학술논문

Sensitivity Analysis Based NVH Optimization in Permanent Magnet Synchronous Machines Using Lumped Unit Force Response
Document Type
Periodical
Source
IEEE Transactions on Industry Applications IEEE Trans. on Ind. Applicat. Industry Applications, IEEE Transactions on. 58(3):3533-3544 Jun, 2022
Subject
Power, Energy and Industry Applications
Signal Processing and Analysis
Fields, Waves and Electromagnetics
Components, Circuits, Devices and Systems
Force
Vibrations
Electromagnetics
Electric machines
Sensitivity analysis
Optimization
Stator windings
Electric machine
noise
vibration
and harshness (NVH)
optimization
permanent magnet synchronous machine (PMSM)
sensitivity analysis
Language
ISSN
0093-9994
1939-9367
Abstract
Acoustic noise and vibration sources in electric machines are mainly of electromagnetic and structural origin. A lumped structural unit response-based sensitivity analysis procedure is proposed in this article, which isolates electromagnetic and structural impacts brought by variation of different design parameters in a fractional slot concentrated winding, surface mounted permanent magnet synchronous machine (SPMSM). The impact of ten design parameters on average torque, torque ripple, synchronous inductance, dominant spatial-temporal order airgap force, cogging torque, friction torque, dominant mode frequency and structural unit response is studied in detail for a 12 slot/10 pole (12s10p) SPMSM. Analysis reveals that on a 12s10p SPMSM, slot opening has a very high impact on dominant airgap force component. A multilevel nonlinear regression model-based fast optimization strategy is introduced considering electromagnetic and structural design objectives and constraints following the sensitivity analysis. The optimized design is prototyped to validate the proposed design approach. Impact hammer-based modal analysis, no load, load and run-up tests are performed on the prototype for experimental validation. Ranging from structural tests to electromagnetic tests, the experimental results closely follow and validate the simulation-based predictions.