How software assists designing motors and generators

Most of the 2D FEA software is good for "quick-and-easy" solutions demanded by shorter design schedules. Although 3D software allows examining more complex models, doing so is not feasible when there is only a day or two to analyze electromagnetic and thermal performance.

QuickField, multiphysics simulation software, has proved efficient at our company for modeling electrical, electromagnetic, thermal, and mechanical problems in electric motors and generators. Discussing techniques for solving a range of design problems highlights a few notable features in the software. For instance, Label Mover assists with parametric analysis of any model created in the software. Parameterized geometry and material libraries are useful when running multiple analyses in the search for higher motor efficiencies. QuickField works well at this task by letting users parameterize almost everything, including geometry, boundary conditions, and loads. It also works after analysis when users see characteristics with several values that need examining.

Dc and synchronous electrical machines frequently serve as examples in magnetostatic FEA models. Magnetostatic refers to operating at constant rotor speed. When investigations focus on time-domain phenomena, transient electromagnetic analysis can be done as well.

Optimizing motor laminations, another analysis task, aims to minimize local saturation, provide sufficient space to insert stator and rotor coils, and improve manufacturability for highvolume production. The optimization solves easily after applying armature and field currents, boundary conditions, and magnetic properties of the motor components. Real challenges show up while modeling the influence of interpoles on commutations, which rely on a properly described current function in the commutating armature coils. This modeling task applies to medium and large dc motors (more than 100 kW) for industrial and traction drives. QuickField also models conventional dc motors, those excited with a stationary field, and reversed brushless dc motors, those with magnets on the rotor.

Many high-efficiency motors for machine tools use a three-phase permanent-magnet synchronous design. It requires a transient electromagnetic solver with a time-domain postprocessor to study torque characteristics such as the effect of magnetic flux harmonics on cogging torque. Ordinarily, eddy currents are low in the rotor core and underneath the magnets, and analysis can be done using magnetostatic FEA with gradual current change in each stator phase.

Magnetostatic features in QuickField analyze demagnetization characteristics, torque constants of dc motors, and steel saturation at no load and full load. Defining these is essential to meet a 10% engineering accuracy figure.

Another use for the magnetostatic module is the full-load operation of a 150-kW, four-pole synchronous generator. To calculate output power at a given rotor current, users can solve the inverse problem by calculating torque for a synchronous motor at a preliminary estimated rotor current and load angle (rotor position). And to optimize the lamination layout, find the steel saturation and output torque. Such simulations also let users predict the number of rotor amp-turns required to meet a specified output power.

In addition to magnetostatics, motor characteristics are also influenced by nonelectric phenomena such as heat. The software has a coupled heat-transfer module that uses the same mesh as the electromagnetic module. Coupling heat with electromagnetics verifies that rotor and stator-winding temperatures are below the maximum allowed insulation temperature. Heat-source coupling is usually done by transferring Joule and iron losses to the heat-transfer model.

The coupling feature is also useful optimizing an amortisseur winding. The task requires running transient electromagnetic FEA due to the eddy-current nature of the motor-starting currents and their damping effect on stator-winding harmonics. Coupling heat-transfer and mechanical-stress analyses would also be necessary to evaluate amortisseur-winding strength. This is especially true when a three-phase generator is used in single-phase operation, and when a motor operates during start-up and braking. The latter may cause thermoelastic stresses inside dampening bars and welded joints. QuickField transfers numerical data from the electromagnetic module to the one for heat transfer, and also to a structural module that calculates thermal stresses and strains that identify mechanical weaknesses.

A challenge for any FEA program is to simulate switched reluctance (SR) motors due to their inherent "switching" nature. Other SR features that make simulations difficult include a relatively small air gap, saliency of rotor and stator poles, and highly saturated tips on commutating poles. The software's transient electromagnetic FEA is needed to refine winding and lamination designs. Initial designs could be generated by lumped-circuit simulators. The Label Mover feature simplifies some of challenges by automating mesh rebuilding to simulate gradual rotor motion with the stator-current change. The feature reduces preprocessing time. The software can generate a mesh of 1 to 2 million nodes to capture the steep spatial gradient of stator-pole saturations in SR motors.

Most motor and generator design problems can be handled by electromagnetic analysis, either static or transient, followed by heat-transfer and stress analysis when needed. Heattransfer analyses, however, depend heavily on assigned convection boundary conditions, normally acquired from laminar or turbulent fluid-flow analyses. Setting up thermal models of motors or generators means obtaining film factors from fullload tests for given machine frames sizes, rotor speeds, fan types, and types of enclosure.

Noise and vibration, other design considerations, are challenges for ac-induction motors. Analytical methods cannot always predict vibroacoustic responses caused by electromagnetically induced vibrations. However, FEA software helps minimize their amplitudes or change harmonic spectra by letting engineers evaluate different slot combinations, winding configurations, and lamination geometries. Such studies are usually based on modal analysis combined with electromagnetic FEA. These approaches are good in theory, but results are sensitive to manufacturing deviations from nominal-and how the motors are mounted.

Correlations can be established between FEA models and actual machines after measuring noise and vibration. Correlations can be used to study the above characteristics for designs having a similar mechanical layout. Accuracies for predicting sound pressure levels turn out to be within 1 to 2 dB for NEMA-frame ac-induction motors. The noise level applies to transient harmonic spectra during start-up, full load, and braking, as confirmed by measurements of the waterfall characteristics of rpm-triggered vibration and sound.

FEA software has become an essential tool for motor and generator design. The programs predict whether or not designs will meet specifications before prototyping. And when simulated designs don't measure up, users can modify the FEA models and solve again.

Users with enough experience can find answers within 10% of physical tests. The error is due to assumptions, simplifications, and approximations made during simulations. Ten percent may be adopted as the norm when you take into account the accuracy of material properties, integrity of the insulation system, and manufacturing tolerances and deviations related to air-gap eccentricity.

QuickField comes from
Tera Analysis Ltd.
(887) 215-8688
quickfield.com
A free student version is also available.

— Mikhail Khanin

Mikhail Khanin is a senior electrical engineer with Baldor Electric Co. (baldor.com), Ft. Smith, Ark.