Surface PM Motor FEM: Flux, Back-EMF, Cogging and Torque

Solve a nonlinear 12-slot, 10-pole motor cross-section by finite elements. Explore flux lines, phase back-EMF, cogging and loaded torque as the rotor turns.

#back EMF #cogging torque #convergence #finite elements #FreeFEM #M270-35A #PMSM #torque ripple

Explore the magnetic field of a 12-slot, 10-pole surface permanent-magnet synchronous motor directly in its cross-section. This PMSM finite element analysis tool is for students learning magnetic circuits, engineers comparing motor proportions, and builders investigating a drone, e-bike or servo motor. It solves the nonlinear magnetic field of the actual slotted geometry using quadratic finite elements.

Set the stator outer and bore diameters, slot depth, tooth width and slot opening first. The rotor dimensions follow from the bore, air gap and magnet thickness. The pole-arc fraction controls how much of each 36° mechanical pole pitch carries a magnet. Set remanence for the magnet at its intended operating temperature; a grade name alone does not specify that temperature. The steel uses the SIWATT M270-35A magnetization curve, continued toward an assumed saturation polarization of 2.0 T. This is typical for the grade, not a saturation value stated in the sheet. The knee indicator shows where iron reluctance becomes significant.

Each tooth carries a concentrated coil. Four coils are connected in series per phase, with one parallel branch. Enter turns per coil and peak phase current, not RMS current. Zero current gives the open-circuit case. Zero current advance gives motoring q-axis excitation; positive advance adds a negative d-axis component. Speed sets the back-EMF calculation, while these magnetostatic field and torque solutions are independent of speed.

Press play or drag the point on the rotor to move through one electrical period, equivalent to 72° mechanical rotation. Red and blue magnets identify outward and inward magnetization. The phase-labelled coil sides show current into or out of the section. Colours show computed flux-density magnitude and contour lines follow the magnetic vector potential. During playback the field is blended between solved positions on the mesh itself. The accompanying markers identify the same angle on the torque and back-EMF curves.

Compare the mean torque from winding work with the mean air-gap stress torque. Their disagreement is a useful numerical diagnostic. The torque constant uses peak phase amperes; the voltage constant uses line-to-line RMS EMF per mechanical radian per second. The displayed induced EMF follows Faraday's minus-sign convention; the counter-voltage term in a motor circuit equation has the opposite reference direction.

For cogging torque simulation, open Advanced to set gap resolution independently of the bulk mesh. Select one of four convergence checks to halve the gap spacing or double the rotor samples for cogging or loaded ripple. The check calculates a fresh torque sweep and reports its change from the baseline. The rotor turns with its own mesh, so the stator's mesh stays the same at every position. Quadratic elements are used throughout; the default gap spacing is half the air-gap width. Repeat both refinements before relying on a small ripple difference. An engine-team FEMM 4.2 comparison of this geometry and steel curve gave about 0.122 N·m peak-to-peak cogging, approaching 0.118 N·m with refinement. This is an internal numerical comparison, not a published or measured motor reference. The ideal slotless air-gap fundamental is only a comparison. This online motor FEA does not calculate end effects, skew, iron or magnet losses, winding temperature, demagnetization, inverter limits, structural strength or efficiency. It is a two-dimensional magnetic design study, not a complete motor rating.

How it is calculated

FreeFEM solves the whole 2D cross-section with quadratic elements (order = 2) throughout, using the magnetic vector potential A_z, with B = (∂A_z/∂y, −∂A_z/∂x), nonlinear steel reluctivity and A_z = 0 on an exterior air circle. Each arc magnet is one radially magnetized region with μᵣ = 1.05. Bore and rotor-side air circles share the fine gap mesh; magnet arcs follow the same chord boundaries, avoiding artificial interfaces. The rotor (every region inside a circle midway across the gap) is drawn at 0°, meshed once and turned to each position; the stator's mesh stays the same, and a band in the middle of the free gap joins the two.

Each coil side carries J_z = ±NI/S_side. A 3 × 3 Gaussian area quadrature averages A_z across each gross side; signed series contributions give ψ = NLΣ(Ā_positive − Ā_negative). Four tooth coils are in series in each phase. Separate loaded and open-circuit FEM sweeps solve a 60° electrical sector. The identity ψA(θₑ+60°)=−ψB(θₑ), phase permutation and half-wave sign reversal reconstruct a full electrical period. An extra independently solved sector endpoint checks this reconstruction.

The displayed induced EMF is −ωₑ dψ/dθₑ, obtained by differentiating the resolved Fourier series. Both torque curves use the whole-gap Arkkio stress integral. The preferred mean torque is Σ∮i dψ divided by the mechanical angle turned; the closed cycle has zero net stored-energy change. The ideal magnet-covered gap field is Br h_m/(h_m+μᵣg), with its rectangular-pole fundamental drawn for comparison.

