With the drastically increasing traffic in the last decades, crucial environmental problems have been caused, such as greenhouse gas emission and traffic noise pollution. These problems have adversely affected our life quality and health conditions. In this paper, modelling of traffic noise employing deep learning is investigated. The goal is to identify the best machine-learning model for predicting traffic noise from real-life traffic data with multivariate traffic features as input. An extensive study on recurrent neural network (RNN) is performed in this work for modelling time series traffic data, which was collected through an experimental campaign at an inner city roundabout, including both video traffic data and audio data. The preprocessing of the data, namely how to generate the appropriate input and output for deep learning model, is detailed in this paper. A selection of different architectures of RNN, such as many-to-one, many-to-many, encoder–decoder architectures, was investigated. Moreover, gated recurrent unit (GRU) and long short-term memory (LSTM) were further discussed. The results revealed that a multivariate bi-directional GRU model with many-to-many architecture achieved the best performance with both high accuracy and computation efficiency. The trained model could be promising for a future smart city concept; with the proposed model, real-time traffic noise predictions can be potentially feasible using only traffic data collected by different sensors in the city, thanks to the generated big data by smart cities. The forecast of excessive noise exposure can help the regulation and policy makers to make early decisions, in order to mitigate the noise level.

A numerical approach to predict broadband trailing-edge noise for low Mach number flows is presented. It is based on the combination of a Helmholtz solver to propagate sound waves with a sound source term computed stochastically. The sound propagation is performed in the frequency domain using a high-order finite element solver. The stochastic approach is based on a random particle-mesh method. The performance of this numerical approach is first examined for a Gaussian source. The numerical approach is then applied to a NACA0012 airfoil for flow Mach numbers of 0.11 and 0.16 and to a controlled-diffusion airfoil for a Mach number of 0.047. The predicted sound levels are compared with experimental data and acoustic results obtained from the acoustic perturbations equations. The mean flow does not significantly modify the acoustic propagation. Using a no-flow propagation model like the Helmholtz equation is therefore a valid approach for low Mach numbers. The reduced computational cost of a Helmholtz calculation, together with the speed of the random-particle mesh turbulence synthesis, allows for fast predictions. While this approach provides relative assessments between different configurations (in particular the spectrum shape), it often requires the inclusion of calibration factors determined from reference measurement or numerical data.