Machine learning-based prediction of drone noise propagation in three-dimensional urban environments

Area of Science:

• Acoustics and environmental engineering
• Artificial intelligence in urban planning

Background:

• Predicting noise from drones and urban air mobility (UAM) is complex due to diverse vehicle types, sound sources, and urban environments.
• Traditional methods for acoustic assessment are computationally intensive and time-consuming.

Purpose of the Study:

• To develop an efficient and accurate method for predicting sound propagation from drones in 3D urban environments.
• To leverage deep learning for improved acoustic assessment in urban studies.

Main Methods:

• A supervised U-shaped neural network, a type of convolutional neural network (CNN), was designed for sound propagation prediction.
• The network incorporates five distinct layers representing different urban altitudes to supervise the learning process.

Main Results:

• The trained neural network accurately predicts sound propagation in various urban settings.
• Achieved a mean error of less than 5% for sound pressure.
• Reduced computation time from hours to minutes compared to traditional methods like Gaussian beam tracing.

Conclusions:

• The developed neural network offers an efficient and accurate solution for predicting drone noise in urban areas.
• This approach shows significant potential for enhancing urban acoustic studies and planning.