A team from Pusan National University has pioneered a new approach to creating effective sound-dampening materials using artificial intelligence. Their research focuses on tackling noise pollution in urban environments—a growing concern linked to traffic, construction, and industrial activities that can lead to health issues like stress and sleep disturbances.
Traditional methods for reducing noise often involve physically blocking sound, which can hinder ventilation. To address this challenge, the researchers introduced a novel acoustic metamaterial called the ventilated acoustic resonator (VAR). This innovative design manipulates sound waves while maintaining airflow, effectively blocking low-frequency noise with a compact structure.
A VAR functions by directing sound waves into a resonant cavity that traps them. For optimal performance, it requires complex geometries tailored for specific frequency ranges. However, conventional design methods struggle with such complexity, limiting their effectiveness.
To overcome this, the research team developed a deep-learning-based inverse design method, led by Associate Professor Sang Min Park. They utilized a conditional variational autoencoder (CVAE) to encode the geometric features of VARs into a lower-dimensional latent space. This model was trained on cross-section images of VAR cavities and corresponding frequency data.
Using this latent space, the researchers applied a genetic algorithm (GA) for optimization, mimicking natural selection to evolve designs over generations. This approach provided high design flexibility while minimizing computational costs.
Their results demonstrated that the optimized VARs exhibited broader bandwidths for sound attenuation compared to traditional designs. Dr. Park emphasized the importance of these findings, stating that the ultra-broadband VARs could significantly mitigate noise pollution in urban settings without compromising ventilation, thus enhancing quality of life.
This innovative design strategy not only promises advancements in sound control but also has the potential to transform AI-driven design processes in various fields, including automotive and aerospace engineering.