A Tacho-Less Instantaneous Phase and Speed Estimation for Diagnosis of Rotating Machineries Under Variable Speed

Instantaneous speed estimation from monitored vibration signals is critical for condition monitoring of variable speed rotating machineries, especially when a tachometer is not accessible. This study proposes a novel approach for estimating instantaneous phase and speed from non-stationary vibration components utilizing a modified dynamic time warping technique. The fundamental mechanism arises from the understanding that these non-stationary components can be perceived as distortions from stationary sinusoidal signals. The proposed approach flexibly aligns a vibration component to a reference with discernible phases. Generalized total least squares regression is employed to pinpoint the instantaneous phase and speed from the warping alignment path. The modification and regression mitigate the degradation of estimation caused by the pathological problem. It also finds that distances along the alignment path provide an indication to incomplete features or mode mixings, which can facilitate the determination of a band-pass filter to extract any non-stationary mono-component. The performance of proposed approach was validated through publicly accessible datasets from wind turbine and aircraft engine systems, demonstrating superior performance over existing techniques, such as Hilbert transform-based phase demodulation and ridge searching from time-frequency distributions, particularly under noisy conditions.