Time-frequency (TF) analysis (TFA) is an effective tool for characterizing non-stationary signals with time-varying features. Due to the characteristics of the aero-engine with the multi-rotor system and frequent changes in working conditions, the actual collected vibration signals are not only of high complexity, strong non-stationarity, and strong noise influence but also have the phenomenon of frequency trajectories crossing. This makes it challenging to analyze such signals using the existing TFA methods. To deal with this problem, we propose an Entropy Matching TF Ridge Squeezing Chirplet Transform (EMRSCT) method. Firstly, a frequency redistribution operator is defined to solve the problem of missing time–frequency features when the widely used frequency redistribution principle deals with TF cross-signals. Next, in the frequency redistribution operator, an adaptive TF cross ridge extraction method is proposed to extract TF cross ridges under strong noise conditions. In addition, a squeezing transform method is constructed for squeezing the components in the TF crossover domain. In summary, we obtain a more accurate TF representation (TFR). The vibration signals of the test rig and real aero-engine are used to validate the effectiveness of EMRSCT, and the results show that the proposed method has good signal reconstruction ability when dealing with strongly non-stationary signals with crossover frequencies. In this occasion, it performs better than some advanced TFA methods in terms of energy concentration, noise robustness, accuracy, and computational efficiency.
 

Time-frequency analysis,Instantaneous frequency estimation,Multi-component strongly non-stationary signals,Crossover frequencies,Non-stationary signal analysis,aero-engine