The M270-35A curve comes from SIJ Acroni's 2023 datasheet: page 2 is digitized to about ±0.01 T in polarization, and page 1 supplies the stated typical points at 2500, 5000 and 10000 A/m. The steel_bh table converts J to B using B = J + μ₀H and supplies FEM in [B; H] order. Both stator and rotor steel regions specify Js = 2.0 T, so the continuation beyond the last knot approaches this saturation polarization while B = J + μ₀H continues to rise. The sheet does not state a saturation polarization: 2.0 T is an assumed typical value for this grade, not a value taken from the datasheet.

Advanced → Check ripple convergence runs one additional full peak-to-peak torque sweep: either halve the gap boundary spacing or double the rotor-position samples, for cogging or loaded ripple. The resolution-status result reports whether the selected change exceeds 5%; the named steps “Refined peak-to-peak torque” and “Refinement change (%)” in How it’s calculated give the numerical comparison. The 5% status is for that one refinement only; repeat both spatial and angular refinements before treating the waveform as resolved. These full torque checks run on demand. Field-solve time is separate from the formula execution limit; models of one geometry are solved as a series on one mesh.

Assumptions

Twelve identical slots, ten identical surface magnets, concentric rotor, twelve tooth coils, four coils in series per phase and one parallel branch. Balanced sinusoidal currents; positive current advance weakens the d-axis field. Magnet recoil permeability is 1.05. The shaft is nonmagnetic. Uniform axial section and stack length; no skew, end leakage, hysteresis, eddy currents, temperature dependence, drive circuit or mechanical losses.

Where it applies

Gap resolution divisions sets the spacing of both gap outlines to gap/n_gap independently of the bulk mesh. The default is 0.5 mm in a 1 mm gap (n_gap = 2), with quadratic elements throughout; adapt = 0 preserves the specified mesh. The bulk default is 2 mm, and the single-position flux-linkage convergence property refines only that bulk mesh. It does not demonstrate torque convergence.

The rotor turns with its own mesh. Its regions are meshed once and turned to each position; the stator's mesh stays the same, and a band in the middle of the free gap joins them. Twelve positions per 60° electrical sector give six samples per cogging cycle. Advanced offers four separate full torque convergence checks: halve the gap spacing or double the rotor-position samples, for cogging or loaded ripple. Each reports its refined peak-to-peak result and change without altering the base curves. At the defaults, the computed cogging peak to peak is 0.11919 N·m; halving gap spacing gives 0.12003 N·m (0.71% change), and doubling rotor samples gives 0.12051 N·m (1.11% change). Loaded peak-to-peak ripple changes by 0.26% and 0.28% under the same respective refinements. Repeat both spatial and angular refinements when changing the design. A single comparison within 5% is not proof of asymptotic convergence. A nonzero mean cogging torque is discretization bias, not usable shaft torque. Percent ripple and relative torque agreement are blank below 0.02 N·m mean torque.

Recalculated with finer numerics throughout (check_accuracy), the defaults move by about 3.7 % on cogging, 1.9 % on the tooth peak and under 0.3 % on ripple, EMF and mean torque: the mesh's numerical error is about 4 % on cogging and under 0.5 % on torque and EMF. The engine team reports an independent FEMM 4.2 run of this same geometry and B–H curve at about 0.122 N·m peak-to-peak cogging, approaching 0.118 N·m as its mesh is refined. This is a user-supplied internal solver comparison, not a published reference or an experimental measurement. The present default result is calculated afresh; no parameter is fitted to that comparison.

Cubic periodic interpolation draws torque and linkage curves; EMF uses a Fourier derivative. The field picture is drawn on the finite-element mesh itself, so material edges stay sharp. Between the solved positions a played frame blends the two nearest solutions and turns the rotor's mesh to the angle between; it solves nothing new. Iron peaks sample interiors, not sharp-corner singularities; 1.6 T is a knee indicator, not a hard saturation limit. The typical 50 Hz steel curve approximates a single-valued magnetostatic law, not hysteresis or core loss. Beyond the tabulated range the assumed Js = 2.0 T controls the continuation; high-excitation results depend on that assumption. The analytic comparison neglects slots, leakage and finite iron reluctance. Independent published checks validate the procedure and winding topology, not a measurement of this motor.

