The testing and extraction method of high-speed wheel-rail rolling noise based on a rolling test rig

Wheel-rail and aerodynamic sources constitute the dominant contributors to noise emissions in high-speed railways. During field testing, the recorded signals are a combination of both, and no effective technique currently exists to distinguish the two. Consequently, high-speed wheel-rail simulation models lack experimental data for validation. To overcome this limitation, a laboratory rolling test rig capable of reaching 400 km/h was constructed. The acoustic response obtained from the rig is unaffected by aerodynamic influence and accurately represents the rotational behavior of the wheelset. Nevertheless, it is still interfered by auxiliary sounds generated from the drive system and cooling fans. To eliminate such interference, this study proposes an extraction approach grounded in the principle of sound-source superposition and the Neural network-based Operational Transfer Path Analysis (NOTPA) model incorporating physical mechanisms. The method effectively eliminates interference components and yields authentic wheel-rail noise. The speed-dependent results show a distinct variation in overall sound pressure levels between low- and high-speed regimes, and the underlying reasons for these differences are preliminarily analyzed. The outcomes of this work not only advance the understanding of high-speed wheel-rail noise but also supply reliable experimental data for model verification. Furthermore, the proposed methodology facilitates the separation of aerodynamic and wheel-rail sound under high-speed conditions, contributing to a clearer understanding of their respective proportions.