On the influence of the housing stiffness of

compact industrial gearboxes on the gear

mesh load distribution

A case study considering spiral bevel and cylindrical gears

Dipl.-Ing. Jürg Langhart and Dr.-Ing. Ioannis Zotos, KISSsoft AG

Summary
The housing stiffness does influence the bearing and therefore shaft positions and shaft alignments
once the gearbox and therefore the housing is subjected to loads. This in itself however does not yet
allow for an assessment of the performance of the gears in terms of e.g. load distribution or vibration.
Further to the shaft misalignment caused by the housing deformation, the shaft and bearing
deformations as well as gear manufacturing tolerances influence the meshing conditions. In the
following study, the different influences are studied and compared to identify of greater relevance.
Note that the findings presented here are valid for compact industrial gearboxes and the findings may
be different if different applications or housing designs were considered.

Figure 1: Bevel helical gearbox modelled in KISSsys release 03-2016 with housing data (CAD file)
imported.
1 Application example
Let us consider a bevel helical gearbox, currently in production by KISSLING AG of Switzerland.
Nominal power is 600 kW and the gearbox is operated in a temperature of 45°C to 60°C, with peaks
up to 80°C. The high speed side gears and shafts may display slightly higher temperatures in
operation. The cast housing is made of EN-GJS 40.3 and features a number of ribs to make the
housing lightweight yet stiff. Considered in this study are the bevel gear stage and the parallel shaft
stages. The parallel stages are achieved using helical gears with normal module mn = 8 mm
(intermediate stage) and mn = 12 mm (output stage). The gears are ground after case carburizing to a
quality grade of Q=6 along ISO 1328 or better. The shafts are supported by two spherical roller
bearings each.

The spiral bevel gear stage is a having a ratio of 17:37 with a gear outer diameter of 460 mm,
manufactured as a Zyklo-Palloid® set. The pinion shaft is supported by a paired taper roller bearing
and a spherical roller bearing while the gear shaft is supported by two spherical roller bearings.

2 Procedure
It is of interest to find out the influence of different factors on the required gear modifications. The gear
modifications are then assessed using a numerical LTCA (loaded tooth contact analysis) as well as
along ISO 6336-1, Annex E (for the calculation of the line load along the face width).

The following factors are considered:

 Influence of shaft and non-linear bearing deformation
 Influence of manufacturing tolerances (of gears and of housing)
 Influence of housing deformation
 Influence of temperature at 80°C

A realistic modeling of the shaft displacement and the shaft deformation (which then leads to a gear
misalignment) in radial and axial direction is required to calculate the gaping in the gear mesh along
ISO 6336-1, Annex E and the resulting contact stress distribution / transmission error. The axial
component of the gear misalignment is not so relevant for parallel shaft gears but for the bevel gear
stage as this influences the H displacement (for the pinion) / the J displacement (for the gear). For
bevel gears, thermal growth and axial bearing deformation are most relevant and require attention.

Calculations are carried out using KISSsoft / KISSsys release 03-2016, using commercial modules
only.

3 Investigation of the parallel shaft stages

3.1 Methods for the assessment of tooth contact

A tooth contact assessment should be done in two steps. Firstly, a line load distribution along the face
width, based on ISO 6336-1, Annex E calculation should be performed. Then, a holistic assessment
of the contact force in the whole plane of action using a LTCA (loaded tooth contact analysis) should
follow.
For the assessment of the lead-wise load distribution, the calculation of the line load w [N/mm] along
ISO 6336-1, Annex E [1], is recommended. It yields realistic values for the load distribution, a KHβ
value needed for the strength rating and consumes little time to perform. This method may be
considered a one dimensional LTCA and has been described in the context of its implementation in
KISSsoft before [2].

A matter of discussion is whether load factors like KA or Kv should be considered in the calculation of
the line load distribution. This is currently (second half of 2016) being discussed and reviewed in the
respective ISO committee (TC 60). In KISSsoft, the educated user may choose how to consider
additional load factors in the line load calculation in the settings [3].

