mode at 680 Hz involves reaction gears with out-of-
phase rotation (namely mesh deflection) and very limited motion of the test gears
(3 % of the reaction gears modal amplitude). The fourth mode at 1,968 Hz
involves test gears similarly with very limited motion of reaction gears (0.2 % of
the test gears modal amplitude). The third and fourth represent modes for the
reaction gear pair and the test gear pair respectively. Limited motion of the test
gears at a resonance for the reaction gears, and vice versa, is a first evidence of
decoupling between the two sides.
The same model was used to obtain the forced-response Transmission Error for
the test gear pair, applying twice the maximum variable contact force expected at
the reaction side. This quantity allows estimating how much of the reaction gears
excitation is able to reach the test gears. A typical value of 5 % was chosen for
gears modal damping ratio [2]. Results in Fig. 5 show an induced Transmission
Error higher than the encoder measurement resolution only in correspondence of
gear resonances. Resonance conditions for the reaction gears must therefore be
502 A. Palermo et al.

Fig. 5 Forced-response transmission error for the test gear pair when exciting the reaction gears
with twice the maximum expected dynamic contact force

avoided. Resonance conditions for the test gears do not represent a problem, since
the Transmission Error amplifies and a decrease in accuracy can be allowed (See
Part II: Instrumentation).

4 Gear Design

Gear design is performed following different purposes for the test side and the
reaction side, which have specifications reported in Table 2. Different tooth
numbers are chosen for test and reaction gears to be able to distinguish excitation
belonging to each gear pair.
Main requirements for test gears are identified in:
1. Providing very repeatable meshing excitation;
2. Possessing a gear mesh stiffness calculated to determine a desired;
3. gear pair resonance frequency;
4. Allowing room for mounting root strain gauges;
5. Not failing under load.

Table 2 Gear specifications

Test gears (Spur) Reaction gsears
Number of teeth 57 64
Normal module 2.60 mm 2.25 mm
Normal pressure angle 20 deg 14 deg
Helix angle at theoretical pitch circle 0 deg 13 deg
Tip diameter 154.50 mm 155.00 mm
Root diameter 141.70 mm 142.00 mm
Face width 23 mm 41 mm
Normal circular tooth thickness at theoretical pitch circle 3.780 mm 3.074 mm
Total contact ratio 1.456 3.0
Average gear mesh stiffness along the line of action 3.3 9 108 N/m 3.5 9 108 N/m
A Novel Gear Test Rig 503

(a) (b)

STE
0 Nm

500 Nm

Fig. 6 a Static transmission error curves from 0 to 500 Nm for a test spur gear pair having
10 lm profile crowning modification. b Example of sub-optimal (5 lm), optimal (10 lm) and
above-optimal (10 ? 5 lm) profile modifications for minimizing peak to peak static transmission
error, with respect to the median applied load. Simulations performed using LDP [6]

Repeatability is substantially enhanced by two choices: first, the gear ratio is

equal to one; second, gears are precision ground to ISO 1328 Quality 3. Unity gear
ratio implies that one tooth will always mesh with the same tooth of the mating test
gear. This excludes deviations introduced by manufacturing variability. Tolerances
for quality ISO 3 were outperformed in the manufacturing execution, with typical
values within ±1 micron for both lead and profile. All the teeth for all the gears
were measured on both flanks along the mid-plane profile and the operating pitch
circle lead to ensure tolerance quality. To achieve a stable and durable tooth
surface, before grinding, gears were made of 16 MnCr5 alloy special steel and
case-hardened to 60 HRC until 0.8 mm depth. The value reached for test gear
mesh stiffness, combined with shaft stiffness, provides a gear pair resonance fre-
quency which can be adjusted in the range described in Par. 6. Mesh stiffness is
tuned mainly by selecting appropriate tip and root circles diameters, operating
centre distance and tooth thickness. Tip and root diameters and centre distance are
also chosen to leave sufficient room for the installation of the root strain gauges.
Both families of spur and helical test gears are composed of two unmodified
gears, two with parabolic profile modifications and two with parabolic lead
modifications, which can be combined to achieve extensive validation of mi-
crogeometry effects. One modification value is designed to be sub-optimal for the
median applied load, the other value to be optimal and the combination to be
above-optimal (Fig. 6).
Main requirements for reaction gears are identified in:
1. Generating the lowest possible meshing excitation;
2. Being insensitive to small misalignments;
504 A. Palermo et al.

