Yong Ren

In fact, multiple clients often use similar devices and collect fault data separately, so joint multi-client collaborative fault diagnosis modeling can solve the problem of data scarcity, but this poses great challenges to data privacy protection. In this paper, we propose a federated transfer fault diagnosis method based on federated learning for cross-domain incomplete data. The proposed method only exchanges the parameters of the local training model, which achieves the privacy protection of the client’s local data. We construct a multi-client collaborative learning framework to address the problem of weak generalization ability caused by the lack of terms in single client training samples. We also propose a targeted semi-supervised fine-tuning strategy based on relative distance to reduce the probability of negative fine-tuning of out-of-distribution samples and improve the accuracy of diagnostic models. The results of cross-condition and cross-equipment experiments demonstrate that the proposed method has obvious advantages over the existing fault diagnosis methods.

Envelope analysis is a widely used method in fault diagnoses of rolling bearings. An optimal narrowband chosen for the envelope demodulation is critical to obtain high detection accuracy. To select the narrowband, the fast kurtogram (FK), which computes the kurtosis of a set of filtered signals, is introduced to detect cyclic transients in a signal, and the zone with the maximum kurtosis is the optimal frequency band. However, the kurtosis value is affected by rotating frequencies and is sensitive to large random impulses which normally occur in industrial applications. These factors weaken the performance of the FK for extracting weak fault features. To overcome these limitations, a novel feature named Order Spectrum Correlated Kurtosis (OSCK) is proposed, replacing the kurtosis index in the FK, to construct an improved kurtogram called Fast Order Spectrum Correlated Kurtogram (FOSCK). A band-pass filter is used to extract the optimal frequency band signal corresponding to the maximum OSCK. The envelope of the filtered signal is calculated using the Hilbert transform, and a low-pass filter is employed to eliminate the trend terms of the envelope. Then, the non-stationary filtered envelope is converted in the time domain into the stationary envelope in the angular domain via Computed Order Tracking (COT) to remove the effects of the speed fluctuation. The order structure of the angular domain envelope signal can then be used to determine the type of fault by identifying its characteristic order. This method offers several merits, such as fine order spectrum resolution and robustness to both random shock and heavy noise. Additionally, it can accurately locate the bearing fault resonance band within a relatively large speed fluctuation. The effectiveness of the proposed method is verified by a number of simulations and experimental bearing fault signals. The results are compared with several existing methods; the proposed method outperforms others in accurate bearing fault feature extraction under varying speed conditions.

In practical applications, the signal measured from a complex mechanical system is usually disturbed by various noises due to the compounded effect of interferences of other machine elements and background noises, especially under varying speed conditions. Resonance-based approaches have been proven to be effective methods to address this problem. However, even if the optimal resonance band is accurately determined, the in-band noise with frequency content in the range covered by the band-pass filter is not eliminated. To avoid missed diagnosis and misdiagnosis of faults in bearings, an iterated SVD (ISVD)-based in-band noise reduction method combined with envelope order spectrum analysis is proposed in this paper. First, the optimal frequency band of a vibrational signal is determined with the help of an enhanced wavelet packet transform kurtogram, in which the kurtosis of each node is calculated based on the envelope spectrum of a signal be reconstructed using the wavelet packet coefficients. The node with the maximum kurtosis value is used to reconstruct the signal. Second, the envelope of a reconstructed signal is calculated by Hilbert transform and the ISVD method is applied to it to reduce the in-band noise. To avoid the destruction of useful information caused by excessive iteration, a threshold is set to determine the number of iterations. After iterative processing, a de-noised signal is reconstructed based on the relationship between the singular value and a frequency component. Finally, the reconstructed signal is resampled and transformed into the fault characteristic order domain where the bearing fault type can be identified from the envelope order spectra. The simulations and experiments were used to validate the efficacy of the proposed method. Compared with the spectral subtraction method, the ISVD method can suppress in-band noise efficiently and beneficial to extract the fault characteristic order under variable speed conditions.

Acceleration sensors are frequently applied to collect vibration signals for bearing fault diagnosis. To fully use these vibration signals of multi-sensors, this paper proposes a new approach to fuse multi-sensor information for bearing fault diagnosis by using ensemble empirical mode decomposition (EEMD), correlation coefficient analysis, and support vector machine (SVM). First, EEMD is applied to decompose the vibration signal into a set of intrinsic mode functions (IMFs), and a correlation coefficient ratio factor (CCRF) is defined to select sensitive IMFs to reconstruct new vibration signals for further feature fusion analysis. Second, an original feature space is constructed from the reconstructed signal. Afterwards, weights are assigned by correlation coefficients among the vibration signals of the considered multi-sensors, and the so-called fused features are extracted by the obtained weights and original feature space. Finally, a trained SVM is employed as the classifier for bearing fault diagnosis. The diagnosis results of the original vibration signals, the first IMF, the proposed reconstruction signal, and the proposed method are 73.33%, 74.17%, 95.83% and 100%, respectively. Therefore, the experiments show that the proposed method has the highest diagnostic accuracy, and it can be regarded as a new way to improve diagnosis results for bearings.