A novel approach for brake emission estimation based on traffic microsimulation, vehicle system dynamics, and machine learning modeling

Abstract

Brake wear is known as the primary source of traffic-related non-exhaust particle generation. Its generation rate is influenced by parameters at different levels: subsystem (type of brakes, pads, materials, etc.), system (vehicles' dynamics, driving style etc.) and suprasystem (road geometries, traffic parameters, etc.). At the subsystem level, we proposed a neural network brake emission modeling, trained and validated through emission data collected from a reduced-scale dynamometer. At the system level, a model of a car dynamics was developed to calculate the wheels’ brake torques and angular velocities. At the suprasystem level, the traffic behavior in a sensitive urban area was characterized experimentally and simulated in a traffic microsimulation software. The vehicle traffic-based records were used to calculate the vehicle dynamic quantities, converted into brake emission through the neural network. To examine the overall traffic impacts on brake emission, the total particle number (PN) and total particle mass were estimated regarding the route choice in the sensitive area and in the whole transportation network. The findings of this study showed significant generation rate of brake emissions (in terms of mass and number) around congested areas (in the order of 10e9 #/s). The brake emission estimation in a real area provides fundamental information to the decision-makers to better insight into the rate of non-exhaust emissions generation.