Unexpected dog clutch disengagement behavior

Hi,

I have noticed an unexpected behavior of the Simscape dog clutch block during disengagement. Basically, there is extra resistance during disengagement around clearance (1mm in the example) that slows down the process, which should not happen since there is no chance for face-to-face contact during disengagement. I have attached an example with the specific set of parameters that will create the phenomenon. The plot is the clutch travel during disengagement. Note that by changing parameters like rotational and translational stiffness and damping, the issue might be alleviated. But it is concerning that for some reasonable set of parameters, the non-physical behavior shows up.

Please let me know if the issue makes sense, what causes it, and potentially how this problem can be fixed (besides finding a set of working parameters through trial and error).

Thanks!

Answer 1

[Martin]

Hello Frank,

Thank you for including the example model to demonstrate the issue. The translational behavior of the shift linkage is limited by hard stops. The maximum travel before hitting the stop is defined as the stroke. When the shift linkage hits the stroke limit, the stiffness and damping specified kick in.

When the clutch is able to engage (checked via tooth alignment, engagement velocity, and if it is currently engaged), the stroke is equal to the ring-hub clearance when disengaged + tooth height. When it cannot engage, the stroke is equal to the ring-hub clearance when disengaged.

To be engaged, the shift linkage must move the ring-hub clearance when disengaged + tooth overlap to engage. In the case that you showed here, it reaches a point where it can no longer be engaged, so the stroke decreased. This caused it to be impacted by the hard stop.

You can reduce this issue by reducing the tooth overlap to engage parameter. Setting it to 1e-3 completely eliminates the issue in your example.

On our side, we will look into making this more robust for a wider range of parameters.

Martin

[Frank]

Hi Martin,

Thanks for your response.

I agree that during the process of engagement, i.e. pushing the ring towards the hub, there could be two hard stops along the travel 1) at clearance (1mm) if the teeth are not aligned, I call it face-to-face contact, and 2) at the end of travel (clearance + tooth height = 4mm) when fully engaged.

However, i am concerned about disengagement process, i.e. pulling the ring out away from the hub. The only hard stop i would expect to see is the one at the very end of travel when it's fully disengaged (0mm), so the linkage will not travel infinitely backward. However, in the plot above with my original setting, there is a second stop at 1mm, which is "hit" at 0.62s, but it should only be triggered when the ring is pushed towards the hub (positive translational velocity), NOT when being pulled out (negative translational velocity). It really should be a one-way hard stop.

I think the behavior should be independent of "tooth overlap to engage", because the ring will NOT hit the intermediate hard stop when being pulled out.

Please double check the source code to see if the intermediate hard stop is disabled during disengagement process.

[Martin]

Hi Frank,

The hit that you are seeing at 0.62s is the fact that the hard stop location moved because the shift linkage disengaged. I fully agree with what you are saying. The effective hard stop location is (partially) based on whether the clutch is engaged or disengaged. Your point is valid that it should see if the clutch is disengaging not just disengaged.

That hard stop is not turned on or off, it is rather its position that moves.

Thanks,

Martin

[Frank]

Ok, so is this a bug in the dog clutch block? Because the behavior shown in my example is not physical, i.e. when disengaging, it should not hit that intermediate hard stop at 0.62s right? If it is implemented as "moving" the hard stop, shouldn't the hard stop be "moved" all the way to the end of travel (0mm) when disengaging?