Rolling bearing is a key component of rotating machinery, and its fault diagnosis technology is very important to ensure the safety of equipment. With the rapid development of deep learning, the convolutional neural network (CNN) is widely used in bearing fault diagnosis, showing obvious advantages in diagnostic accuracy. However, the deep CNN model generally requires a lot of computing resources and storage space, and it is not easy to apply in practical engineering. Aiming at this problem, a lightweight CNN model for rolling bearing fault diagnosis is designed. This model is combined with a network pruning algorithm and neural architecture search, which not only ensures the accuracy of diagnosis but also reduces computing resources. By constructing the search space of the complete Cell class unit, using the multi-objective reinforcement learning search strategy, and applying the deep learning pruning method to prune and search the network, the lightweight CNN model with higher accuracy is efficiently searched. The rolling bearing fault data set is utilized to validate the use of the lightweight CNN model for diagnosing rolling bearing faults. This model significantly enhances operational efficiency without compromising accuracy, achieving a fault diagnosis accuracy of up to 98.56%.

The vibration signal of an early rolling bearing is nonstationary and nonlinear, and the fault signal is weak and difficult to extract. To address this problem, this paper proposes a genetic mutation particle swarm optimization variational mode decomposition (GMPSO-VMD) algorithm and applies it to rolling bearing vibration signal fault feature extraction. Firstly, the minimum envelope entropy is used as the objective function of the GMPSO to find the optimal parameter combination of the VMD algorithm. Then, the optimized VMD algorithm is used to decompose the vibration signal of the rolling bearing and several intrinsic mode functions (IMFs) are obtained. The envelope spectrum analysis of GMPSO-VMD decomposed rolling bearing fault signal IMF1 was carried out. Moreover, the feature frequency of the four fault states of the rolling bearing are extracted accurately. Finally, the GMPSO-VMD algorithm is utilized to analyze the simulation signal and rolling bearing fault vibration signal. The effectiveness of the GMPSO-VMD algorithm is verified by comparing it with the fixed parameter VMD (FP-VMD) algorithm, complete ensemble empirical mode decomposition adaptive noise (CEEMDAN) algorithm and empirical mode decomposition (EMD) algorithm.

The occurrence of the motor shaft voltage and bearing current caused by the inverter will aggravate bearing damage and lead to the premature failure of bearings. Many types of equipment are being shut down due to bearing currents, such as filters, insulated bearings and grounding brushes. Traditional suppression measures cannot eliminate the bearing current and the bearing damage mechanism under the bearing current is not clear. In this paper, the damage caused by the bearing current to bearings is analyzed in detail. The influences of different working conditions on the bearing current and the damage caused are discussed. The source of bearing currents is introduced and the bearing current model under different working conditions is reviewed. An outlook for future studies is proposed, based on the current research status and challenges.

High-speed blades are often prone to fatigue due to severe blade vibrations. In particular, synchronous vibrations can cause irreversible damages to the blade. Blade tip-timing methods (BTT) have become a promising way to monitor blade vibrations. However, synchronous vibrations are unsuitably monitored by uniform BTT sampling. Therefore, non-equally mounted probes have been used, which will result in the non-uniformity of the sampling signal. Since under-sampling is an intrinsic drawback of BTT methods, how to analyze non-uniformly under-sampled BTT signals is a big challenge. In this paper, a novel reconstruction method for non-uniformly under-sampled BTT data is presented. The method is based on the periodically non-uniform sampling theorem. Firstly, a mathematical model of a non-uniform BTT sampling process is built. It can be treated as the sum of certain uniform sample streams. For each stream, an interpolating function is required to prevent aliasing in the reconstructed signal. Secondly, simultaneous equations of all interpolating functions in each sub-band are built and corresponding solutions are ultimately derived to remove unwanted replicas of the original signal caused by the sampling, which may overlay the original signal. In the end, numerical simulations and experiments are carried out to validate the feasibility of the proposed method. The results demonstrate the accuracy of the reconstructed signal depends on the sampling frequency, the blade vibration frequency, the blade vibration bandwidth, the probe static offset and the number of samples. In practice, both types of blade vibration signals can be particularly reconstructed by non-uniform BTT data acquired from only two probes.

A gear fault diagnosis method based on kurtosis criterion variational mode decomposition (VMD) and self-organizing map (SOM) neural network is proposed. Firstly, the VMD algorithm is used to decompose the gear vibration signal, and the instantaneous frequency mean is calculated as the evaluation index, and the characteristic curve is drawn to screen out the most relevant intrinsic mode functions (IMFs) of the original vibration signal. Then, the number of VMD decompositions is determined, and the kurtosis value of IMFs are extracted to form the feature vectors. Then, the kurtosis value feature vectors of IMFs are normalized to form the kurtosis value normalized vectors. Finally, the normalized vectors of kurtosis value are input into SOM neural network to realize gear fault diagnosis. When the number of training times of SOM neural network is 100, the gear fault category is accurately classified by SOM neural network. The results show that when the training times of SOM neural network is 100 times, the gear fault diagnosis method, based on the kurtosis criterion VMD and SOM neural network is 100%, which indicates that the new method has a good effect on gear fault diagnosis.

When operating under harsh condition (e.g., time-varying speed and load, large shocks), the vibration signals of rolling element bearings are always manifested as low signal noise ratio, non-stationary statistical parameters, which cause difficulties for current diagnostic methods. As such, an IMF-based adaptive envelope order analysis (IMF-AEOA) is proposed for bearing fault detection under such conditions. This approach is established through combining the ensemble empirical mode decomposition (EEMD), envelope order tracking and fault sensitive analysis. In this scheme, EEMD provides an effective way to adaptively decompose the raw vibration signal into IMFs with different frequency bands. The envelope order tracking is further employed to transform the envelope of each IMF to angular domain to eliminate the spectral smearing induced by speed variation, which makes the bearing characteristic frequencies more clear and discernible in the envelope order spectrum. Finally, a fault sensitive matrix is established to select the optimal IMF containing the richest diagnostic information for final decision making. The effectiveness of IMF-AEOA is validated by simulated signal and experimental data from locomotive bearings. The result shows that IMF-AEOA could accurately identify both single and multiple faults of bearing even under time-varying rotating speed and large extraneous shocks.