A more refined method is the two dimensional LTCA where also the effective contact ratio under load
is considered [4]. It also yields results like transmission error TE, peak to peak transmission PPTE,
local power losses due to friction, local contact temperature, local lubricant film thickness / micropitting
risks values and others. The two dimensional LTCA also yields a more accurate line load distribution
compared to the ISO 6336-1, Annex E method as it does consider the true contact ratio which is
affected by the tooth deformation (this effect is not considered in the one dimensional LTCA along ISO
6665-1, Annex E calculation where the theoretical contact ratio is considered only).

It is therefore obvious that the two above methods may yield different results for the load distribution
and may therefore result in different required gear modifications. Once however good modifications
are applied, the differences are small. It is hence recommended to first design modifications based on
the findings elaborated using the ISO 6336-1, Annex E one dimensional LTCA (which is very fast to
use) and then to use – for a final check and for the profile modifications – the two dimensional LTCA
(Figure 2).

Figure 2: Calculation f the line load wb [N/mm] along the gear face width on the operating pitch cylinder. Left:
Using one dimensional LTCA along ISO 6336-1, Annex E. Right: Using two dimensional LTCA as
implemented in KISSsoft using numerical approach.

Key
“Linienlast” = line load
“Breite”= face width
3.2 Influence of shaft and bearing deformation

The below investigations are documented for the output stage. There, the shaft deformation influence
is studied first. The shaft deformation is a function of shear, torsion and bending loads and is
calculated with a semi-analytical / semi-FEM method using a Timoshenko beam model. Then, the
non-linear bearing stiffness is considered by looking at each contact between a rolling element and the
races along [5]. The compression of the rolling elements results in a movement of the inner ring with
respect to the outer ring. In this, also the bearing operating clearance (as a function of bearing
clearance, fit on shaft and in housing, thermal expansion and centrifugal effects) is considered. The
housing deformation is not yet considered in order to be able to separate the different influences.

A proposal for an optimal lead modification is provided by KISSsoft based on an algorithm along ISO
6336-1, Annex E. For the parallel shaft output stage of the example gearbox shown above, a helix
angle modification of cHβ = -11 μm and a crowning of cβ = 19 μm is calculated (Figure 3).

Figure 3: Erforderliche Breitenkorrekturen aufgrund der Wellendurchbiegung und Lagernachgiebigkeiten

Key:
„erforderliche Korrekturen“ = required modifications
„Korrektur“ = modification
„Schrägungswinkel“ = helix angle (modification)
„Balligkeit“ = (lead) crowning

The line load distribution wb with the above mentioned influences is shown in Figure 4, left. In Figure
4, right, the resulting line load distribution wb with the above mentioned influences and modifications
applied is shown in comparison. The load distribution factor KHβ drops from KHβ=1.17 (without
modifications) to KHβ=1.01 (with modifications) confirming the suitability of the modifications
automatically proposed by KISSsoft.
Figure 4: Line load distribution calculated using ISO 6336-1, Annex E. Left: Considering shaft and bearing
deformation but without modifications applied. Right: Considering shaft and bearing deformation and with
modifications applied.

Key
“Linienlast” = line load (distribution)
“Breite” = (face) width

3.3 Influence of manufacturing tolerances

Further to the shaft and bearing deformation, the influence of manufacturing errors are to be
considered. Said errors may in fact be the governing factor when determining the gear misalignment
and suitable modifications to compensate those, they must therefore not be neglected. In general two
groups of tolerances are considered

 Helix slope deviations f H of the gears

 Shaft inclination and deviation errors f,f

Helix slope deviations along ISO 1328 [6] are a measure related to the gear quality and are influenced
by the manufacturing process, Figure 5. For the gearbox considered here, for the output stage with
mn=12 mm, the permissible helix slope deviation is fH =15 μm for the pinion and fH=16 μm for the
wheel. In KISSsoft, these two values may be considered in a worst case scenario (resulting in a total
misalignemnt of the flanks of 31 μm) or in a statistical approach such that 99.7% of meshes are
covered, resulting in 22 μm misalignment. The later approach is used here.

