Introducing a procedure for predicting and reducing tire/road noise using a fast-computing hybrid model.

Abstract

Prediction and reduction of tire noise are some of the main concerns of tire designers nowadays. Due to tire noise's sophisticated nature, low-noise tire design is like a maze path improbable to achieve without a scientific understanding of the underlying causes. This paper develops a knowledge-based hybrid model, incoherent summation of sounds generated and amplified by texture impact, tread impact, air-pumping, Helmholtz resonance, pipe resonance, horn effect, and air cavity resonance. The required data have been carried out by measuring C1 radial tires' noise levels in a semi-anechoic chamber. The developed model [with about a 1.7 dB(A) error on total noise prediction] presents mechanisms' contributions to the overall sound. The model is substituted with a fast-computing statistical model employing data generated based on Taguchi design. Machine learning methods are implemented for this aim, and the support vector machine provides the most accurate model. The proposed fast-computing hybrid model, developed based on a scientific description of underlying mechanisms, is applicable for noise reduction. The model's sensitivity to 21 tire parameters is analyzed, leading to valuable tips on lower tire noise. The results show the critical role of tread pattern characteristics, especially groove angle, in tire noise.