A computationally efficient reduced order dynamic model for NVH and psychoacoustic predictions in electric powertrains

Electric vehicles are gaining popularity due to their cleaner transportation and lower emissions. However, the shift from internal combustion engines to electric (e-) powertrains presents new challenges in Noise, Vibration, and Harshness (NVH). E-powertrains produce distinctive tonal noises due to e-motor whistling and gear whining, making it essential to address these NVH concerns and associated psychoacoustics. The computational burden of three-dimensional, fully flexible dynamic powertrain model simulations leads to the proposal for a reduced order model (ROM). The latter is developed by integrating lumped parameter and finite-element (FE) modelling methods, considering electromagnetic effects, local nonlinearities, structural flexibility and housing mobility. The ROM provides a way to predict the radiated noise from the combined e-motor, drivetrain system and housing, offering a more economical way to predict the powertrain’s sound quality. The utilisation of the prominent ratio in psychoacoustic analysis for the radiated noise is completed by adopting established neural network techniques. Moreover, the ROM results are validated by multi-body dynamics simulations and experimental measurements. Thus, for the first time a computationally efficient ROM is presented for predicting results for psychoacoustic metrics and NVH performance in e-powertrains.