Figure 5: Definitions of flank line deviations along ISO 1328

Shaft alignment tolerances along ISO/TR 10064-3 [6] define persmissible deviation and inclination
errors with respect to the bearing span or gear face width (Figure 6).

Figure 6: Definition of deviation and inclination error along ISO/TR 10064-3

ISO/TR 10064-3 determines permissible values for the above deviations from a permissible flank line
error Fβ, considering the bearing span L and the common gear face width b as follows:

 L
f   0.5  F
b
f  2 f

The above components are combined to represent a misalignment in the plane of action, fΣ:

1
f   f   f 
3

Values for the bearing span L, common face width b and permissible flank line deviation Fβ are
calculated based on the drawing data yielding a deviation error of fΣβ = 30 μm and fΣδ = 59 μm. The
resulting misalignment in the plane of action is then fΣ = 50 μm. Again, this is a worst case approach
while a statistical approach yields a value of fΣ = 33 μm, the later is used.

Alternatively to the tolerances as per ISO/TR 10064-3, position tolerances from the drawings (or
measured values) may be used, Figure 7.
Figure 7: Determining the shaft misalignment tolerances from housing tolerances as given on the housing
drawing.

To compensate manufacturing errors, the use of helix angle modifications is not suitable as those
errors are random by nature. To compensate random errors, either a lead crowning or a symmetrical
end relief is required. The optimal lead crowning value is calculated by defining a range defining a
lower and upper limit and a step size. The different crowning values are then combined with the
above, constant helix angle modification of cHβ = -11 μm. In our example, ten different crowning values
(minimum at 19 μm, maximum at 55 μm) are created and combined with the constant helix angle
modification.

Figure 8: Combining a constant helix angle modification with a varying crowning in the KISSsoft modifications
sizing function.

Key
„Nr.“ = no. “Art der Korrektur” = type of modification
„Zahnrad“ = gear “Schrägungswinkelkorrektur” = helix angle modification
„Rad“ = gear no. “Betrag” = value
“Flanke” = flank “Faktor” = factor
“beide” = both “Synchronisiere mit” = synchronize with
“Breitenballigkeit” = lead crowning „Anzahl Schritte“ = no. of steps
From the above calculations, we find a face load factor for each of the ten solutions (solution no. 0 is
the one without any modifications for comparison), we find that the face load factor KHβ is minimized if
the lead crowning is set to cHβ = 43 μm, Figure 9.

Figure 9: Resulting KHβ and gear life as a function of the crowning.

Key
“Breitenlastfaktor” = face load factor
“Rad” = gear (wheel)
“Betrag” = amount
“Minimale Lebensdauer” = minimum lifetime

A final check with these modifications, considering all combinations of the flank line error fH and the
shaft misalignment error fma (+fma +fHβ, +fma -fHβ, -fma +fHβ, -fma -fHβ) shows that the
modifications are most suitable to compensate both shaft / bearing deformations as well as random
manufacturing errors. Under all conditions, dangerous edge loading is avoided, Figure 10.

Figure 10: Line load distribution considering shaft / bearing deformation, modifications and different combinations
of gear manufacturing errors and housing manufacturing errors.
We conclude that the required modifications are c Hβ = -11 μm and cHβ = 43 μm. In below Figure 11,
these values are shown together with the crowning of cHβ = 24 μm reqruied to compensate
manufacturing tolerances.

Figure 11: Required lead modifications in total and crowning required to compensate random manufacturing
errors / tolerances.

Key
“erforderliche Korrekturen” = required modifications (corrections)
“Korrektur” = modification (correction)
“Schrägungswinkel” = helix angle (modification)
“Balligkeit” = (lead) crowning
“Balligkeit für tolerances” = crowning for (compensating) tolerances

The resulting face load factor increases to KHβ=1.07 to KHβ=1.16 (depending on how the random
manufacturing errors are combined). Both values are acceptable considering that the resulting design
is “robust”, meaning, it is suitable for any combination of random manufacturing errors within the
specified tolerances.

