Choosing a Machine Topology for an Industrial Application

The general industry approach to machine types is quite stable as machine design is now quite well developed.

Induction (Asynchronous)

Most machines are induction machines, they’re inexpensive and relatively easy to control but have poor starting characteristics and fairly poor power factor. Almost all the problematic losses are on the stator which is good, they can be removed easily either by natural, forced air or liquid cooling. They’re robust and can be made fault tolerant with a little thought. They’re good for continuous service applications with constant speed loads like HVAC, machine tools etc. Control of induction machines is thought to be fairly easy.

Synchronous (PM)

Permanent magnet (PM) machines are more expensive than induction machines, have better power factor, starting characteristics and are good for servo applications, they’re also good for high speed if the magnets can be retained properly. Control at low speeds is often fairly easy several arrangements are possible including direct torque control, this is similar to the induction machine, at high speeds consideration needs to be given to de-fluxing the PM machine properly. If a wide speed range above the base speed is needed, a considerable amount of energy may have to be put into the machine to oppose the PM flux and keep the EMFs within the converter limits. PMs don’t like to get too hot, it upsets their domains.

Synchronous (Wound)

Wound field synchronous machines are not as popular as PM machines but have a couple of subtle advantages. The first disadvantage is that a large electromagnet is required to make the same flux as quite a modest PM. The electromagnet will be lossy and the PM wont. These losses may well be on the rotor where they’re hard to remove by liquid cooling. Currents need to be coupled to the rotor perhaps with brushes or a rotating transformer. This may necessitate a diode bridge on the rotor. On the other hand it’s possible to do some neat control work to force unity power factor under almost all conditions. There is no sensitivity to the price or supply of PMs, but iron and copper are still required. Speed is not a problem, if the rotor coils can be retained along with any other components that end up on the rotor to enable brush-less operation, the field weakening region can be large, ten times the base speed is often possible. WFSMs have been used in at least one fully electric vehicle and at least one manufacture of wind turbines uses them.

Reluctance Machines (Synchronous and Switched)

The synchronous reluctance machines have poor characteristics all-round including the possibility of a rather complicated rotor setup depending on the kind of SynRM but at least they run from a standard 3 phase 6 switch inverter.

The switched reluctance machines are the most troublesome of all radial field machines. They’re the easiest to explain but the hardest to control by far. They have several advantages, no PMs, very simple rotor, can go quite fast without too much difficulty on the machine side. The drive is bespoke which deleteriously increases the cost, each phase often has a full H bridge. The power factor is miserable, and the starting characteristics are nothing to write home about. The control is very troublesome to set up, the noise, vibration and harshness (NVH) are the worst of all machines especially the torque ripple and the resulting conducted acoustic vibrations through the stator. It’s possible to change the drive strategy to use a sine wave drive and a 6 switch inverter to run an SR in a variety of arrangements rather than the classical square wave drive but these don’t fully ameliorate the NVH problems and they usually give away torque density along the way. Consider the area under half of a square wave compared to the area under half of a sine wave. In the SR machine’s favour it is probably the most robust. The best way to kill one is to break its resolver.

Axial Flux Machines

Axial flux machines have not received nearly as much commercial attention, this may be because the development of the machine often requires 3D FEM simulations which have only become feasible with reasonable computing resources in the last decade or so. These may be synchronous, asynchronous or SR. The shape of the machine is conducive to certain applications (in wheel hub motor for example). Another perceived difficulty is with regard to the mechanical forces the machine experiences when operating, but these may be unfounded. Certainly there is scope for ease of modularity in the axial flux machines and this may make winding much less expensive (no needle winding). This ease occurs because the flux doesn’t have to travel out of one modular segment and into the next hence the connection between segments is not so critical — flux is not expected to pass through. In a modular radial machine, flux usually passes across modular segments and the mating of segments _must_ offer a low reluctance. An axial flux machine might find applications in megawatt level ship propulsion and electrical generation for on-board systems for example. Examples of high-speed axial flux machines are probably very few in number.

Conclusion

In conclusion, a commercial designer is very probably going to use a radial flux machine. If cost is the key thing, use an induction machine. If weight/volume (i.e. torque density) are very important and the cost is not so critical (and the temperature is below about 300 C) or a high performance servo is needed, it might be worth using a PM machine. Induction machines are also used as servos though. In some very special situations a case may be made for one of the others (especially the wound field machine) but these are the exceptions not the rule.

If someone is pushing a fancy machine topology, their either an academic and it’s their research area/life’s work or they’re a company who are using the fancy topology as their USP.

Axial flux machines might become more common over the next couple of decades in applications where they’re considered advantageous but this can only happen if the price of their design and manufacture is competitive with a radial flux machine that will perform the same task sufficiently well. Since modularity is one of their big advantages, I wouldn’t expect to see them making a mark in high speed applications because the mechanical implications of modularity are not usually compatible with high speed work.