An Acoustic Diagnosis Scheme for Rubbing in Dual-Rotor Systems Based on Posterior Approximating

This paper aims to provide insights into the challenging inverse problem that arises when analyzing the acoustic responses of dual-rotor systems in aero-engine applications. Since the acoustic data measured from actual industrial sites is seriously disturbed, it is difficult to obtain the target features for condition diagnosis through processing the measured data. This paper provides an acoustic sensing scheme that can recover acoustic signals from distorted ones so as to support status identification in realistic applications. Specifically, in this paper, a diffusion probabilistic process is proposed for recovering signals from distorted acoustic sources with a scheduled sampling strategy. Considering the external interference and potential transformation during sound propagation, a novel strategy is adapted to approximate the posterior distribution. The effectiveness of this method is validated on a public dataset as well as signals obtained from a realistic dual-rotor test rig. Experimental results indicate that the developed method not only boosts the performance of SI-SNRi in recovered signals more than other contributions in acoustic signal processing but also retains cognizable and representative features. Identification models with recovered data achieve leading results than that with distorted data, which proves the feasibility and effectiveness of the scheme for practical applications in status identification tasks.