3.4 Influence of the housing deformation

Bearing forces act on the housing and result in a deformation thereof, resulting in the bearing (outer)
ring to be displaced with respect to its original position.

The deformation of the housing is calculated inside KISSsys by using a reduced stiffness matrix of the
housing and the above mentioned bearing forces (compiled into a bearing force vector). The housing
stiffness matrix is exported from the FEM software (ABAQUS, ANSYS or NASTRAN).

An iterative calculation is required as the deformed housing influences the shaft deformation which
again influences the bearing forces (which again influences the resulting housing deformation), Figure
12. The approach as implemented in KISSsys has been presented earlier in [8].
Figure 12: Iterative calculation of the housing deformation due to the bearing forces applied.

Key
“statische Wellenberechnung” = static shaft calculation “Ende” = end
“Lagerkräfte” = bearing forces “Ja” = yes
“Berechnung Gehäusedeformation” = calculation (of) housing deformation “Nein” = no
“neue Lagerposition” = new bearing position
“Sind Lagerkräfte und Positionen unverändert” = are bearing forces and positions unchanged

3.5 Relevant details of the housing stiffness matrix

In the FEM model of the housing, the bearing centers need to be modeled with a single node,
connected to the surrounding bearing seat using e.g. RBE2 or RBE3 type elements (depending on the
FEM code used). A certain error due to the modeling is to be accepted also because radial and axial
forces may in reality be supported by different surfaces.

In the gearbox shown here, the shafts of the two cylindrical gear stages are supported by spherical
roller bearings. Their inner ring is using a p6 fit on the shaft and a J6 fit on in the housing. The
transition fit in the housing will result in the axial loads being transmitted by the shoulder shown below
(red path) while radial loads are transmitted directly into the housing walls (green path).

Figure 13: Left: Axial forces and radial forces acting on different planets on the housing. Right: Assumption made
in the FEM model, axial and radial load act both on the same surface.
In the FEM model, the spherical roller bearing is supported only on the radial surface, resulting in a
somewhat to stiff model for the axial forces. However, as the axial movement of the bearing (due to
the axial deformation of the housing) is not having a significant influence on the mesh of cylindrical
gears, this modelling approach is acceptable. After importing the reduced stiffness matrix into
KISSsys, the matrix is aligned with the KISSsys model by matching the centre of the bearings in the
KISSsys model with the nodes located at the bearing centers in the FEM model. Then, the kinematic
calculation is executed in KISSsys, the bearing forces are calculated and the bearing force vector is
multiplied with the housing stiffness matrix giving a bearing displacement vector.

Figure 14: Left: FEM model of the gearbox housing. Stifness matrix and housing CAD model imported in KISSsys.

Before assessing the influence of the housing deformation on the gear meshes, let us look at the
resulting housing deformations as calculated in KISSsys for the shafts no. 3 and no. 4 (output shaft).
Figure 15 shows the total radial displacement of the shaft without (blue) and with consideration of the
housing deformation (red). In both cases, the movement of the bearing inner ring with respect to the
outer ring due to bearing operating clearance and deformation is considered. We observe that the
change in shaft displacement is less than 10 μm when we also consider the housing deformation.

Figure 15: Radial displacement of the shafts due to bearing operating clearance and deformation and with /
without housing deformation.
Key
„Wellenverlagerung mit und ohne Gehäusesteifigkeit“ = shaft displacement with and without housing stiffness
„Verschiebung“ = displacement „ohne“ = without
„Welle“ = shaft „mit“ = with
„Lager“ =bearing „Gehäusesteifigkeit“ = housing stiffness

As the change in shaft displacement is small once the housing deformation is considered, the required
change in the modifications will also be small. In the example considered, the required change in helix
angle modification is only ΔcHβ = -2 μm to compensate the change in the gear alignment. This change
has no practical relevance and is well within any manufacturing tolerance. The required total helix
angle modification is now cHβ = -13 μm and cβ = 43 μm, Figure 16.