References

  • D. Meeker, FEMM 4.2 User's Manual, 2015, magnetostatics and torque integration: https://www.femm.info/Archives/doc/manual42.pdf
  • SIJ Acroni, SIWATT M270-35A technical data sheet, 2023, pp. 1–2: https://cms.sij.si/storage/302/SIWATT-M270-35A.pdf
  • D. Meeker, dipole torque benchmark, 2018, including TorqueBenchmark.nb results: https://www.femm.info/doku/doku.php?id=torquebenchmark
  • Nabla E01, independent 12-slot/10-pole FE example, winding factor only; its motor geometry and steel differ: https://www.nabla-fem.com/examples/E01/index.html
  • Design background: J. R. Hendershot and T. J. E. Miller, Design of Brushless Permanent-Magnet Machines, Motor Design Books, 2010, ISBN 9780984068708. Publisher: https://www.motordesignbooks.com/
  • Design background: J. Pyrhönen, T. Jokinen and V. Hrabovcová, Design of Rotating Electrical Machines, 2nd ed., Wiley, 2014. DOI: 10.1002/9781118701591

Inputs

InputSymbolUnitDefault
Show numerical and secondary geometry inputsadvancedfalse
Stator outer diameterD_outmm100
Stator bore diameterD_boremm55
Air gapgapmm1
Magnet thicknessh_pmmm4
Nonmagnetic shaft diameterD_shaftmm15
Slot depth
Measured radially along the slot centreline.
h_slotmm15
Tooth-tip depthh_tipmm1.1
Tooth body widthw_toothmm6.4
Slot openingw_openmm2
Stack lengthL_stackmm50
Bulk mesh sizemesh_sizemm2
Turns per tooth coilN_coil20
Peak phase currentI_peakA15
Current advance from q-axis
Positive advance adds negative d-axis current; zero is motoring q-axis current.
gamma_ideg0
Magnet pole-arc fraction
Magnet arc divided by the 36° mechanical pole pitch.
alpha_pm0.8
Magnet remanenceBr_pmT1.2
Speedn_speedrpm2000
Rotor angle
Mechanical angle. One electrical period is 72° of it.
theta_mdeg7
Positions per 60° electrical sector
Twelve positions give six samples per cogging cycle; increase for ripple convergence.
n_sector12
Gap resolution divisions
Both gap outlines use the air-gap width divided by this number.
n_gap2
Check ripple convergence
Runs one full refinement comparison; the main curves keep the selected resolution.
resolution_modeNo refinement check

Results

With the default inputs above, the calc gives:

ResultValue
Mean torque · winding work4.265 N·m
Mean torque · air-gap stress4.268 N·m
Loaded ripple (%)6.6
Loaded ripple · peak to peak0.28 N·m
Cogging torque · peak to peak0.12 N·m
Open-circuit phase EMF RMS28.39 V
Phase fundamental flux linkage0.0381 Wb
Kt per peak phase ampere0.286 N·m/A
Ke · line-line RMS / mechanical speed0.234 V·s/rad
Tooth interior peak1.783 T
Stator yoke peak1.581 T
Ideal magnet-covered gap field0.95 T
Work–stress disagreement (%)0.07
Saturation indicatorIron above knee
Torque resolution statusNo refinement check selected

Checked against sources

Each case below was calculated by this calc and agrees with the value from its source (within the tolerance where one is shown).

Independent FEMM dipole torque · 30°

Source: Meeker, FEMM TorqueBenchmark.nb, page 1: 30° gives 0.500019 N·m (analytical 0.5). Same 10 × 5 × 20 mm bar, Hc=1 MA/m, μr=1 and 1 T external field. This calc generates the uniform field through a finite circular magnetized shell with an analytically corrected amplitude. A 1% torque tolerance allows this separate mesh; no parameter is fitted to the reference result.

Inputs: the default inputs above.

ResultSourceThis calc
Independent dipole benchmark: 30°0.5 ± 0.005 N·m0.4997 N·m

Independent FEMM dipole torque · 90°

Source: Meeker, FEMM TorqueBenchmark.nb, page 1: 90° gives 1.00004 N·m (analytical 1). Same bar and field as the 30° benchmark. Checks the magnet source and Arkkio torque reader; it is not a measurement or calibration of the 12-slot motor.

Inputs: the default inputs above.

ResultSourceThis calc
Independent dipole benchmark: 90°1 ± 0.01 N·m0.9996 N·m

Independent 12-slot/10-pole winding factor

Source: Nabla E01, Winding and materials: twelve tooth coils, four series coils per phase, q=0.4 and winding factor 0.933. Compare both the library winding and the actual phase/sign assignments used in this field model; the source's motor geometry and steel differ. https://www.nabla-fem.com/examples/E01/index.html

Inputs: the default inputs above.

ResultSourceThis calc
Fundamental winding factor0.933 ± 0.0010.933
Fundamental factor of the model's actual coils0.933 ± 0.0010.933

Related Resources

Responses

Your email address will not be published. Required fields are marked *

L ading...