FEM analysis of a gearbox housing, for the

calculation of stress and deflection characteristics

G. Camp1 , N. Drivakos2
1
Volkswagen A.G., 2 BETA CAE Systems S.A.
Abstract: Stresses and deflections developed on gearbox housing play a crucial role in the load
carrying capacity of its components, as well as the durability of the gearbox itself. Building and
testing actual prototypes is a procedure that requires a considerable amount of resources (i.e.
personnel, time and consequently money). On the other hand, current market needs, impose short
time development cycles and cost reduction. To minimize the number of prototypes built, an initial
virtual model is created (CAE) and analysed for a number of analysis cases. The above procedure,
leads to an increasing need for the creation of complicated models for high simulation accuracy.
More and more components are included in the models while on some studies, machine elements
like gears, bearings and bolts have a detailed FE representation.
In the present study, a real case automotive gearbox housing is modelled, in order to investigate
stress and deflection characteristics for a number of load cases. It incorporates the FE modeling
of the housing, contacts simulation, assembly and operating conditions loading. Solution results
are obtained and evaluated in accordance to the design requirements. Finally, a conclusion
whether to further modify the design or proceed with building a prototype for testing is reached.
The study was done utilizing ANSA pre-processor and Abaqus/Standard.
Keywords: Abaqus/Standard, ANSA, Automotive, FEM, Gearbox, Powertrain.

1. Introduction

The creation of a new automotive gearbox is a complicated and demanding operation. A gearbox
is designed to withstand a specific range of load carrying capacity and drive characteristics. The
first step in the design cycle is to calculate the main components like gears pairs, bearings and
shafts. Following the gearbox housing is calculated. One of the requirements to be met is that the
housing must be stiff enough to ensure flawless operation of the gear pairs due to the elastic
deflections and vibrations of the shafts and the housing. Additionally, the housing design should
also guarantee optimal behavior in durability and fatigue. Finally, the gearbox should meet the
size, weight and cost restrictions imposed by the engine and the car manufacturer.
Building and testing actual prototypes is a procedure that requires a considerable amount of
resources (i.e. personnel, time and consequently money). On the other hand, current market needs,
impose short time development cycles and cost reduction. To minimize the number of prototypes
built, first a virtual model is created (CAE) and analyzed for a number of analysis cases. From the
results obtained, the engineers try to optimize the behavior of the model, by continuously
modifying the design and rerunning the analysis. Only after the results are satisfactory, the first
actual prototype is constructed and tested.
The above procedure, leads to an increasing need for the creation of complicated models for high
simulation accuracy. More and more components are included in the models while on some

2010 SIMULIA Customer Conference 1

studies, machine elements like gears, bearings and bolts have a detailed FE representation.
Preparing a gearbox-housing model for FEM analysis is a complex and tedious process that used
to require a lot of man-hours from experienced engineers and usually involved the combination of
different software for each step of the simulation (mesh generation, connections, loads, model
built up). All these, in combination with the ever-increasing model sizes that are necessary for
realistic and accurate simulation, significantly raise the complexity of the process, making it error
prone and stiff. In the last couple of years many evolutions have taken place and today there are
pre-processors that offer not only the majority of these functionalities, but they incorporate them
in a fully or semi-automatic way, while solvers allows us to combine all the above and give us fast
and accurate results.
This study, presents the modeling of a gearbox housing for FEM analysis of stresses and
deflection characteristics, realized using one pre-processor and one solver. Meshing and model
build up was done using ANSA, one of the leading commercial pre-processors. The model was
solved using Abaqus/Standard and following, µETA post-processor was used for viewing and
evaluating the results obtained.

2. MODEL BUILD UP

Table 1: Model build up phases

Geometry Handling - CAD data input
- Erase not needed parts
- Check and repair geometrical errors

Mesh - Surface meshing

- Check and fix mesh quality

- Volume Meshing

Model Build Up - Contacts / Connectors

- Boundary conditions
- Manipulation of solver entities
- Checking of the model
- Solver Header

Solve - Solve model

Evaluate Results

The pre-processor used for the FE model creation was ANSA version v13.0.3. The process
followed through the study consists of the steps summarized in Table 1. A brief description of
each step is presented in the following paragraphs.
2.1 Geometry Handling
Volkswagen AG provided the CAD input files in Pro/E format. The CAD files were directly
imported into the preprocessor using an application (CAD to ANSA Translators), which translates

2 2010 SIMULIA Customer Conference

native CAD data to ANSA. With the use of native Pro/E data (as opposed to exporting an
intermediate neutral format like iges or step and importing the latter), apart from the geometry, the
meta-data that exist in the CAD file are transferred to the pre-processor. Color and layer
management done in the CAD as well as all information regarding part positioning, hierarchy,
multi-instantiated parts, automatically pass in the pre-processor, so both the designer (CAD
department) and the analyst (CAE department) share the same model organization, resulting in an
uneventful cooperation.

