ACOUSTIC NOISE AND VIBRATIONS OF ELECTRIC POWERTRAINS Focus on electromagnetically-excited NVH for automotive applications and EV/HEV Part 4 NVH experimental characterization of electric chains LE BESNERAIS Jean contact@eomys.com Note: this presentation is based on extracts of EOMYS technical training https://eomys.com/services/article/formations?lang=en EOMYS ENGINEERING 2013- www.eomys.com 1
NVH EXPERIMENTAL CHARACTERIZATION OF ELECTRIC POWERTRAIN EOMYS ENGINEERING 2013- www.eomys.com 2
Introduction To fully characterize the vibro-acoustic behaviour of an electrical machine or build accurate simulation models (whatever their type, FEA or analytical), some experiments are necessary Acoustic measurements of Sound Power Level of noise source -> Techniques described in Standards depend on environment The characterization of structural modes can be done based on two main techniques: - experimental modal analysis (EMA) - operational modal analysis (OMA) No standard for experimental modal analysis (EMA) nor operational modal analysis (OMA) EMA: use of external excitation (e.g. hammer, shaker) / OMA: use of internal excitation (e.g. magnetic forces) EMA gives Frequency Response Functions but not OMA The characterization of the excitation forces can be done using the Operational Deflection Shape analysis (ODS) EOMYS ENGINEERING 2013- www.eomys.com 3
Experimental Modal Analysis To identify stator circumferential modes, the number of accelerometers depends on the wave number to be captured To identify basic cylinder modes without torsion one can use radial accelerometers only To identify the circumferential mode m, at least 2m accelerometers are required 2m+1 advised to avoid a sensor being at a node position Impact hammer technique may not be adapted to large electrical machines (large noise to signal ratio) EOMYS ENGINEERING 2013- www.eomys.com 4
To identify lamination main cylinder structural modes it is advised to run the EMA before the full assembly, especially if a high frequency mode has to be studied A high modal density of the full system (incl. windings, frame, cooling system, boxes, etc) can make the modal identification tedious Running EMA along the different assembly steps (welding of ribs, coil insertion, VPI, etc) allows to identify the contribution of each new element to stiffness / mass / damping The boundary conditions of the full assembly may differ from the one of the lamination package alone EMA allows to identify all the structural modes, even those which cannot be excited by magnetic forces EOMYS ENGINEERING 2013- www.eomys.com 5
Operational Modal Analysis The meshing procedure is the same as for EMA but the internal excitation coming from magnetic forces is used Therefore only structural modes that be excited by internal forces are characterized Vibration data can be recorded during run-ups in order to excite the structure over a wide range of frequencies The real operational conditions (current level & load state, boundary conditions, temperature etc) can be used The FRF cannot be normalized to a force magnitude (e.g. m/s 2 /N) except if they are calculated independently of the tests mode 2 mode 3 EOMYS ENGINEERING 2013- www.eomys.com 6
Operational Deflection Shapes ODS consists in visualizing how the structure deflects at a given frequency Far from any resonance, the vibration pattern corresponds to the excitation force pattern At resonance the visualized deflection is the corresponding modal deflection shape The number of nodes, the nature of the vibration wave (rotating / pulsating) and the rotation direction can be observed From [F4] EOMYS ENGINEERING 2013- www.eomys.com 7
From [F4] EOMYS ENGINEERING 2013- www.eomys.com 8
HW & SW minimum set-up 1 radial accelerometer in the middle of the stator lamination (or rotor yoke for outer rotor machine) 1 microphone 1 m away from the outer frame, in the middle of the frame axis 1 tachometer 1 current probe for phase current Torque ripple characteristics can be indirectly recorded through angular speed derivation and tangential accelerometer on the stator Fine band FFT Averaged FFT Spectrograms Sound pressure level with and without A weighting 1/3 octave analysis is generally only useful to check standard requirements EOMYS ENGINEERING 2013- www.eomys.com 9
