stresses in a transient analysis due to torque

[NewMechE4]

Hello,

We have a customer who is experiencing fatigue failures at the hub weld of some centrifugal fan impellers which are driven by electric motors. They believe that the cause is that they are using what they call a "hard start" (they are not controlling the startup acceleration, and instead they are dumping full power to the motor.) They have requested a study that correlates the motor startup time to the fatigue life of the impeller. To do that, I performed a transient FEA of the impeller to determine the stress at the hub weld when subjected to different values of torque. The issue is that the results aren't showing any difference in stress whether I apply 1000 ft-lbs of torque or 5000 ft-lbs of torque. The stress increases as the square of the rotational velocity as expected, so the stress does increase faster with respect to time when I apply a higher torque. But I thought I might see an initial spike in stress when the torque is first applied and is opposed by the inertia of the impeller. I applied the torques as step loads, although now I am thinking maybe I should apply them as ramped loads.

Any thoughts?

Thanks for your time

 

[FEA way]

Which software do you use for this simulation ? Can you provide some details like type of dynamic solver (implicit or explicit), boundary conditions, interactions and the way load is applied to the model ?

 

[NewMechE4]

Sorry, should have included more details.

I'm using ANSYS Mechanical with the implicit solver. I'm applying a moment to the cylindrical face where the shaft makes contact with the hub. Internally, I believe ANSYS puts a remote point at the center of the cylinder which is connected to the face with constraint equations. Then a torque is applied to the remote point. There is a keyway which I defeatured, but I'm really only interested in the stresses at the welds, so I thought this would be ok. I constrained the nodes on the hub in the direction of the shaft axis. There is also a remote point which is scoped to the outside diameter of the baseplate which has 0 displacement constraints in in the x and z directions (in the plane of the plate). The idea was to allow radial displacement of the nodes, but to require the average of the radial displacements to be zero to prevent rigid body motion.

Step End Time: 0.1s
Initial Time Step: 0.005s
Min Time Step: 0.005s
Max Time Step: 0.005s
Time Integration: On

 

[FEA way]

In dynamic analysis you don't have to worry about rigid body motions so you can relax the boundary conditions and make them represent real life case without additional constraints included only to fix RBMs. However your BCs in general seem correct. When it comes to the problem with results, make sure that density is defined correctly so that inertia effects are significant in the analysis. It's also important to use realistic amplitude for load, as you've mentioned in the first post.

 

[TomLee777]

If you consider air resistance, the stresses will go up significantly.

 

[NewMechE4]

Thanks for the feedback. I'll continue working on the problem

I believe I defined density correctly, but I will doublecheck.

Air resistance is also a good consideration, which I was not considering.