Introduction
Manatee e-NVH CAE collaborative platform has been initially developed by EOMYS to support its consulting activities aiming at troubleshooting and solving on electric machine noise problems. Manatee has therefore been successfully applied to predict and mitigate noise and vibration due to magnetic forces on more than 200 electrified systems.
Manatee software is a specialized software solution and its use requires to be trained on e-NVH phenomena. Noise due to electromagnetic excitations is a complex multiphysic process, large discrepancies between test and simulation can be observed if the comparison is not carried with care. This article reviews some points that should be checked in case of discrepancies between test and simulation.
Sound pressure versus sound power level
Comparison between tests and simulations must be carried on comparable figures: Manatee acoustic noise level can be given as a sound power level (SWL), or as a sound pressure level (SPL). SPL measurements are easy to carry but can include significant uncertainties (e.g. background noise, reverberant field, directivity). Using SPL comparison requires more attention. If analytical SPL model is used, the room constant where NVH tests have been carried must be specified to include reverberation effects.
When comparing measured and simulated sound pressure levels, be sure to input in Manatee:
- the right distance from the middle of the e-machine to the microphone
- the right directivity coefficient
- the right room constant for reverberant field
Electrical machines are usually tested at factory where directivity coefficient is not ideal and reverberation field is uncertain, in that case simulation results may differ from experiments with up to +/-10 dB difference.
If you want to estimate your room constant to improve quantitative noise calculations with Manatee, EOMYS can run a specific measurement campaign to quantify reverberation effects in your test facilities.
Sound power level measurements run according to acoustic standards (e.g. ISO 3745:2012, ISO 3744:2010, ISO 3746:2010) are recommended for a valid comparison. In loaded case, sound power level of tested machine must not include background noise and loading machine noise. Intensimetry technique is therefore recommended and EOMYS can help you to carry such NVH measurements.
Mechanical and aerodynamic parasitic sources
When comparing measured and simulated SWL due to electromagnetic forces, be sure that you have also excluded non-magnetic acoustic noise sources (aerodynamic noise, mechanical noise) that might be present in experiments. Note that aerodynamic noise may occur at same frequencies as electromagnetic noise in some specific applications.
Manatee software also makes it possible to import non-magnetic noise such as gear noise.
Outer-borne Vs inner-borne noise
Manatee software includes different modelling levels suitable for early design and detailed design stages. In early design phase based on semi-analytic vibroacoustic models, only air-borne noise radiated by outer structure is included. Experiments always include a part of Structure Borne Noise (SBN) which can explain differences between test and simulations. SBN due to rotor excitation by magnetic forces can be included when using Manatee combined with a 3D FEA mechanical model using Electromagnetic Vibration Synthesis algorithm for instance. The rotor FEA model (in particular rotor bending modes and Rotor Housing Coupling mode) should be fit with experiments for good estimation of SBN. The presence of the rotor may also affect some of the stator modes (e.g. the bending mode of a clamped-free stator), explaining differences between calculated and measured airborne noise if the rotor was not included in calculations.
If emotor is placed in a casing and sound power level calculation under Manatee is carried using outer enveloppe nodes of the casing, simulation results only include structure-borne noise neglecting acoustic leakages. Acoustic leakages can make part of the airbone sound radiated inside the casing, resulting in higher noise levels.
Damping
An important simulation parameter to obtain an accurate value of the absolute sound and vibration power levels due to magnetic excitations is modal damping (from 0.5 to 4% in electrical machines). Damping cannot be calculated numerically, and it depends on several parameters (temperature, resin type, winding technology, mode / frequency etc). Manatee default simulation workflows uses a default average damping value of 2%. In case case, simulation results may lead to 20 
=-12 dB to 20 
=6 dB gaps compared to tests at resonance peaks.
A step by step Experimental Modal Analysis is highly recommended to quantifiy modal damping of your application. When measured damping is used in your simulation, vibration and sound levels accuracy can be brought down do +/-3 dB.
Again, EOMYS can help you to quantify damping in your specific application by special NVH tests.