Figure 16: Required lead modifications in total and due to the housing deformation alone.

Key
“erforderliche Breitenkorrektur” = required lead corrections (modifications)
“Korrektur” = correction (modification)
“Schrägungswinkel” = helix angle (modification
“Balligkeit” = (lead) crowning
“Schrägungswinkel für Gehäuse” = helix angle (modification) for (compensation of) housing (deformation)

Finally, we may conclude that for the output stage, the influence of the housing deformation is not
significant in this example where the housing was well design and is hence stiff.

4 Bevel stage
4.1 Introduction to bevel gear misalignment

Due to the gear mesh force direction governed by shaft offset, pressure angle, pitch cone angle and
helix angle (mean spiral angle), the bevel gears will be misaligned. This misalignment is obviously also
a function of the support of the gear shafts. The resulting deformations are typically defined along
Klingelnberg [9] using VHJ displacement conventions as shown below. Other conventions exist but
are not used in this example, Figure 17.

Figure 17: Forces acting on the bevel gears along ISO/TR 22849 and definition of VHJ displacements as per
Klingenlberg definition

The total of the VHJ displacements are formed from the respective displacement components of
pinion and wheel. For the H and J displacement, the axial movement of the gear shaft is relevant. This
means that – opposed to the cylindrical gear stages – the axial movement of the shafts is relevant. A
detailled modelling and calcualtion thereof is therefore required. The values for VHJ displacements are
calculated in KISSsoft automatically from the deformations and displacements of the pinion and wheel
shaft.

The VHJ displacements are also used to explain shifts in the contact patterns of bevel gears. In our
example, a LH bevel gear set is used. If the drive flank is loaded, a upward movement of the pinion
will result, this again results in the contact pattern moving towards the heel, Figure 18.

Figure 18: Contact pattern shift on the RH wheel, for V and H displacements

In order to avoid a edge loading due to the shift in contact pattern, a longitudinal crowning is applied.
Furthermore, the spiral angle is modified. Further modifications like Bias are not discussed here.
4.2 Methods for assessing the contact patterns

In the strength rating method ISO 10300 [10] for bevel gears, the load distribution as calculated is not
directly considered. The load distribution is considered in a simplified manner only, giving rough
values for KHβ depending on the type of shaft supports (e.g. overhung gears will result in a higher
vlaue for KHβ ). Hence, in the bevel gear strength rating, there is no equivalent to the detailed line load
distribution calculation as per ISO 6336-1, Annex E as available for cylindrical gears.

The assessment of the contact pattern for bevel gears is typically done after a test run. The contact
pattern is checked visually whether it is at the desired location and of desired size and e.g. not
extending to the edges on toe and heel side.

In KISSsoft, the tooth contact of bevel gears is assessed using a two dimensional LTCA under load.
Thereby, the stiffness calculation of the teeth, which is the basis of the LTCA, is calculated according
to Weber/Banaschek. This method is based on cylindrical gears and has been successfully adapted
for bevel gear calculations. Important results of the numerical calculation are the relative position and
alignment of the bevel pinion and gear under load as well as the above mentioned VHJ
displacements. In this calculation, the modifications as defined by the user are of course considered.
Results include – in the same manner as for cylindrical gears – e.g.: Hertzian contact pressure,
transmission error TE, line load distribution and others. Also, the maximum line load wmax is then
known and, combined with the mean line load wbm it is possible to calculate a load distribution factor
KHβ in the same way as for cylindrical gears.

