AI-powered electric motor design

1. Example of an automatic motor design workflow

Traditional electric motor automatic optimization typically requires users to define the exact ranges for each design variable, necessitating some preliminary studies on the motor’s geometry to determine sufficient bounds for the design variables. Too narrow bounds increase the risk of suboptimal solutions, while overly wide bounds lead to extensive searches and potentially overlook optimal designs due to the vast search space. Too wide optimization variable bounds will also lead to a large percentage of designs with inconsistent geometry dimensions. To address these challenges, basic motor sizing rules are employed, and ratio-based design variables not directly tied to specific motor geometry dimensions are introduced. This approach effectively narrows the search space while minimizing the risk of overlooking optimal design candidates.

This article introduces a MATLAB code example, utilizing the dedicated MotorXP-MATLAB API, to automate the design process for motors and generators with surface-mounted permanent magnets. By integrating basic motor sizing rules with advanced optimization algorithm, the code optimizes motor design by minimizing the motor weight and maximizing efficiency under user-defined operating conditions. The code offers versatility, allowing users to automate specific motor or generator designs, explore various design spaces, or develop customized routines for automation of the design workflows.

Key motor sizing rules implemented in the MATLAB code include:

• Magnet Volume Calculation
• Tooth Width and Slot Pitch Ratio
• Relationship between Stator Back Iron Depth and Tooth Width
• Relationship between Rotor Back Iron Depth and Stator Back Iron Depth
• Magnet Arc Angle

2. From motor sizing rules to ratio-based design variables

This section provides a detailed description of the motor sizing rules implemented in this example, along with corresponding design variables listed in Table 1 below.

Table 1. Default motor sizing design variables

1. Magnet Volume Calculation:

The volume of the magnet material required to build the motor is determined based on the approach described in [1]:

magnetVolume = Cν ·
targetPower
f₁ · B_r · H_c

where:

• targetPower – the output power the motor is optimized for
• f₁ – operational motor frequency
• B_r – residual flux density of the magnet
• H_c – coercivity of the magnet
• Cν – magnet volume coefficient (refer to Table 1)

2. Tooth Width and Slot Pitch Ratio:

The relationship between the areas occupied by stator teeth and slots is defined as a ratio between the tooth width angle and slot pitch angle (as show in Figure 1):

ratio_toothWidth =
toothWidthAngle
slotPitchAngle

3. Stator Back Iron Depth:

The relationship between the stator back iron depth and tooth width is determined by the magnetic flux carried by these parts of the motor. For concentrated windings (coil span equals one), the relationship is expressed as:

statorBackIronDepth = coef_statorBackIronDepth · toothWidth

For distributed windings (coil span higher than one) statorBackIronDepth also depends on the number of stator slots (Ns) and number of pole pairs:

statorBackIronDepth = coef_statorBackIronDepth ·
toothWidth · N_s
4 · nPolePairs

4. Rotor Back Iron Depth:

Ideally, the rotor back iron depth should equal the stator back iron depth since they carry the same magnetic flux. However, some variation is allowed to consider leakage flux, defined as a ratio between the rotor back iron depth and the stator back iron depth:

rotorBackIronDepth = ratio_rotorBackIronDepth · statorBackIronDepth

5. Magnet Angle:

The magnet angle, representing the angle of the magnet arc in electrical degrees, indicates the portion of the rotor pole occupied by the magnet. Gaps between magnets help reduce magnet flux leakage. Note that a full rotor pole pitch corresponds to 180 electrical degrees (refer to Figure 1).

To complete the proposed motor design workflow setup, the user should also specify the outer and inner diameters of the motor, lamination length, and air gap. These parameters can either be fixed values or defined as design variables within a specific range for inclusion in the optimization process. Additionally, users can specify desired slot opening dimensions and utilize fixed magnet weight for all design candidates.
During the execution of the proposed design workflow, the optimization algorithm generates various combinations of design variables, each corresponding to a different design candidate. An iterative search is then conducted for each design candidate to find the motor geometry satisfying the specific combination of design variables and constraints. In cases where the search leads to erroneous motor geometry with inconsistent dimensions, predefined geometry limits interrupt the search, excluding the corresponding design candidate from the optimization process.
For a more detailed description of the proposed automatic motor design workflow, refer to the MotorXP documentation, and the full source code is available with a free one-month MotorXP trial.

3. Case study

The proposed automatic motor design workflow is applied to design an electric bike traction motor. The input data and requirements for the motor are as follows:

• Outer diameter: 70 mm
• Motor length: ≤110 mm
• Motor weight (active materials): < 2kg
• Air gap: 0.6 mm
• Maximum speed: 8000 RPM
• Maximum supply frequency @8000 RPM: ≤ 1000 Hz
• Rated DC voltage: 48V
• Iron core material: M19, stacking factor ~0.95
• Magnet material: N42UH
• Slot fill factor: ~0.35
• Winding type: concentrated
• Estimated winding temperature: 80°C
• Estimated magnet temperature: 60°C

The task is to design the motor optimized for operation at 2000 RPM and 7 Nm.
After analyzing the possible slot-pole combinations for concentrated winding, considering that the supply frequency should not exceed 1000 Hz for maximum rotor speed of 8000 RPM, only two optimal slot/pole combinations are identified: 12 slots / 10 poles and 12 slots / 14 poles. There is a useful article on selecting slot-pole combinations for concentrated windings.
With the motor being either of an inner rotor or outer rotor configuration, four potential motor configurations emerge:

• Outrunner with 12 slots and 10 poles
• Outrunner with 12 slots and 14 poles
• Inrunner with 12 slots and 10 poles
• Inrunner with 12 slots and 14 poles

The task is to assess these four motor configurations and determine the most suitable one for the given application. This results in four design spaces to be explored using the proposed automatic motor design workflow. To set up the MATLAB script, the user defines a range only for two motor parameters: the inner diameter of the motor and the lamination length. These ranges, which are provided in Table 2, are selected to maintain the motor weight around 2 kg for the specified outer diameter of 70 mm.

Table 2. User-defined design variables

Additionally, the initial slot opening parameters are set as follows

• Slot opening depth: 1 mm
• Slot opening width: 2 mm

The Pareto plots generated by the optimization algorithm for each design space are presented in Figure 3, providing a comprehensive comparison of the analyzed motor configurations to facilitate an informed design decision, aiming to minimize motor weight and maximize efficiency.

click on image to enlarge

Based on the results, the outrunner motor configuration with 12 slots and 14 poles is selected, and the corresponding initial design candidate is shown in Figure 4. This chosen design candidate will undergo further analysis and optimization.

References

[1]. C. C. Mi and L. Luo, “Analytical design of permanent magnet traction drive motors,” 2005 IEEE Vehicle Power and Propulsion Conference, Chicago, IL, 2005, pp. 5 pp.-, doi: 10.1109/VPPC.2005.1554577.

Related articles

From Simulation-Driven to Optimization-Driven Electric Motor Design

Overview of the automatic electric motor optimization using the MotorXP software. Read article>>

Mystery of the Tesla’s Next-Gen Zero-Rare-Earth Electric Motor: Feasibility Study of Transitioning to Ferrite Magnets for EV Traction Motors

Designing an axial flux permanent magnet motor with ferrite magnets to match the efficiency and power density of the Tesla Model 3 motor using the automatic optimization features of the MotorXP software. Read article>>