Psychoacoustics-guided noise and vibration source identification for battery electric vehicles

Reduced engine-noise masking in battery electric vehicles (BEVs) makes narrowband tonal and modulated components more perceptually salient in the cabin. Conventional energy-based noise source identification (NSI) may overlook components critical to perceived sound quality. This study proposes a psychoacoustics-guided NSI method using synchronous measurements from 13 microphones and ten accelerometers on a test vehicle under 24 operating conditions across four road types. Subjective annoyance evaluation of ear-position noise recordings by 32 participants identified loudness and tonality as perception-relevant psychoacoustic metrics. The dominant bark bands of these metrics and a coherence criterion screen of narrowband candidates extracted by ensemble empirical mode decomposition were combined with independent component analysis, yielding perceptually salient frequencies. Noise contributors are identified through partial coherence, validated using multiple-input single-output linear optimal estimation, and traced to upstream vibration contributors through partial coherence. Dominant bands concentrated within 0.5-4 bark (approximately 20-402 Hz). Reconstruction agrees with measurements over 20-450 Hz, with a root mean square error of no more than 1.882 dB(A) and psychoacoustic relative errors below 20% for both loudness and tonality. These results indicate that perception-based NSI better captures cabin noise salience relevant to subjective annoyance and provides a basis for targeted noise, vibration, and harshness optimization in BEV cabins.