Run-ups / order analysis To characterize magnetic noise & vibration, recording during run-ups (variation of the speed from 0 to max speed) are useful as they allow to excite the structure at variable frequency For PMSM the open-circuit excitation magnitude is independent of speed The ideal run-up is linear increase of speed with constant magnetic excitation A constant excitation allows to distinguish structural resonances (e.g. noise level is not reduced due to defluxing, but because we go further away from resonance) In SM (or IM at no-load) the magnetic forces frequencies linked to the fundamental current are all proportional to f s The proportional factor between the vibration frequency and the mechanical frequency is known as an «order» (WARNING this has nothing to do with the spatial frequency) In order to obtain the magnitude of these vibration as a function of speed, the sampling frequency is retuned proportionally to the rotational speed = «order tracking analysis» EOMYS ENGINEERING 2013- www.eomys.com 10
Linear run-up: - all magnetic force harmonics linked to fundamental current are straight lines crossing 0 rpm, 0 Hz - Natural frequencies appear as vertical lines - Resonance appear at the crossing between «diagonal excitations and vertical modes» From [F3] From [F3] Corresponding order analysis: - all magnetic force harmonics linked to fundamental current are verticals - Natural frequencies appear as parabolas EOMYS ENGINEERING 2013- www.eomys.com 11
Due to finite response time of structural eigen modes, run-ups must be «long enough» to capture resonances The maximum sweep rate [Hz/s] can be defined using ISO79626-2 ξ i f i [Hz] S max [Hz/s] N max -N min [rpm] Sweep time [s] 0.5% 35 0.11 2500 370 0.5% 50 0.22 20 1.5 0.5% 250 5.6 500 1.5 0.5% 500 22.5 3000 2.2 The higher the natural frequency and the damping, the faster the run-up can be EOMYS ENGINEERING 2013- www.eomys.com 12
rpm rpm rpm rpm rpm Spatiograms Spatiogram = spectrogram by circumferential wavenumber = r=0 + r=-2 Hz Hz Hz + r=+2 + r=+4 + r=-4 Hz Hz Hz EOMYS ENGINEERING 2013- www.eomys.com 13
rpm rpm rpm rpm Spatiograms can also represent negative wavenumbers with negative frequencies = r=0 r=0 Hz Hz + r=-2 r=+2 + r=-4 r=+4 Hz Hz EOMYS ENGINEERING 2013- www.eomys.com 14
Spatiograms can be obtained by running classical spectrograms on spatially filtered acceleration signals EOMYS ENGINEERING 2013- www.eomys.com 15
NVH type tests for synchronous machines Standard test Run-up at no-load (+ order analysis) Run-up at nominal load (+ order analysis) Run-down in open-circuit from maximum speed Investigation test Effect of PWM switching Field weakening / effect of current angle ODS at resonances OMA EOMYS ENGINEERING 2013- www.eomys.com 16
WARNING, for PM machines a real no-load run-up does not exist (or the rotor must be rotated by an external motor in generator mode), it must be made with a coast-down Armature field introduces additional vibration harmonics due to PWM + pole/armature interaction Additional lines EOMYS ENGINEERING 2013- www.eomys.com 17
NVH type tests for induction machines Standard test Run-up at no-load (+ order analysis) Run-up at nominal load (+ order analysis) Run-down from maximum speed by shutting off current Investigation test Effect of PWM switching Effect of E/f (saturation) ODS at resonances OMA EOMYS ENGINEERING 2013- www.eomys.com 18
Ex of a data acquisition system project set-up Fine band spectra & tacho Spectrogram + averaged FFT of the phase current Spectrogram + averaged FFT of the radial & tangential acceleration Spectrogram + averaged FFT of the SPL Overall SPL (dba and db, to avoid artificial resonance due to dba increase) Acceleration and current Time analysis of SPL Coupling with converter control to have Id / Iq Extraction of LCM(Zs,2p) orders EOMYS ENGINEERING 2013- www.eomys.com 19
