In this study, a deep learning-based acoustic analysis method is proposed to predict the acoustic natural modes and natural frequencies of a structure given only its shape information. The effectiveness of the proposed method is proved by applying it to identification of the acoustic characteristics of a vehicle. The acoustic characteristics of a closed space vary depending on the shape, size, and location of the partitions existing therein. Although a designer may possess no knowledge of acoustic theory or acoustic analysis programs, the redesigning time of a mechanical structure, such as a vehicle, can be dramatically shortened if the acoustic characteristics of the candidate shape can be identified. A deep learning model is developed to perform this task on a two-dimensional acoustic cavity. It is trained with appropriate input and output data to verify the feasibility, and subsequently applied to the two-dimensional vehicle model to demonstrate its validity.
