Cyclostationary Analysis-Based Gearbox Fault Diagnosis Under Varying Speed

Abstract

Under real-time conditions, fault detection of wind turbine gearboxes is a challenge, due to non-stationary vibration signals, exhibiting modulations. Thus, an attempt has been made in this manuscript by experimentally detecting the localized gear tooth faults under uncertain wind speed, not yet explored. This manuscript thus investigates the capabilities of cyclostationary analysis, resampled-time synchronous averaging (TSA), and spectral analysis at varying speed conditions. Further, kurtosis was used to analyze the processed vibration signals statistically. Due to uncertain speed variation the improvised version of TSA developed for varying speed also reflected ambiguity in extracting fault transients. The frequency spectrum and spectral kurtosis (SK) were not found effective in noticing the fault symptoms excited by the gear faults. First-order cyclostationary components indicated the presence of fault. By evaluating fourth-order cyclostationary components and kurtosis, fault was detected. Thus, the experimental results exhibit that the cyclostationary analysis completely outperforms the results of TSA, SK—kurtogram under varying speed. © 2021, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.