Week 2 Bevel Gear Challenge

GRID DEPENDENCY TEST ON BEVEL GEAR SIMULATION

GRID DEPENDENCY TEST

Grid dependency test is done to find out how the solution depends on the grid size by running the simulation with refined mesh. Generating a finer mesh, results in higher accuracy of the analysis but with the additional cost on computation time and computational power.We know that the solutions are interpolated at nodes to get an approximate result, when there is higher number of nodes and elements, higher is the computational time and power required by the system to complete the solution.

The grid dependency test helps in determining optimal mesh size where the accurate result could be obtained keeping in the mind the cost of time and power. Since the results obtained are approximate, refining mesh further would result in negligble percentage change in the results and only increase the computational cost.

Objective: -

1. Perform a grid dependency test for the Bevel gear simulation for mesh sizes of 6,5 and 4 mm.

2. Solve for Von-Mises stress, Equivalent elastic strain and Total Deformation using the previously mentioned mesh sizes.

3. Compare our findings for the three scenarios with appropriate figures and charts.

Procedure: -

• Open the Ansys workbench and drag and drop the static structural analysis system in the project schematic.Rename the project as "Bevel Gear analysis".
• Double click on engineering data to add the material or we can cllck right mouse botton and choose edit option from the drop down menu.We will use structural steel as working material for this analysis.

• Go to spaceclaim and edit out the extra material as this will increase the computational time and its not contributing anything to the analysis.

CONNECTIONS

• Contacts at the interface were set up between two gears with small gear having 16 faces and the big gear having 24 faces with type frictional with coefficient zero.

Meshing

If we use the default mesh the meshing will be very coarse in the teeth part, which will result in a lot of convergence issue. So we will insert a local sizing mesh of 1.75mm for the teeth parts.

Global mesh size-6

• Local mesh size-1.75

Global mesh size-5

Global mesh size-4

JOINTS

• Revolute joints were set up with connection type ‘body-ground’ for each gear. Big gear is given a constant moment of 100 N-mm and small gear is given rotation on 120 degrees in 6 steps (i.e. 20-degree increment per step). To change the direction of rotation, we can switch the axis by reference coordinate system.

Analysis setting: -

• The solution initially is not converged for the provided time step. So we later modified the data to
• initial time step-0.2
• minimum time step-5e-2
• maximum time step-0.3

Solution-

Simulation mesh result is shown for different mesh sizes in order 6,5 and 4mm.

Equivalent stress for 6mm global mesh size on Bevel gear.

Total deformation for 6mm global mesh size on Bevel gear.

Equivalent elastic strain for 6mm global mesh size on Bevel gear.

Equivalent stress for 5mm global mesh size on Bevel gear.

Total deformation for 5mm global mesh size on Bevel gear.

Equivalent elastic strain for 5mm global mesh size on Bevel gear.

Equivalent stress for 4mm global mesh size on Bevel gear.

Total deformation for 4mm global mesh size on Bevel gear.

Equivalent elastic strain for 4mm global mesh size on Bevel gear.

Tabular data for all the obtained values for the three mesh sizes ; -

Sr. No.	Mesh Size (mm)	Maximum Equivalent Stress (MPa)	Total Deformation (mm)	Maximum Equivalent elastic strain
1	6 mm	4.5255	47.869	3.008e-5
2	5 mm	4.8218	47.869	3.182e-5
3	4 mm	5.1393	47.869	3.313e-5

CONCLUSION: -

• Grid Dependency test has been performed and from the above table, we can conclude that performing grid dependency test on bevel gear by varying the global mesh size we can achieve much better accuracy of the result. We can observe that as the mesh is refined from e.g 5mm to 4mm we achieve better approximate of the result for equivalent elastic strain and equivalent stress, however, performing the simulation with further refining the mesh would result in a very close approximation of the result but with very negligible change in the result with an additional on the computational power and time.