"Repmag" Part III

Further work required for a physical prototype

The modelling done so far, especially the magnetostatic/dynamic modelling described in Part I, is encouraging, and confirms the basic concept of significantly less repulsive force between adjacent magnets within the shield than outside of it. However, I would want to do further modelling before building a physical prototype. This would include larger devices, and ferrite magnets as well as neodymium-iron-boron ones. One obvious variation should also be modelled: the case where adjacent magnets all had alternate polarities, instead of the same polarity, so that they would attract instead of repelling.

Further work should also be done on the energy delivered by a magnet entering the shield separated from its neighbours, compared with the energy consumed when it exits the shield with closely adjacent neighbours. This is an important issue that needs to be fully resolved.

Various airgaps between the magnet and shield faces should also be modelled. It seems likely that very small airgaps will be desirable in a prototype, thus requiring very small flexibility/deformation of components.

Eddy currents

One reason for looking at ferrite, and also bonded, rather than sintered NdFeB magnets, would be to combat a potentially serious problem that can already be foreseen:— eddy current losses. In a device like this, magnetic flux density will vary greatly as the device operates, in the magnets themselves, and also in the shield. That means that eddy currents, and the resistive I²R losses they cause, will occur in all these components when running.

The traditional methods of reducing eddy currents to tolerable levels are to laminate conducting components, or else to substitute them with non-conducting components (such as ferrite).

Drawing 1

Repmag drawing 1

The "multi-layer" 2D drawing above is a start for a possible physical prototype. It was drawn to this stage mainly to show that, as required for this approach, the spider and its components can have a smaller radius than the distance from its pivot to the nearest magnet. Not all components are shown, such as bearings, framework etc.

The magnets (red) are attached to shafts, all pivoted to a fixed point at the black solid circle. The shafts carry linear motion bearings (purple) which are carried in equally-spaced pivots on the spider (green), which has an offset fixed pivot at the green solid circle.


Drawing 2

Repmag drawing 2

Here is another "multi-layer" 2D drawing of a slightly different approach which would give better rigidity (less deformation of components), although it would require the magnets themselves to have radial holes.

The magnets (red) have internal ball spline bearings so they can slide without rotation along the splined arms of a spider (blue) centered at O. At their outer radii the magnets carry rollers which bear against an offset circular track (green) centered at P. The shield (grey) is also centered at P.

Shield lamination options

Shield lamination

Various lamination options for the shield are shown above.

The best simple option for minimum reluctance of the magnetic circuits, i.e. Alt. A, with wedge-shaped laminations, is almost certainly impractical. Neither Alt. B or Alt. C are desirable because of their greater reluctance over the complete magnetic circuit between magnet faces (across the many inevitable small gaps between laminations, for these alternatives). Something like Alt. D could be the best option — although somewhat difficult, it should be achievable.

"Repmag" Part II

As I said last time, I have done some initial mechanical design on my "Repmag" idea. The ultimate goal of course would be to prove and develop the design sufficiently to enable a successful physical prototype to be built. However, for now, I'll post a bit more about what I have done on it so far.

Magnetic forces replaced by spring forces

silux model with spring forces replacing magnetic forces

Here is an early silux model I made to verify the basic concept that stronger repelling forces between adjacent magnets outside of the shield, and weaker repelling forces inside the shield, would cause the rotor assembly to turn as a whole. It was assumed that the energy gained by each magnet entering the shield would be balanced by the energy lost by another magnet exiting the shield, so these entry and exit forces were not modelled. (I will have more to say about that later).

In this model compression springs, substituting for forces of magnetic repulsion, are added between adjacent objects representing the magnets. All springs have a natural length of 0.04m, and a minimum compressed length of 0.01m. Outside of the shield, the springs (black) can exert maximum force of 9N, giving spring constants of 300N/m. Inside the shield, the springs (pink) are weaker, at 6N and 200N/m.

The macro stops the model every 30º of clockwise spider rotation. Then the spring which has exited the shield is changed manually from a weaker to a stronger one, and vice versa for the spring which has entered the shield. The model is then re-started for the next 30º etc.

This model, with light magnets and rollers, but with an artificially heavy 10kg spider, is identical to the one shown in Part I previously, apart from the added springs. Starting with all components at rest, with no other torques or forces, it gave the following graph of spider rotational speed vs time:— 

Graph of Repmag spider rotational speed vs time

In this graph, the spider has undergone smooth acceleration from rest, completing just under seven revolutions, reaching 10.246 radians/sec (nearly 100 rpm) in 8.532 seconds.

The spring parameters were chosen only as "reasonable" values, and could have been a bit stronger. This "changing-spring" driven model delivers energy (½Iω²) of 0.5 × 0.0375 × 10.246² = 1.9684J in 8.532 seconds. This gives a power output of 0.231 watts, only about half the 0.507 watts obtained last time from the magnet driven model.

Next time I'll look at more details for a possible physical prototype.