A Method for Estimation of Motor Rotational Speed from STFT Spectrogram at a Non-stationary Conditions

摘要

Vibration signals acquired from rotational machines may be measured without a tacho signal for different reasons - either the speed sensor cannot be mounted due to construction reasons, speed was not measured, or the tacho signal is corrupted or not accurate enough in case of higher speeds. This paper shows STFT spectrogram based approach to determine speed profile from multiple time-varying frequency components that can be used for instance for direct processing by a Vold-Kalman filter (VKF). The method functionality is demonstrated on a simulated signal as well as on a vibration signal measured using an accelerometer on electric motor during the run-up conditions, while the rotational speed was not measured. The only known information is maximum rotational speed, presence of run-up in the signal and the motor geometry. Conducted experiment showed that despite the simple approach that resides in utilization of only one method that results directly the speed profile, the method can handle larger rotational speed changes while having enough resolution needed for VKF processing.