4.3 Influence of bearing pre-tension on the axial position of the gears in

mesh

As mentioned above, the axial position of bevel gears needs to be considered in detail (opposed to
cylindrical gears, where the axial position is of little relevance) when calculating the tooth contact
patterns. The axial position is much influence by the bearings supporting the shaft. In the gearbox
studied, the pinion shaft is supported in axial direction by taper roller bearings. Typically, they are
assembled with a slight pre-tension in order to achieve a more even load distribution in operation. The
pre-tension is achieved by shimming. In the KISSsys model, the pre-tension is modeled such that the
position of the bevel pinion does not change, that is, the pre-tension is modeled by moving both outer
rings by half of the pre-tension amount in opposite directions.

4.4 Support of the axial force on the housing cover

The axial force from the bevel pinion is supported by the housing through a bearing cover attached to
the housing by means of bolts as shown below. To consider this force flow in the housing correctly, a
non-linear FEM model is elaborated where contact formulation including friction (assuming a
coefficient of friction μ=0.10). In this model, the cover and the bolts are modeled, the bolts include a
pre-tension of FV=44.1 kN, Figure 19, left. The master node of the left taper roller bearing is now
attached to this cover, Figure 19, right, marked in blue. With this modelling approach, it is possible to
lead the axial force over the cover while the radial force is still directly transmitted to the housing,
Figure 19, right.
Figure 19 Detailed FEM model with cover and bolts (left). Definition of master node on the cover,
extended towards the right side to match master node position to bearing centre position.

The FEM calcualtion is now done in two steps. In the first step, the bolt pre-tension is applied, then,
the FEM model is linearised and the reduced stiffness matrix is exported. The effect of the bolt pre-
tension is in the below deformation plot, Figure 20.

Figure 20 Deformed FEM model due to bolt pre-tension

As expbected, the deformation calculation in KISSsys using the above modified stiffness matrix yields
that the three master nodes, 1, 2 and 3 (located in the centres of the bearing supporting the bevel
pinion shaft) have higher axial displacement. However, the effect is minimal and of just a few microns,
Figure 21.
Figure 21 Axial displacement of the bearing centres 1, 2 and 3 considering the bearing cover elasticity.

Key
“Axiale Verschiebung der Ritzelwellenlager” = axial displacement of the pinion shaft bearings
“Lagerverschiebung” = Bearing displacement
“FE Modell ohne Deckel” = FEM model without cover
“FE Modell mit Deckel” = FEM model with cover

4.5 Temperature influence

Due to the different power losses inevitably present in the gearbox during operation, the gearbox parts
will heat up and expand. In this example, it is assumed that the gearbox temperature increases from
20°C to 80°C (for all parts in an uniform manner) resulting in expansion of housing and shafts. It is of
interest how the bevel pinion shaft is getting displaced due to temperature effects. For this, the FEM
model is calculated in three steps
- Bolt pre-tension is applied
- Temperature increase is applied
- Model is linearised and stiffness matrix is exported

In this, the housing bores are left free to deform. The resulting bore centre / bearing position change is
shown in below figure and is reaching a value of 0.85 mm at the highest location.
Figure 22: Housing deformation due to thermal load

In the KISSsoft shaft calculations, thermal expansion of shafts is defined with respect to a reference
point that remains fixed in the local coordinate system. In case of bevel gears, this reference point is
typically selected as the pitch cone apex. For the assumed operating temperture of 80°C, the shafts
expand with respect to the pitch cone apex which results in an axial movement of the gears,
influencing their VHJ displacement values, see Figure 23.

Figure 23: Displacement of bevel pinion and bevel gear shaft under 80°C temperature.

Key
“Verschiebung” = displacement
“Längsachse” = longitudinal axis

The below contact patterns show two typical operating conditions. The contact patterns are checked at
assembly, without load. Housing and other parts temperature is then at 20°C. V and H displacement
values are very low and within a few 10 μm. See Figure 24 below, left side.
With external torque load applied and in operating temperature of 80°C, the V and H displacement
values are V = -0.25 mm and H = 0.21 mm. The contact is slightly moved towards the heel side, see
Figure 24 below, right side.

