Modelling with machine learning techniques of the acoustic performance of rubberized asphalt pavements and their physical characterization to reduce their impact and extend their lifetime

Exposure to noise has significant impacts on the physical, mental health and well-being of humans and animals. Between the different anthropogenic noise sources, transport noise plays a major source due to its extension. An increment of car transportation and road freight traffic highlights the need to enhance the transportation system, as this is connected to air and noise pollution. As the number of passenger vehicles increases, this urges to address the traffic problem in different ways. Implementation of low-noise pavements and pavement monitoring are ways to reduce the noise emission due to the tyre pavement interaction and the air pollutant emissions due to the tyre consumption and the vehicle engine. In this thesis, pavement monitoring and tyre rolling noise mechanisms are studied with methods different regarding traditional ones. The two main elements presented are the use of tyre cavity microphones for pavement monitoring and the implementation of machine learning algorithms to investigate the tyre rolling noise mechanisms. For this, an initial research focused on experimentally verifying the tyre cavity noise acoustic classification in relation to the standard pavement condition index. The research aimed to develop a measurement device able to relate pavement surface condition with its emission. The results show good agreements and the need to develop an improved system. Following, to increase the knowledge about tyre cavity noise signal, various measurements at different speeds were carried out on four low-noise pavements. The objective was the further comprehension of the speed relationship of the tyre cavity noise and its relationship with the exterior noise measured with the standard Close ProXimity method. The study also focused on the resonance modes in loaded and travelling conditions. To introduce to the complexity of the tyre road noise, the next research topic focused on the acoustic and surface description of an innovative low-noise pavement that is characterized by its low thickness and high porosity. The results in time of the pavement surface and noise emission are presented, confirming the good acoustical benefits of the pavement. Finally, the use of machine learning algorithm was applied to distinguish the noise emission of rubberized asphalt pavements with crumb rubber insertion (Dry and Wet method), and pavements with no crumb rubber insertion. The method here presented gives an introductory way to compare different pavements mix by predicting the tyre road noise with a selected pavement texture surface, air temperature and tyre hardness.