Axial flux vs radial flux motors

Axial and radial describe the main air-gap flux direction relative to the shaft. An axial machine uses disc-like active surfaces; a radial machine uses cylindrical active surfaces. Rotor and stator counts, excitation, slots, winding and cooling remain independent design choices.

Yokeless adds a narrower construction detail: a shared stator yoke is removed while individual magnetic teeth can remain. YASA describes a segmented-armature approach. An ironless active stator is another option, such as the patterned conductor concept described by Infinitum. Neither label makes all axial machines interchangeable.

Use the same shaft output points and the same allowed thermal conditions for both candidates. If one supplier quotes a cold-start peak and another a stabilized continuous point, a single kW/kg column hides the most important difference.

The table below is our proposed procurement method, not a published ranking. An unknown condition should remain unknown until clarified. Record measured, simulated and requested values in different fields. A simulation can help choose a prototype but is not the same evidence as an instrumented test.

Mechanical power is P = T × 2πn / 60, with torque T in Nm and speed n in rpm giving watts. The familiar rounded conversion is P(kW) = T(Nm) × n(rpm) / 9550. At zero speed, shaft power is zero even while the motor produces holding torque and generates heat.

Illustrative arithmetic: 200 Nm at 3,000 rpm is about 62.8 kW. It is not 200 kW. A hypothetical 20 kg housed motor at this point has 10 Nm/kg and 3.14 kW/kg. If its dedicated inverter and cooling auxiliaries add an assumed 10 kg, the corresponding complete-drive values become 6.67 Nm/kg and 2.09 kW/kg. No real model or thermal rating is implied.

The force lever arm is one reason disc geometry can be attractive. It does not establish a general scaling law for a purchasable motor: allowable diameter, loading, speed, stress, temperature and included hardware all constrain the result. Compare the actual candidates instead of converting a topology name into a density promise.

“Liquid cooled” does not specify the thermal resistance from winding to fluid. A jacket, a cooled stator support and direct-fluid contact have different interfaces. Ask where the hot spot is measured, which temperatures are inferred, and what flow and inlet temperature were used.

An independent EMRAX datasheet separates voltage variants and cooling-dependent output conditions, while its manual sets installation and operation requirements. Use that documentation style as a reminder to request conditions; do not transfer its numbers to another supplier. YASA also describes direct oil cooling for its own architecture, which does not prove all axial motors share that capability.

A concrete document check: the EMRAX 228 version 1.7 sheet says its motors require adequate airflow, including liquid-cooled installations, and warns against complete enclosure. Its front page also notes a product-generation transition. Confirm the exact model revision, installation instructions and available configuration before using a downloaded outline or rating in an integration decision.

For a disc-shaped assembly, ask how the rotor faces, bearings, housing and stator carrier retain clearance over temperature and load. A nominally symmetric dual-rotor concept still needs a stiffness, tolerance and load review. Symmetry is not an inspection record.

Start with mounting datums, thrust and radial loads, speed limits and the controlled assembly revision. Request separate cold dimensional records and evidence appropriate to the hot or rotating condition. This is a proposed release workflow; the correct limits depend on the design and application.

These are decision scenarios, not customer case studies. Use them to identify the next evidence request rather than to predict a successful design. A candidate should leave the shortlist if it fails a hard system constraint, even when a headline density looks favorable.

Separate engineering work, tooling, sample units, inspection fixtures, test reports and recurring units. Include the inverter, feedback, cables and cooling hardware only when they are inside the agreed scope. A frameless kit can shift integration work to your team rather than remove it.

For low-volume procurement, request drawing-release terms, change-control rules, sample acceptance and who owns production fixtures. No universal price premium or NRE saving can be concluded from flux direction alone.

This is a qualitative engineering and procurement framework, reviewed on 20 October 2026. Manufacturer sources establish their own topology descriptions and documentation practices. The examples and decision tables are our analysis, with explicitly assumed inputs; they are not cross-vendor dynamometer results.

No fixed efficiency, weight-saving or torque-density advantage is assigned to either topology. We have no matched YokelessMotor versus competitor test set to support one. The practical next step is to obtain candidate-specific evidence under a common brief and then decide whether prototype testing is justified.