3. Providing the lowest possible mesh stiffness;

4. Having long service life.
Meshing excitation is limited first by using helical gears with high and integer
total contact ratio (the latter property was verified under load and including mi-
crogeometry effects). Subsequently, profile modifications are optimized to minimize
gear transmission error peak to peak value. Sensitivity to angular misalignment is
reduced by applying end relief and parabolic lead crowning to the teeth surfaces.
Mesh stiffness is reduced (to lower the gear pair resonance frequency) by reducing
root circle diameter, tooth thickness and normal pressure angle.
Gear strength was verified using three software tools: Ansol Helical 3D [3] for a
Finite Element full-field verification, OSU Gearlab LDP [2] for a confirmation and
KISSsoft [4] for rating according to the norm ISO 6336. Ratings are provided
hereafter for the already manufactured family of spur gears and for reaction gears.
Helical gears have been designed, but will be manufactured after experimentally
verifying the design for the spur family. According to the ISO 6336 norm, spur
gears have high static safety factor (3.7 against yield at tooth root), but short
service life (8.5 h at maximum speed and load) due to sharp handover between
tooth pairs. This compromise is reached for test gears to verify the above-men-
tioned requirements, since test gears are appointed to operate for a limited time
and then to be changed. Reaction gears have high static safety factor (5.3 against
yield at tooth root) and infinite service life at maximum speed and load. Stresses
were lowered for reaction gears by increasing the facewidth value. Higher face-
width value provides the additional advantage of increasing the total contact ratio,
but also the disadvantage of increasing the mesh stiffness. The trade-off is bal-
anced by tuning the helix angle.

5 Setting Misalignments

Angular misalignment in an arbitrary plane and parallel misalignment in the

transverse plane can be set by a technique proposed by [5]. Bearings are mounted
in eccentric caps; these caps can be turned to orient the eccentricity at a given
angle. Same angles for the two caps on the same shaft result in imposing a parallel
misalignment; opposite angles result in imposing an angular misalignment. Angles
in-between result in a combination of angular and parallel misalignment. Angular
misalignments motivate the compulsory need of spherical roller bearings. Axial
parallel misalignment can be imposed thanks to expansion locking units to fasten
the gears on the shafts. Each unit is composed of two self-centring tapered conical
elements with cylindrical bore for the shaft and cylindrical outer surface for the
gear. Gears can first be axially displaced of a known quantity by using a calibrated
thickness and then fastened to the shafts.
Given the significant sensitivity of gears to small amounts of misalignments,
manufacturing and assembly tolerances acquire paramount importance. Geometric
A Novel Gear Test Rig 505

tolerance constraints were calculated quantitatively and will be discussed here only
qualitatively for the sake of conciseness. Holes in bearings support plates require
tight positioning tolerance to avoid initial misalignments and cylindricity to centre
bearing caps; cylindrical surfaces for bearing caps require tight cylindricity to
centre bearings and accurate eccentricity to set misalignment; bearings and shafts
require tight turning accuracy with very limited radial runout. Alignment and
cylindricity for bearing support plates holes are achieved by assembling the plates
and finishing rough holes with a single operation of wire electrical discharged
machining (EDM). Wire EDM is also used to manufacture bearing caps. High
precision bearings are selected to restrict to 30 % the normal amount of runout for
the inner race (Kia parameter). High precision pre-ground and hardened shafts are
selected for minimized radial runout and improved cylindricity.

