CTNet: A data-driven time-frequency technique for wind turbines fault diagnosis under time-varying speeds.

Nonstationary fault signals collected from wind turbine planetary gearboxes and bearings often exhibit close-spaced instantaneous frequencies (IFs), or even crossed IFs, bringing challenges for existing time-frequency analysis (TFA) methods. To address the issue, a data-driven TFA technique, termed CTNet is developed. The CTNet is a novel model that combines a fully convolutional auto-encoder network with the convolutional block attention module (CBAM). In the CTNet, the encoder layer is first designed to extract coarse features of the time-frequency representation (TFR) calculated by the general linear Chirplet transform (GLCT); second, the decoder layer is combined to restore and conserve details of the key time-frequency features; third, the skip connections are designed to accelerate training by linking extracted and reconstructed features; finally, the CBAM is introduced to adaptively explore channel and spatial relationships of the TFR, focusing more on close-spaced or crossed frequency features, and effectively reconstruct the TFR. The effectiveness of the CTNet is validated by numerical signals with close-spaced or crossed IFs, and real-world signals of wind turbine planetary gearbox and bearings. Comparison analysis with state-of-the-art TFA methods shows that the CTNet has high time-frequency resolution in characterizing nonstationary signals and a much better ability to detect wind turbine faults.

Bearing multi-fault diagnosis with iterative generalized demodulation guided by enhanced rotational frequency matching under time-varying speed conditions.

The rotational frequency (RF) is an important information for multi-fault features detection of rolling bearing under varying speed conditions. In the traditional methods, such as the computed order analysis (COA) and the time-frequency analysis (TFA), the RF should be measured using an encoder or extracted by a complex algorithm, which bring challenge to bearing fault diagnosis. In order to address this issue, a novel iterative generalized demodulation (IGD) based method guided by the instantaneous fault characteristic frequency (IFCF) extraction and enhanced instantaneous rotational frequency (IRF) matching is proposed in this paper. Specifically, the resonance frequency band excited by bearing fault is first obtained by the band-pass filter, and its envelope time-frequency​ representation (TFR) is calculated using the Hilbert transform and the short-time Fourier transform (STFT). Second, the IFCF is extracted using the harmonic summation-based peak search algorithm from the envelope TFR. Third, the time-varying RF ridge is transformed into a line paralleling to the time axis using the IGD with the phase function (PF). The PF is calculated by the IFCF function and fault characteristic coefficient (FCC). Lastly, the iterative generalized demodulation spectrum (IGDS) is obtained using the fast Fourier transform (FFT) for identifying fault type corresponding to the extracted IFCF. Based on obtained fault type and FCC ratios, new PFs and frequency points (FPs) are calculated for detecting other faults. Both simulated and experimental results validate that multi-fault features of rolling bearing under time-varying rotational speeds can be effectively identified without RF measurement and extraction.

Novel Method for Vibration Sensor-Based Instantaneous Defect Frequency Estimation for Rolling Bearings Under Non-Stationary Conditions.

It is proposed a novel instantaneous frequency estimation technology, multi-generalized demodulation transform, for non-stationary signals, whose true time variations of instantaneous frequencies are unknown and difficult to extract from the time-frequency representation due to essentially noisy environment. Theoretical bases of the novel instantaneous frequency estimation technology are created. The main innovations are summarized as: (a) novel instantaneous frequency estimation technology, multi-generalized demodulation transform, is proposed, (b) novel instantaneous frequency estimation results, obtained by simulation, for four types of amplitude and frequency modulated non-stationary single and multicomponent signals under strong background noise (signal to noise ratio is -5 dB), and (c) novel experimental instantaneous frequency estimation results for defect frequency of rolling bearings for multiple defect frequency harmonics, using the proposed technology in non-stationary conditions and in conditions of different levels of noise interference, including a strong noise interference. Quantitative instantaneous frequency estimation errors are employed to evaluate performance of the proposed IF estimation technology. Simulation and experimental estimation results show high effectiveness of the proposed estimation technology.