Experimental interpretations Interpretations of run-ups (sonagrams / spectrograms): WRSM [F3] p=2 Zs=48 (m,f)=(4,44*fr)=(-4,22*fs) and (4,52*fR)=(4,26*fs) in resonance with the circumferential mode m=2p=4 (0,48*fR)=(0,24*fs) in resonance with breathing mode m=0 EOMYS ENGINEERING 2013- www.eomys.com 20
Interpretations of run-ups (sonagrams / spectrograms): SPMSM SPMSM (p=7, Zs=42) with external rotor 200 Hz (mode 2) 400 Hz (mode 3) 1200 Hz (mode 4) 3200 Hz (mode 0) LCM(Zs,2p)/p=6 Resonance between 18*fs with mode 0 EOMYS ENGINEERING 2013- www.eomys.com 21
Interpretations of run-ups (sonagrams / spectrograms): IPMSM [F1] Traction application, acoustic noise sonagram Zs=60, p=5 LCM(Zs,2p)=60 Abrupt magnitude changes corresponds to control changes (MTPA, constant power, constant voltage) 0-th order no-load magnetic forces are proportional to 12fs PWM lines appear around 12 khz and 5 khz 6fs and 18fs are here due to an unbalance of the current flowing in the double layer windings [F2] EOMYS ENGINEERING 2013- www.eomys.com 22
Traction application, acoustic noise sonagram Zs=48, p=4 LCM(Zs,2p)=48 Gearbox mesh frequencies H10,5 H21 H29 H58 vibration noise EOMYS ENGINEERING 2013- www.eomys.com 23
Interpretations of run-ups (sonagrams / spectrograms): SCIM From [F4] Sinusoidal supply, vibration spectrogram Z s =27, Z r =21, p=2 EOMYS ENGINEERING 2013- www.eomys.com 24
Traction application in sinusoidal supply, acoustic noise sonagram Z s =36, Z r =26, p=3 1: saturation line of order Zr-Zs+4p=2 and frequency f s (Z r /p+4) resonating with ovalization mode at 700 Hz 3: slotting line of order Zr-Zs+2p=4 and frequency f s (Z r /p+2) 2: winding line of order Zr-Zs-p+5p=-2 and frequency f s (Z r /p-2) From [F4] EOMYS ENGINEERING 2013- www.eomys.com 25
From [F4] Traction application with asynchronous PWM, acoustic noise sonagram Z s =48, Z r =38, p=2 Saturation line of order Zr-Zs+4p=-2 and frequency f(zr/p+4) resonating with ovalization mode at 600 Hz Slotting+PWM interactions have same slope than slotting lines EOMYS ENGINEERING 2013- www.eomys.com 26
slotting 1 calculated angle fc=1 f s 2 calculated angles fc=5 f s 3 calculated angles fc=7 f s 5 calculated angles fc=11 f s Synchronous PWM fc=15 f s Synchronous PWM fc=21 f s Asynchronous PWM fc=600 Hz Traction application with asynchronous / synchronous / CA PWM, acoustic noise sonagram Change of PWM strategy gives sudden noise change (discontinuity of the equivalent switching frequency) Continuous acoustic lines are linked to the fundamental current (slotting, saturation, etc) EOMYS ENGINEERING 2013- www.eomys.com 27
NVH source discrimination methodology on induction machines Check background noise of motor (e.g. pumps) Check non magnetic noise sources (e.g. fans, gears, etc) Shut down the current to confirm magnetic origin of noise Vary saturation level / switching frequency to identify saturation & PWM excitations Analyse the spectrum & compare with theory to find wavenumbers & frequencies Analyse the current to identify eccentricities, switching, principal slotting harmonics Run order analysis / ODS to confirm a resonance or a forced mode EOMYS ENGINEERING 2013- www.eomys.com 28
NVH source discrimination methodology on synchronous machines Check background noise of motor (e.g. pumps) Check non magnetic noise sources (e.g. fans, gears, etc) Shut down the current to obtain acoustic noise without armature field If possible make a run-up with demagnetized magnets to obtain acoustic noise without magnet field Vary saturation level / switching frequency to identify saturation & PWM excitations Analyse the spectrum & compare with theory to find wavenumbers & frequencies Analyse the current to identify eccentricities, switching, principal slotting harmonics Run order analysis / ODS to confirm a resonance or a forced mode EOMYS ENGINEERING 2013- www.eomys.com 29