A Physics-ly Impossible Device

I may be wrong, but what I have read and understand seems to imply that this wont work, but whatever really...




Idea for a "kinetic propulser".

Parts:
    Red and Green large figures, large magnets;
    Smaller red figures, magnetic weights;
    Purple figures, adjusters for moving the large magnets;
    Yellow rod, sliding rod connecting weights to dark-green rod (slides through a hole in the green rod or such);
    Dark green shaft, drive shaft from motor (blue).

Assertions:
When spinning in the presense of a magnetic field, the magnetic weights should naturally center themselves;
Likewise, when spinning, the magnets should put enough pull on the weights to slighly displace the rotation.

At each end of the rotation, the weights will both be pulled in the same direction, thus the weights will slide, at the same time, because of the rotation of the shaft they will flip around, and thus be pushed the other way. At the same time, the weights slide, they also excert a force on the magnets, and because the friction is low against the shaft, the force is not really excerted on the motor itself.

Claim:
I will claim that, since I haven't really seen much to show that it wont work (though my physics knowlege is lacking), so I will guess it "might" work.
The device will perform a conversion of torque to a linear propulsive force by excerting a force on matter within the device itself.


Modified propulser

An idea came up for a simple way to modify the device.



This one shouldn't be too hard, just propulsion is based on moving the ring vs. trying to clamp down on it...
This one would probably be more stable as well, as at all points it is repelled by the edge of the ring.

This would likely represent the "idle" phase.


a detail part of the claim was computed by me as an example:
one has 2 1kg weights spinning at around 60/PI rpm on a centered rod at a distance of 1m on each side (estimate, each weight is moving with approx 1Nm force).
one then excerts enough force to move the rod 1 cm. it should continue spinning, but at this time will provide approx 0.99Nm forwards torque and 1.01Nm backwards torque, or cost about 0.10Nm.

as a claim, if the rod were fully extended (eg: like in a flail) and could thus move 100cm, the total backward torque would be about 1.0Nm, at this point I assume, because there are 2 weights, that without additional energy the rpm will have dropped by about half.

I may be wrong here, but this seems to make sense.


Work

Assert:
Work is being done primarily in one direction.
I now realize the device is dependent on this claim.

A force in the direction of the offset will cause a net work in that direction, eg, by the sliding of the weights.
During the same time as the slide, there will come to be an indnetation during the time of the slide, this will cuase work in one direction and a reduction of the velocity (and thus work, in the perpendicular direction).

The weight pushed further will have a higher energy.
As it rotates, this higher energy will continue to rotate and move upwards, will be counteracted by the magnets.

Issue:
Work does not depend on velocity, this assertion fails.
The only way it could hold would be if one assumed that mass dialation might occure and be able to effect net work.