Figure 1: Parts participating in model analysis.

In the present study, only the parts that contribute to the housing stiffness are used. These are, the
two cases of the gearbox (parts A, B), the engine parts where the housing is mounted (parts C, D)
and some auxiliary parts bolted on the gearbox (parts E, F), Figure 1. All remaining parts were
neglected alongside with the geometrical definition of the different kind of bolts. Bolts were
modeled with a combination of rigid and beam elements with specific characteristics.
The last step of the geometry handling was to ensure that there was no geometrical error due to an
incomplete design of a part in CAD. A check for initial penetration of the parts was also
performed in case some parts had been wrongly positioned or intersected each other. In latter case,
erroneous parts were either correctly positioned, or modified to ensure an intersection-free
assembly.
2.2 Meshing
As mentioned earlier, in the virtual prototype development cycle (CAE cycle), after the evaluation
of an analysis results, slight design modifications may be introduced to some parts to improve

2010 SIMULIA Customer Conference 3

their behaviour. The new design must be re-evaluated and therefore the FE-model must be updated
to accommodate all design changes. Furthermore, there are cases that some parts might be meshed
using different set of mesh specifications. For this reason it is significant for the engineer to be
able to keep track of the parts versions and the different meshes used in the model.
To maintain compatibility with the PDM system, part id and version is automatically transferred in
ANSA during the CAD translation phase. Furthermore, utilizing a number or the preprocessor’s
features, (ANSA Data Management) [1], it is possible to monitor each part’s versioning and
automatically update the assembly when a new component becomes available. The product
structure with all related meta-data is directly accessible, giving the user an overview of the model
in-hand (parts used, versions, mesh specification), Figure 2.

Figure 2: Model organization (ANSA Part Manager).

As mentioned previously, the parts included in this analysis are presented in Figure 1. Depending
on the part and its position in the assembly, components were meshed using three different mesh
parameters. To better simulate the gearbox housing and get more detailed results, the model was
meshed with second order solid elements. To create a solid mesh of high quality, it is imperative
to start from a surface mesh of also high quality. The main parts of the gearbox housing (part A,
B) were surface meshed with second order triangles with a mean element length of 3.5 mm. In
order to keep the total number of elements low, parts C, D and E that only contribute to the
stiffness, and we are not interested in the stresses developed on them, were meshed with second
order triangles with target element length of 4.5 mm. Gaskets and pins were meshed using
hexahedral elements. The different mesh types used in the analysis are summarized in Table 2.
Table 2: Mesh specifications used

No Mean Length Surface Mesh Solid Mesh Parts

Fine 3.5 mm Triangles (STRIA65) Tetras (C3D10M) A, B

Coarse 4.5 mm Triangles (STRIA65) Tetras (C3D10M) C, D, E

Hexa 3 mm Quads (S4) Hexas (C3D8) F

In order to obtain the best ratio of quality versus time spend, surface mesh of all parts was
performed using ANSA automatic mesh creation tool (Batch Meshing) [2]. With Batch Meshing
the user specifies the required mesh characteristics and the quality criteria that the resulting mesh

4 2010 SIMULIA Customer Conference

has to meet and ANSA automatically generates a mesh of the prescribed quality. Since no fully
automatic procedure is able to guarantee a result with zero elements violating the quality criteria,
after batch mesh is finished, the generated mesh is inspected and any remaining violating elements
are automatically isolated and manually corrected. This is necessary because most solvers accept a
limited number of elements below a quality threshold and in the case that this number is exceeded
the solver either stops completely or the calculated results are of poor quality.
Having finished with the surface mesh improvement, the next step is the generation of the solid
mesh. Since all parts have a 3D geometrical definition, the definition of the volumes, and the
generation of the solid elements are made automatically. As a last step any violating solids are
corrected by the user, to comply with the mesh specification imposed by the solver and analysis
type. The total number of elements per part and the time spend for the automatic creation of the
mesh is summarized in Table 3.
Table 3: Element per part and time needed for automatic creation of the mesh
Parts Surface Solid Automatic (h:min) Manual (h:min)
A 199 998 757 560 0:12 1:30
B 237 028 627 478 0:15 1:45
C 25 962 167 034 0:02 0:10
D 16 184 37 776 0:01 0:10
E 4 425 4 425 0:01 0:30
F 31 112 52 348 0:06 0:05
Total ~516 000 ~1 515 000 0:37 4:10

