Engines condition monitoring based on spectrum analysis of sound signals

Abstract

This paper proposes the use of spectrum estimation with cross spectrum correlation to monitor the condition of automotive engines based on their sound signatures. This approach is convenient and economical as it is nonintrusive and does not require high precision sensors, The nonparametric spectrum estimation is chosen because the sample available is sufficiently large such that there is no leakage and frequency resolution problem. The methods chosen are periodogram, Bartlett, Welch and Blackman-Tukey.