A generalized approach for extracting deterministic components in nonstationary conditions
Fadi Karkafi, Jérôme Antoni, Quentin Leclère, Mahsa Yazdanianasr, Konstantinos Gryllias, Mohamed El Badaoui · 2025
Analyzing vibration signals from rotating machinery requires accurately extracting individual orders, which are sinusoidal components whose amplitude and phase are modulated by machine dynamics. These deterministic components play a vital role in diagnostic and prognostic tasks by providing insights into the machine’s operational state. Traditional order tracking methods often depend on predefined parameters, such as rotational speed or fixed window lengths. However, these approaches struggle when faced with rapid fluctuations or new influencing factors that are unaccounted for, leading to degraded performance. This paper introduces an automated approach that adapts to signal characteristics in a generalized framework. The proposed methodology employs a smoothing operator that dynamically adjusts to the envelope change rate, ensuring smooth Fourier coefficients and enabling adapted extraction of deterministic components in nonstationary conditions. The effectiveness of the method is validated through numerical simulations and a real-world application on the CFM56 turbojet engine. Results demonstrate the capability of the proposed approach to accurately extract cyclic orders and their nonstationary envelope modulations. A comparative analysis with the Vold-Kalman Filter, Local Synchronous Fitting, and Sliding Window Tracking highlights its superiority in attenuating unwanted sources and adapting to complex modulations without requiring manual tuning.