2.3 Bolt Modelling

In the gearbox assembly, all parts, except the bearings, are connected using bolts. The bolts are
modeled using a combination of beams and rigid elements. For the thread length of the bolt in
contact with the hole thread, as well as the bolt head, rigid elements (*MPC) are used. The free
body of the bolt, the part between the head and the thread rigid elements, is modeled with a beam
combination (three B31 beams). These beams have section characteristics imposed from the actual
bolt while a pre-tension entity (*PRE-TENSION SECTION) is also applied in one of the beams to
correctly simulate the bolt connection. The beam characteristics and pre-tension values, for each
bolt type, are summarized in Table 4.
Table 4: Bolt characteristics
Bolt type Radius (mm) Pretension (N)
M6 (9 pcs) 2.53 9 000
M8 (26 pcs) 3.41 17 361
M10 (3 pcs) 4.30 25 526
M12 (5 pcs) 5.18 41 400

Creation of the bolts presented two difficulties. The first was that the model had four different bolt
types at 43 various locations. The second was that during the various phases of the analysis some
parts had to be replaced by new or updated versions. The bolt connections representation would
have to be recreated and adapted to the new design every time there was a part update.
The above problems were solved using a number of automatic features of ANSA. For each bolt
type an ANSA Connector Entity [3] was created. The characteristic of such entities is that they

2010 SIMULIA Customer Conference 5

hold information regarding the connected parts, the “representation” element to be generated and
the “interface” elements; i.e. the intermediate elements between the representation and the
connected parts. In this application the representation elements was the beam and the interfaces
were the rigid elements. Each ANSA Connector Entity created consists of two parts. One part
creates the rigid element that represents the thread in contact and the other creates the bolt head
MPC, connects it using the beam elements to the MPC of the thread and applies the desired pre-
tension to the beam. In this way, the user has to only define a Connector Entity for each bolt type.
Following, the defined connector is copied to all the needed positions either manually or via a
script (that reads a list with the locations coordinates) and in the end all connectors are
automatically realized. During realization, ANSA identifies for each connector the proper nodes
from the participating parts and creates the desired connection representation. The resulting bolt
modelling is presented in Figure 3. At this point it must be noted that the creation of the ANSA
Connector entities is done only once during the initial model creation.

Figure 3: Bolt representation

2.4 Contact definition
To correctly simulate the behaviour of the bearings that are pressed in the bores of the housing, tie
contacts (*TIE) have been created between each bore and the corresponding bearing.
A tie contact is a boundary condition used by solvers to connect nodes from one surface to
elements of the other in order to restrict any movement between them. The tie contact as shown in

6 2010 SIMULIA Customer Conference

Figure 4 is defined between the outer solid face of the bearing and the solid face of the housing
bore.

Figure 4: Tied definition between bearing and hole.

For the part pairs connected by bolts, contact definitions are created (*CONTACT PAIR), Figure
5. Contact definitions prevent the penetration of elements of one part into elements of the other,
due to deflection. A friction model was also defined for the contact to represent the friction
developed during sliding between the two parts. In the present study a simple friction model was
used with a constant friction coefficient value of 0.1. Even if during the meshing phase the model
was checked and fixed for initial penetrations a clearance value (*CLEARANCE) was also
specified in the contact pairs to ensure no penetrations are present at the beginning of the
simulation.
During model built up the user must be careful how to use contacts definitions or tie since it
greatly influences model solution. For example, in reality part F (a gasket) is used to prevent direct
contact of the gearbox (part A) to the engine (parts C and D), is kept in its place due to the
pretension of the bolts that hold parts A and C together. If this is simulated via two contact
definitions then solution will stop since part F at the beginning of the simulation will be
considered as not connected to the rest of the model. To overcome that, these cases were modeled
using a combination of one tie and one contact definition. The tie connects one side of part F to
part A (ensuring that part F is not free) and the contact definition is used to prevent intersections
between F and parts C - D. The above was used for the retaining pins, and the gaskets / flanges.