Figure 24: Tragbildlagen der Kegelradverzahnung bei Montage (links) und Betrieb (rechts)

Key
“aussen” = heel side “innen” = toe side
“Breite“ = face width

4.6 VHJ-displacement considering housing deformation

The influence of the housing deformation can easily be shown for the bevel gear stage by looking at
the VHJ-displacement values. In below Figure 25, the VHJ-displacement values are shown, once with
and one without housing deformation.

AS expected, if the housing deformation is considered through the housing stiffness matrix, the
displacements are higher once external loads are applied. However, the influence of the housing is
small and do not change the contact pattern noticeably.

Figure 25: Bevel gear mesh displacement after assembly and in operation (with and without housing influence)

Key
“bei Montage, last frei” = at assembly, without load
“unter Last, ohne Gehäuse” = with load applied, housing not considered
“unter Last, with Gehäuse” = with load applied, housing considered
“Verlagerung” = displacement

4.7 Bevel gear modifications

In the production of bevel gears, gear modifications are often applied by the manufacturer based on
experience or general guidelines found in the respective literature. This because the designer typically
lacks tools to calculate required and optimal modifications. The designer often has to accept that the
modifications are decided by the manufacturer and is left to wonder whether the part machined will
fulfill his requirements.

®
For the Zyklo-Palloid bevel gear set used in this example, the literature (Klingelnberg) recommends a
crowning of 1/350th to 1/800th of the face width for “small” displacements. Often, in practical
applications, a value of 1/1000th is used. This would correspond to a crowning value of 84 μm,
resulting in the contact to be somewhat on the heel side with a maximum contact pressure of 1360
2
N/mm (Figure 26, left). Assessing the line load distribution, a KHβ value of 2.7 is found. The peak to
peak transmission error is calculated at PPTE = 6.8 μm. With this modification, a contact pattern
without edge load and stress peaks is achieved.

In order to maximise the power density of the bevel gear set, the modifications are now optimised. The
contact pattern is moved to the centre of the flank and the crowning is reduced to 20 μm (Figure 26,
2
right). The resulting contact stress now drops to 1140 N/mm , or an improvement of almost 20%. KHβ
drops considerably to 1.7 and the peak to peak transmission also is reduced to PPTE = 2.45 μm. We
may therefore not only expect an increased load bearing capacity but also a reduced noise level.

Figure 26 Tooth contact pattern with crowning as recommended by literature (Left) and after optimisation (Right)

4.8 Influence of gear manufacturing tolerances

In ISO 17485 ‘Bevel gears − ISO system of accuracy’, no values for the manufacturing tolerances in
lead direction are given, an assessment of the effect of the gear quality on the contact is hence not
directly possible.
As a simplification, in the scope of this study, the values listed in ISO 1328 for cylindrical gears are
used. As a reference, the equivalent cylindrical gear is used to determine the applicable helix slope
deviation at fHβ=10 μm for the pinion and 12 μm for the wheel.

A combination using a statistical approach covering 99.7% of all cases results in a total of fHβ= +/- 17.5
μm. This manufacturing error has very little influence on the load distribution when combined with the
above optimized lead modifications, see below Figure 27.
Figure 27: Tooth contact pattern with fHβ = -17.5 μm (left) and + 17.5 μm (right).

Key
“aussen” = heel side “innen” = toe side
“Breite“ = face width

5 Summary
The above study illustrates that it is quite a complex task to perform a detailed and accurate
calculation considering bearings, shafts, temperature and housing influence on the gear mesh. While it
is very simple and fast to consider e.g. gear machining errors or bearing stiffness in the calculations,
the integration of housing information takes more time. Still, this may well be justified in critical
applications.

Considering a compact and stiff housing as used for the industrial gearbox investigated in this study,
we find that the housing deformation is not the most relevant factor influencing the gear mesh. Current
practice (not to consider the housing influence in such products) is therefore supported. Note however
that these findings are not applicable for other types of gearboxes where the influence of the housing
deformation may be more relevant. Shaft and bearing deformations as well as variations in the bearing
clearance, manufacturing errors and thermal influences are more significant. Furthermore, the
commonly held wisdom that a soft housing leads to meshing problems is not supported as a soft
housing may keep shafts (and the gears) nicely parallel even if the centre distance is increased
considerably (but due to the nature of the involute gearing, this does not affect the mesh negatively).

