QUESTION:
I am attempting to simulate the operation of an electric motor by rotating (repositioning) the rotor in successive transient restarts. The nodal spacing in the circumferential direction along an intentionally modeled "sliding interface" is the airgap between the rotor and stator is uniform (1 degree). For each restart I reposition the rotor so that sliding interface nodes are coincident, use CPINTF to couple AZ, and increment time by an amount corresponding to the desired rotor angular velocity. I find the following:
1) Use nonlinear BH data, steady state: flux lines contiguous as intended - OK
2) Use linear permeability, transient simulation with incrementally positioned rotor: fluxlines contiguous as intended - OK
3) Use nonlinear BH data, transient simulation with incrementally positioned rotor: fluxlines are NOT contiguous - ERROR!
When I switch to CEINTF (use constraint equations in the sliding interface rather than couple degree of freedom sets), then case 3 above shows contiguous fluxlines on the sliding interface, so I have a workaround. However, I am still curious to know why CPINTF does not work in the moving rotor transient when nonlinear materials are present.
ANSWER:
This from development:
I studied this model and I can see that the coupling equations are created OK and are consistent from the start to finish of each solve command. What happens is that in the second load step the coupled DOF do not have the same AZ solution values. The incremental AZ solution values are coupled but not the total solution values. It turns out that this is expected since in a nonlinear problem coupling cannot properly keep the total AZ solution values together for each coupled set of nodes. Constraint equations can do so with the extra constant term involved with CEs.
In a nonlinear analysis, newly introduced CPs (and therefore CPINTF) simply couple the upcoming *incremental* AZ (what we are solving for in a nonlinear analysis using the Ne
|
QUESTION: I am attempting to simulate the operation of an electric motor by rotating (repositioning) the rotor in successive transient restarts. The nodal spacing in the circumferential direction along an intentionally modeled "sliding interface" is the airgap between the rotor and stator is uniform (1 degree). For each restart I reposition the rotor so that sliding interface nodes are coincident, use CPINTF to couple AZ, and increment time by an amount corresponding to the desired rotor angular velocity. I find the following: 1) Use nonlinear BH data, steady state: flux lines contiguous as intended - OK 2) Use linear permeability, transient simulation with incrementally positioned rotor: fluxlines contiguous as intended - OK 3) Use nonlinear BH data, transient simulation with incrementally positioned rotor: fluxlines are NOT contiguous - ERROR! When I switch to CEINTF (use constraint equations in the sliding interface rather than couple degree of freedom sets), then case 3 above shows contiguous fluxlines on the sliding interface, so I have a workaround. However, I am still curious to know why CPINTF does not work in the moving rotor transient when nonlinear materials are present. ANSWER: This from development: I studied this model and I can see that the coupling equations are created OK and are consistent from the start to finish of each solve command. What happens is that in the second load step the coupled DOF do not have the same AZ solution values. The incremental AZ solution values are coupled but not the total solution values. It turns out that this is expected since in a nonlinear problem coupling cannot properly keep the total AZ solution values together for each coupled set of nodes. Constraint equations can do so with the extra constant term involved with CEs. In a nonlinear analysis, newly introduced CPs (and therefore CPINTF) simply couple the upcoming *incremental* AZ (what we are solving for in a nonlinear analysis using the Newton method K*dU = R-F ) and enforce them to be the same. The *total* AZ is what it was and is not affected by these new CP. With CEs, however, we internally introduce a constant term in the equation equal to the amount that the AZ of the newly constrained AZ's differ, thereby enforcing the *total* AZ to be the same.Since CPs do not have this concept of a "constant term", there is no way to do this with them. (Note though that in some applications coupling the DOFs from "here on out" is a desired effect hence we do not block it!). |
||
|