6 Adjusting Shaft Compliance

Expansion locking units are used to allow axial positioning of the gears and thus to
adjust shaft stiffness. In particular, bending and torsional stiffness are maximum
when the loaded shaft segment is shortest (i.e. the test gears are closest to the
bearings within the loaded path). Two configurations will be used: gears at half
shaft length and gears closest to the bearings. In the maximum shaft stiffness
configuration, bending and torsional shaft stiffness values are substantially higher
than the mesh stiffness. In the minimum shaft stiffness configuration, bending and
torsional shaft stiffness values are substantially lower than the mesh stiffness.
Using the beam theory of elasticity, bending and torsional stiffness assume values
reported in Fig. 7. Corresponding resonance frequencies for the test gear pair are
expected to be close to 1,968 Hz for maximum shaft stiffness and close to 800 Hz
for minimum shaft stiffness.
Two additional aspects need to be clarified: the effects of bearing stiffness on
test gears resonance frequency and on load-dependent misalignment. Average
bearing stiffness in the loading range is three times the mesh stiffness for the test

Fig. 7 Comparison of gear mesh stiffness with shaft bending and torsional stiffness with respect
to the axial position of the gears
506 A. Palermo et al.

gears. Load dependent misalignment can be induced only in the minimum shaft
stiffness configuration. In the maximum shaft stiffness configuration gears are
equidistant from bearings and rotation induced by bending is zero. In the minimum
shaft stiffness configuration, the closest bearing absorbs almost the complete
meshing loads, while the opposite bearing is virtually unloaded. Bearing deflec-
tions at maximum load are below 10 microns, leading to a maximum misalignment
of 0.03 mrad (three times below uncertainty on imposed misalignment).

7 Conclusions

The main steps of the design process for a cylindrical gear pair test rig have been
described. Solutions already available in literature have been used to impose
angular misalignment in an arbitrary plane and parallel misalignment in a trans-
verse plane. A new solution has been proposed to impose axial misalignment and to
select the influence of shaft compliance on gear mesh stiffness. The use of flexible
couplings and flywheels has been discussed to isolate the test side from stray
excitation belonging to the reaction side. The main criteria identified for gear design
have been reported for the test and the reaction side. A 4-degrees-of-freedom
rotational model has been used to estimate rotational resonance frequencies of the
test rig. Contributions on mesh stiffness coming from shafts have been considered
for test gears. Special care has been paid to tolerances, for ensuring repeatability of
measurements and accurate values of imposed misalignments.

Acknowledgments The authors gratefully acknowledge the IWT (Agency for Innovation by
Science and Technology in Flanders) for the financial support through the project HEV-NVH
(Agr. 110360).
The authors are also grateful to Dr. Giorgio Bonori, Italian distributor for Ansol LLC, for
providing free academic license and support for the software Transmission 3D. Similar
acknowledgement goes to Mr. Ivan Saltini, Italian distributor for KiSSsoft AG, for providing free
academic license for the software KISSsoft.

References

1. Palermo A, Mundo D, Hadjit R, Desmet W (2013) Multibody element for spur and helical gear
meshing based on detailed three-dimensional contact calculations. Mech Mach Theo 62:13–30
2. Parker RG, Vijayakar SM, Imajo T (2000) Non-linear dynamic response of a spur gear pair:
modelling and experimental comparisons. J Sound Vib 237:435–455
3. Parker RG, Agashe V, Vijayakar SM (2000) Dynamic response of a planetary gear system
using a finite element/contact mechanics model. J Mech Des 122:304–310
4. Contact Analysis in the Cylindrical Gear Calculation (2011) KissSoft AG, http://
www.kisssoft.ch/english/downloads/
5. Hotait M, Kahraman A (2008) Experiments on root stresses of helical gears with lead crown
and misalignments. J Mech Des 130:1–5
6. Load Distribution Program (2002) Users Manual, Ohio State University, GearLab