2010 SIMULIA Customer Conference 7

 Figure 5: Contact definition between the two main parts.
2.5 Boundary conditions
Figure 6 shows the mounting boundary conditions of the gearbox housing. The housing is fixed at
five positions, which represent the mounting of the gearbox to the engine. At each of the five
positions an edge is selected and all the nodes on that edge have their translational (x, y, z) degrees
of freedom fixed (*BOUNDARY) during solution, Figure 7.

Figure 6: Mounting boundary conditions.

8 2010 SIMULIA Customer Conference

 Figure 7: Fixed nodes in x, y, z directions.
2.6 Part replacement - Version Update
The development of a new gearbox is usually based on an already existing design. Old parts are
re-designed in order to comply with the new requirements or completely new parts are introduced.
This process is repeated many times throughout the development cycle. So, the analyst has to
continually update his model with the new versions of the parts. Furthermore, when changing a
part, he also has to re-apply all boundary conditions, loads, connection entities, and update
previously defined grouping of entities, i.e. sets of elements that are used by contact definitions
(*SURFACE). This was taken into account when the housing model was created. Connectors
entities were used for bolts, while for the boundary conditions and the contacts set definition
special ANSA entities were used that allow the automatic recreation of these entities if any change
occurs.
When a new part becomes available, the user should execute the following steps. Start with
meshing the new / updated part using batch meshing, and then continue with the manual repair of
any remaining violating elements. Then, through ANSA Parts Manager replace the existing part
with the new version. Any dependent entities of the outgoing component are automatically applied
on the new the part. Connectors, boundary conditions (generic entity builders) [3] and contact
definition sets are automatically recreated using the nodes / elements of the new part. When the
procedure is over, a report of any failed operation is printed and the user knows those areas / cases
that need further treatment.
2.7 Load cases
To examine the stresses and deflections characteristics of the gearbox housing, the following cases
have to be investigated: For every gear pair two cases, one for acceleration and a second for

2010 SIMULIA Customer Conference 9

deceleration. This is done for all seven gear pairs, for the reverse, as well as for parking uphill and
downhill. Summing the above cases result in a total of 18 different loads cases.

Figure 8: External forces on the bearings.

For the gearbox housing calculation, as mentioned earlier, gear pairs, shafts and bearing's inner
rings are not included in the model. The loads (*CLOAD) are applied on the inner solid face of the
bearing outer rings as shown in Figure 8 and are the reaction forces developed at the bearings on
each load case.
These reaction forces are previously calculated using a simulation software for the 18 load cases.
This simulation software is supplied with the design characteristics of the gears, shafts, bearings,
the stiffness characteristics of the participating parts as well as the external loads of the gearbox
and calculates the reaction forces on bearings rings for each load case.
The fact that the shafts and all the components mounted on them are not included in the model
means that the same model of the housing can be used for all 18 calculations by simply changing
the applied loads (loading step).

3. Results and Discussion

3.1 Solution
Abaqus/Standard v6.8-3 [4] was used to solve the model. Solution was made into two stages
(steps).
The first (STEP 1) simulates the gearbox before loading due to operation, when only bolt
pretension and initial boundary conditions are present. This step is made in order to calculate the
stresses and deflection of the gearbox casing due to the bolts pretension. Since this stage is
common for all loadcases it was calculated only once and the results derived were used then for

10 2010 SIMULIA Customer Conference

each loadcase. As mentioned above the model consisted of 1.5 million second order tetras
(C3D10) which resulted in a 350 MB Abaqus input deck. A 16 cpus cluster was used to solve this
step. The step was solved in only one increment to keep the output file as small as possible. The
results were obtained after 8 hours and the size of the file was 1.4 GB, Table 5.
Table 5: Solution steps information
Step No. of cpus Time (hours) Input file Results size
Step 1 (Pretention) 16 8 350 MB ~ 1 400 MB
Step 2 (loadcase) 32 3.5 Restart step 1 ~ 1 400 MB

Since contacts and pretensions have been established in the initial step (Step 1) restarts were made
from these results, for each loadcase, resulting in a STEP for a each of the 18 different loadcases.
To further speed up solution, a 32 cpu cluster was used, resulting in lowering the time needed for
each step to 3.5 hours. Again the size of each result file was around 1.4 GB, Table 5.
3.2 Results
As mentioned above, the first step simulates the gearbox before loading, when only bolt
pretension and boundary conditions are present. Due to the bolts pretension, stresses are developed
around the bolt holes and the contact interfaces, Figure 9.

