Dymola_InlineAfterIndexReduction

Question

This question is related to an issue that I encountered when I was playing with some blocks. Here's the model that I have,

As you can see, there are two kinds of connections, the inputs of the first connection(from the top to bottom) is u[1],u[2],u[3], other blocks are quite self-explanatory (all default values, except that startTime = 5 for the step input block).

From my knowledge, the first kind of connection only outputs angular velocity but not angle and angle acceleration(they are both zero), which is a bit not realistic(I'll explain why I did this). The second connection outputs an angular velocity as well.

My problem was that, in the Second connection, the clutch seems working all right(after 5 seconds the clutch is engaged(relative angular velocity w_rel = 0))

However, the first connection behaves quite differently. We can see that they are all flange connections, and angular velocities are all calculated from flange_a/b.phi, so we should expect that there is no angular velocity difference in the clutch no matter what the input (realExperssion1) is. But the interesting thing is that when I simulate the model, the left flange of the clutch is not moving, the right flange is rotating instead. Here're two plots of my results.

Connection1

Connection2

Actually, I should expect to see the flange_a.phi and flange_b.phi are all zero, and then I accidentally removed the annotation __Dymola_InlineAfterIndexReduction = true in the move block, then the model behaves as what I expected. I wound be really appreciated if anyone could help me explain what I saw. Thanks A LOT!

Answer 1

The documentation for the Move model clearly says

The user has to guarantee that the input signals are consistent to each other

In your case, they are not consistent. So I'm not too surprised you get a strange answer. It was not clear to me why you even attempted to go this route. You imply in the message you would explain why, but I certainly didn't understand your motivation. I suspect that the Move model exists to allow the user to provide their own explicit functions for position, velocity and acceleration that Dymola will use during index reduction instead of generating those functions from the underlying equations. Unless you can provide consistent functions, you really shouldn't use this block at all.

You should really be using a source where you specify only one of position, velocity and acceleration. If that isn't possible, then I'm afraid you'll have to explain why so we can try to understand what you are really trying to achieve here.