An algorithm of searching instantaneous frequency based on sparse transform of TFD and fuzzy decision

Abstract

Searching for instantaneous frequency (IF) ridges accurately from the time–frequency distribution (TFD) of vibration signals is crucial for analyzing faults in variable-speed rotating machinery. However, the IF is easily interfered by noise and neighboring frequencies in practical TFD, which causes difficulty in searching. To solve this problem, an improved IF ridge search algorithm based on sparse transform of TFD and fuzzy decision (STTFD-FDSA) is proposed in this paper. First, STTFD is proposed to construct a sparse representation of TFD (STFD) to improve the IF discriminability and reduce the amount of data. Then, the frequencies and directions of the ridge points to be searched are predicted based on the segmented curve fitting (SCF) method. Finally, the fuzzy decision model is constructed based on 5 features, including frequencies, directions, magnitudes, and predicted frequencies and directions of the points in the STFD to determine the optimal ridge points. Simulations and experiments show that the proposed method can search the weak IF components accurately under noise interference, and the effect and robustness are better than the 3 existing methods.