In this study, an assessment method is proposed to quantify rattle noise by applying variational mode decomposition (VMD) to automotive structural sound radiations. The complete prevention of rattle sounds in assembled plate structures excited by vehicle operation conditions is difficult. It is important to design automotive structures that induce minimal acoustic contributions to the vehicle interior noise. To investigate occurrence of unpredictable short-duration sounds, precise decomposition without distortion is required. The VMD has advantages on disintegration of complex signals. To confirm the decomposition performance for sounds, synthetic acoustic signals composed of pink and impact sounds were analyzed. The separation performance was compared to conventional spectral filtering methods and complete ensemble empirical mode decomposition (CEEMD). The sound radiation from an automotive structure was measured using the vibration test in an anechoic chamber. The measured sound was decomposed into background and rattle sounds using VMD. The broadband background sound was quantified by level. The rattle sound was evaluated using Prony analysis. To investigate the masking effect, just noticeable differences (JND) were obtained with variations in the levels of the decomposed sounds. The algorithm to derive annoyance index from evaluation of the decomposed sounds was proposed. Auditory experiments were conducted to obtain perceived annoyance of the noise emitted by an automotive structure. The annoyance index predicted the actual perception precisely compared with the objective sound quality metrics. The loudness alone did not effectively quantify the impact sound. The results were used to confirm the reliability of the proposed method, demonstrating the advantages of VMD on impact sound annoyance quantification analysis.
