Joel_Lapointe
Mechanical
We have an aluminium die casting housing that receives some machining. From time to time, the machining is notching a wall. A hole is produced so 2 empty volumes are communicating.
Since all of this is sealed with 2 shaft seals and a cover, we may accept that hole as is. However, we aim to make sure that the fatigue behavior is documented.
We do not have a fatigue package with our CAD package. We don't do fatigue enough to justify the cost. So our strategy is to model the first fatigue cycle as a static analysis (2 simulations, 1 for pushing force, 1 for pulling force, the full alternate cycle is made with a push-pull motion).
Then,the nodes stresses are extracted with a minimal value and maximal value. An equivalent fatigue stress is computed for "Sa" (alternate stress) ans "Sm" (mean stress). We then plot the S-N curve for these nodes. Same thing for the modified Goodman diagram.
My problem is that Kt (stress raiser) isn't taken in account in FEA since the geometry is resolved enough with meshing (no singularity). However, I have the difficulty to not consider the notch sensitivity factor Ke = 1/Kf. In my head, there is 2 competing ways :
First way:
- If Kt = 1 (because of FEA is well resolved with mesh) then, Kf also = 1 and Ke = 1. Hence, no need of notch sensitivity in fatigue based on static FEA.
Second way:
- Simulate without notch (Gives stress S_nonotch). Get the worst node of the interest area (make sure there is no singularity)
- Simulate with notch (Gives stress S_notch). Get the worst node of the same interest area (make sure there is no singularity)
- The ratio of the 2 stresses is Kt = S_notch/S_nonotch > 1
- Then, Kf = q(Kt-1) + 1
- Then, Ke = 1/Kf
- Then, the endurance limit: Se = Ka*Kb*...*ke*Se' is modifying the S-N curve to fit the non ideal case of real aluminium.
- Then, the Modified Goodman diagram is computed with the modified "Se" to compare with the nodes "Sa" and "Sm".
- Then, the minimal factor of safety is taken from the worst node of the area.
Well...
Since all of this is sealed with 2 shaft seals and a cover, we may accept that hole as is. However, we aim to make sure that the fatigue behavior is documented.
We do not have a fatigue package with our CAD package. We don't do fatigue enough to justify the cost. So our strategy is to model the first fatigue cycle as a static analysis (2 simulations, 1 for pushing force, 1 for pulling force, the full alternate cycle is made with a push-pull motion).
Then,the nodes stresses are extracted with a minimal value and maximal value. An equivalent fatigue stress is computed for "Sa" (alternate stress) ans "Sm" (mean stress). We then plot the S-N curve for these nodes. Same thing for the modified Goodman diagram.
My problem is that Kt (stress raiser) isn't taken in account in FEA since the geometry is resolved enough with meshing (no singularity). However, I have the difficulty to not consider the notch sensitivity factor Ke = 1/Kf. In my head, there is 2 competing ways :
First way:
- If Kt = 1 (because of FEA is well resolved with mesh) then, Kf also = 1 and Ke = 1. Hence, no need of notch sensitivity in fatigue based on static FEA.
Second way:
- Simulate without notch (Gives stress S_nonotch). Get the worst node of the interest area (make sure there is no singularity)
- Simulate with notch (Gives stress S_notch). Get the worst node of the same interest area (make sure there is no singularity)
- The ratio of the 2 stresses is Kt = S_notch/S_nonotch > 1
- Then, Kf = q(Kt-1) + 1
- Then, Ke = 1/Kf
- Then, the endurance limit: Se = Ka*Kb*...*ke*Se' is modifying the S-N curve to fit the non ideal case of real aluminium.
- Then, the Modified Goodman diagram is computed with the modified "Se" to compare with the nodes "Sa" and "Sm".
- Then, the minimal factor of safety is taken from the worst node of the area.
Well...
