Improving vehicle aeroacoustics using machine learning

This paper presents a new approach to improving the overall aeroacoustic comfort of a vehicle, an important feature of vehicle design. The traditional improvement process is extended to benefit extensively from machine learning, information retrieval and information extraction technologies to assist the wind tunnel engineers with difficult tasks. The paper first describes the general approach and then focuses on providing a detailed description of the most important task of assessing the degree of discomfort for a human caused by wind noise in a vehicle, when the noise spectrum is known. For this purpose a novel approach of learning linear regression models that are consistent with expert's domain knowledge is presented. The results of the end user evaluation of the entire system are also presented to reflect the strengths of this approach.