Figure 9: Calculation of the stress due to the bolts pretension (Step 1).
Following, based on the results of the first step, each loadcase is applied (loads on the bearing
rings) and the calculation continues to obtain the final stresses developed on the gearbox housing,
Figure 10. In this way the model better simulates the reality since the gearbox is first assembled
and mounted on the engine, developing stresses due to the bolts pretension (Step 1), and finally is
loaded during engine operation (Step 2).
The results presented for Step 2 are for full acceleration using the first gear pair of the gearbox.

2010 SIMULIA Customer Conference

11
 Figure 10: Von Misses stresses developed on housing (Step 2).
In Figure 11, the Von Misses stresses developed on the gearbox housing are shown. In the bigger
part of the housing, stresses have a value lower than 25% of the tensile strength limit. At areas
around bearing bores where usually cracks are developed due to high stresses, the values
encountered are around 65% of the material’s tensile strength limit (100%).

Figure 11: High stresses developed near bearing housings (Step 2).

12 2010 SIMULIA Customer Conference

 Figure 12: Deformation of the gearbox housing.

Figure 13: Contact Pressures after Step 1 (pretension).

2010 SIMULIA Customer Conference

13
 Figure 14: Contact Pressures after Step 2 (loading).
In Figure 12 the deformed over the undeformed (wire frame) housing is presented. The
calculations also show that, the highest eccentricity of a pair of bearings centers, due to the
housing deflections, is within the designer’s tolerances.
Another problem encountered in gearboxes is that due to elastic deflection of the housing there is
the chance that two parts might be separated, and let the lubricant escape. This will, eventual lead
to the failing of the enclosed machine elements (gears, bearings) due to lack of lubrication. For
this reason, the value of the contact pressures between the parts is very important. In Figure 13 the
contact pressures developed between part A and part B are shown as a result only of the bolt
pretension loading. The contact pressures are evenly distributed on the contact area getting
maximum values around the bolt holes. As a result of the external loads, Figure 14, the contact
pressure distribution changes. This leads to higher contact pressures in lower part of the contact
area and lower pressures on the upper region. The values also show that separation of the two parts
under load is avoided.

4. Conclusion

In this study the modeling of a gearbox housing for FEM analysis of stress and deflection
characteristics was presented. The model was created using the ANSA pre-processing software.
The model set-up was done in such a way to support fast and accurate synchronization of the CAE
model with CAD updates, and easy application of the forces for the 18 different load cases.

14 2010 SIMULIA Customer Conference

The gearbox model was solved using Abaqus/Standard v6.83. Solution was conducted into two
steps. The first simulated the pretention of the bolts. The second step simulated each loadcase.
Each loadcase step differed only in the applied loads and it used the results obtained from the first
step (*RESTART). Complete analysis of all 18 loadcases took 63 hours (3.5 hours x 18 = 63) for a
model that consisted of 1.5 million, tetras (C3D10M), and 16 TIE orCONTACT pairs.
Stresses and deflections characteristics were calculated for all cases, while the results for the first
load case (full power acceleration of the first gear) were also presented in this study.

5. ACKNOWLEDGEMENTS

The authors would like to thank Volkswagen AG for the support given throughout this work and
for granting permission to publish this study.

6. References

1. Makropoulou I., Kiouptsidis K., Skolarikis K. “Capturing the best practices for the
organization and analysis of crash-test simulation load-cases.” Proc of the 2nd ANSA &
µETA INTERNATIONAL CONGRESS,Chalkidiki 2007
2. Saltiel S.-T., Tryfonidis M.. F2006M141: Latest Developments in CAE Process Automation:
An Insight into Batch Meshing of BiW Models, FISITA 2006, Yokohama 2006.
3. ANSA v13.0.3 User’s Guide, BETA-CAE Systems S.A., 2009.
4. Abaqus/Standard version 6.8 Documentation, Simulia

2010 SIMULIA Customer Conference

15