Structural modes
Discrepancies between simulation and tests can be obtained if the simulated modal basis is not representative of the reality. This can be due to the following issues:
- missing rotor in 3D FEA mechanical model
- 3D FEA mechanical model has not been fit with experiments
- 3D FEA mechanical model fitting has been carried in different boundary conditions than operational ones (e.g. free-free)
- stiffening effect of magnetic pre-stress on structural modes in some specific geometries
- effet of coolant (e.g. oil film or water jacket) on structural modes and damping
Magnetic and geometrical asymmetries
It is known that eccentricities and geometrical / magnetic asymmetries can introduce new resonances due to new magnetic force harmonics, thus changing significantly vibration and noise levels. In particular, eccentricities modulate all pulsating forces with UMP harmonics, which can easily excite different structural modes. If you have simulated a symmetrical machine in Manatee, some resonances may be missing compared to experiments. The following measurements are recommended:
- phase current / resistance / inductance measurements (current unbalance)
- uneven turn distribution for manufacturing constraints
- stator bore radius measurements (non uniform airgap)
- rotor balancing, static & dynamic eccentricity levels (direct mechanical measurement or indirect electrical measurements) including conical eccentricity
- IPMSM rotor magnetization along axial and circumferential directions (non uniform magnetization)
EOMYS can design and run these e-NVH tests as well as check their impact using e-NVH simulations .
Current waveforms
Current waveform is responsible for magnetic excitation harmonics that may differ between tests and experiments, especially when using Manatee with sine supply. Differences can be due to:
- unbalance phase currents
- presence of back emf phase belt harmonics, or Rotor Slot Harmonic (called RSH or PSH) in induction machines
- presence of converter-induced low frequency component such as 5f/7f voltage harmonics
- presence of parasitic harmonics due to faults
In that case, it is recommended to perform measurement of three-phase currents and to apply these current inside Manatee simulation to check the effect on e-NVH.
Data acquisition post-processing
The signals obtained with a Dynamic Acquisition System are generally post processed using specific algorithms such as Short Time Fourier Transform (STFT), RPM extraction and order tracking analysis. These post processing have some specific parameters which can have a significant impact on results. Besides test set-up accuracy, test results should also account for post processing accuracy.
In particular, STFT used in spectrograms requires a trade-off between time accuracy and frequency accuracy. When comparing order levels, a synchronous sampling (not fixed sampling) is advised. Order extraction must be done integrating energy on a given bandwidth. It is advised to do a sensitivity study on these parameters before comparing dB levels from tests and simulation.
Miscellenaeous effects
The following phenomena may also impact the e-NVH of your design:
- temperature (magnet temperature impacts remanent flux and the amplitude of magnetic force harmonics). For a detailed discussion of thermal effects on e-NVH see EOMYS article.
- B(H) curve (in case of high dependency with fundamental frequency, be sure that it is included in magnetic calculations)
- axial magnetic forces due to skewing or axial misalignement
- speed ripple or load fluctuations that may not be included in numerical simulation
- strong electromechanical coupling (e.g. combined effect of centrifugal forces and eccentricities through Unbalanced Magnetic Pull)
- gyroscopic effects for high speed machines, or magnetic circuit deformation under centrifugal forces
- strong rotor vibrations that are not related to magnetic forces, that modulate the magnetic flux
- strong fluid / structure interaction (e.g. underwater electric motors, or water-cooled electrical machines)
Modelling accuracy
Manatee software proposes different modelling levels suitable for early design stage and detailed design stage applications. When comparing Manatee results with experiments, the degree of modelling details should be progressively increased in case of mismatch. In particular, electromagnetic loads should be calculated with electromagnetic FEA. Structural response should be calculated with a 3D mechanical FEA model including rotor and stator that has been fit with experiments. Acoustic response should be caclulated with 3D acoustic FEA or acoustic BEM (case of free field) in the following cases:
- noise issue occurs at “low” frequency where Equivalent Radiated Power model overestimates sound level
- significant acoustic leakage occurs (e.g. through casing where emotor is enclosed)