Computational Framework for Urban Acoustic Wave Propagation and Noise Mapping Using GPU Acceleration

The acoustic environment in a city is complicated because of irregular construction geometries, changing traffic sources, and the environment. Conventional noise mapping methods, which are usually founded on simplified propagation laws, do not have the accuracy to serve well at analysing urban soundscape. The present paper presents a computation model based on a GPU architecture of urban acoustic waves propagation and noise mapping on a large scale. The scheme uses finite-differentiated time-domain (FDTD) algorithms with perfectly matched layers in order to reduce artificial reflections. Using the parallelism of graphic cards, the system can simulate the dispersion of sound in urban areas with a distance of kilometres in real-time. Comparison to in-situ measurements of noises in traffic metropolitan areas demonstrates a high level of improvement in predictive accuracy over more traditional methods based on ISO. The uniqueness of the work is the high-resolution of FDTD acoustics that is applied in conjunction with the use of a computer with a high level of acceleration, allowing a city scale prediction of noise, without excessive calculations that force inexpensive computing. The framework helps urban planners and policymakers to evaluate exposure to noise on the ground in a dynamic, mitigation control policies like a noise barrier or a green wall, and optimal urban planning policy to complement livability. The study will help in the development of the computational acoustics of urban environment, which is a sustainable objective in the application of smart cities.