Even for the bevel gear stage, it was found in this particular design that the influence of the housing
deformation on the meshing condition is marginal. Recommendations from literature as to the amount
of crowning to be applied were confirmed, they are suitable to compensate manufacturing tolerances
and the housing deformation. The bevel gear LTCA function implemented in KISSsoft has again
proven itself as a most valuable tool for the assessment of spiral bevel gears in contact.

The above findings are in line with common practice in general gearbox engineering that the housing
deformation is not of prime relevance when designing gear modifications.

The resulting improvements in load distribution as designed using KISSsoft and KISSsys are shown
below, Figure 28.
Figure 28: Improvements achieved in face load factors for all three gear stages.

Key
„Getriebeoptimierung“ = gearbox optimisation „vorher“ = before (optimisation)
„Breitenlastfaktor“ = face load distribution factor „nachher“ = after (optimisation)
„Kegelrad“ = bevel gear (stage)
„Stirnradstufe“ = cylindrical gear stage

The focus of this study was to show the influence of the housing deformation on the required
modifications. The following figure shows the influence of the housing on the required modifications.
These are shown while not considering the housing deformation in blue and while considering the
housing deformation in red. It is well visible that out of the total modifications required, the effect of the
housing deformation is small in this gearbox / housing design.

Figure 29: Influence of the housing deformation on the gaping in the different meshes / required modifications.
Key
“erforderliche Flankenwinkelkorrektur” = required helix angle correction (modification)
“Korrektur” = correction (modification) “ohne” = without
„Kegelrad“ = bevel gear (stage) “mit” = with
„Stirnradstufe“ = cylindrical gear stage “Gehäusesteifigkeit” = housing stiffness

In conclusion, we recommend to
- Pay great attention to the modeling of bearings, in particular their operating clearance and
non-linear stiffness in operation, considering e.g. thermal effects and pre-tension.
- Consider gear and housing manufacturing errors on the gear misalignment.
- Consider the housing stiffness as a final step to optimize a product in case housing
deformations are high and uneven, the product is for a critical application or for high volume
production

Again, it should be noted that other types of products, the housing deformation may be of greater
relevance. Then, the use of KISSsoft and KISSsys allowing for direct import of reduced housing
stiffness matrices from ABAQUS, ANSYS and NASTRAN is a most helpful tool to calculate the
housing deformation under load and to assess the resulting gear mesh misalignment to achieve a best
possible meshing situation under various operation conditions.

References
[1] ISO 6336, Teil 1: Calculation of load capacity of spur and helical gears; ISO Genf, 2006

[2] Kissling, U.: Sizing the gear micro geometry – Step by Step; BGA GEARS, 2015

[3] Handbuch KISSsoft/KISSsys, Berechnungssoftware für den Maschinenbau, www.KISSsoft.AG

[4] Mahr, B.: Kontaktanalyse; Antriebstechnik 12/2011, 2011

[5] ISO/TS 16281: Rolling bearings − Methods for calculating the modified reference rating life for
universally loaded bearings; Genf, 2008

[6] ISO 1328-1: Cylindrical gears – ISO system of flank tolerance classification; Genf, 2013

[7] ISO/TR 10064-3: Recommendations relative to gear blanks, shaft centre distance and parallelism
of axes; ISO Genf, 1996

[8] Langhart, J.: Einfluss der Gehäusesteifigkeit auf die Verzahnungslaufeigenschaften; DMK 2015

[9] Klingelnberg, J.: Kegelräder; Springer Verlag, Berlin, 2008

[10] ISO 10300, Teil 1: Calculation of load capacity of bevel gears; ISO Genf, 2014