The society is now in the data-rich environment, and deep learning is widely used in bearing fault diagnostic technology due to the advancement of information technology. These methods typically need a large amount of data to support. However, in some practical cases, only few of samples are frequently available when a fault occurs rather than adequate data to be analyzed. This situation indicates that bearing fault diagnostic problems are frequently few-shot problems. In this work, a generic meta-transfer learning model with special neuronal processing parameters (MSNPP) is proposed. MSNPP avoids the issue of overfitting commonly encountered in traditional meta-learning approaches when solving few-shot problems and maintains excellent performance when extracting features with deep networks. Moreover, MSNPP discovers the connection between different tasks by analyzing a few samples and quickly adapts to new tasks. In MSNPP, a technique known as neuron transfer (NT) is used to manipulate neurons by scaling and shifting them. The scaling and shifting parameters are used as meta-learning hyperparameters to transfer within different tasks, which is the work of NT. Experimental result shows that MSNPP prevents the issue of overfitting in conventional meta-learning approaches and achieves satisfactory accuracy and robustness when solving few-shot problems in fault diagnosis.

As critical components in modern aerospace productions, rolling element bearings (REBs) generally work under varying speed conditions, which brings great challenges to their operating health monitoring. Some novel time–frequency decomposition (TFD) algorithms are established recently to extract nonlinear features from the non-stationary signals effectively, which are promising for realizing fault diagnosis of REBs under varying speed conditions. However, numerous personal experiences must be incorporated and the anti-noise performance of these methods needs to be further enhanced. Given these issues, a synchronous chirp mode extraction (SCME) - based REB fault diagnosis method is proposed for the health monitoring of REBs under varying speed conditions in this study. It mainly consists of following two parts. (a) The shaft rotational frequency (SRF) is initially estimated from the low-frequency band of the vibration signal. Simultaneously, an adaptive refining strategy is incorporated to obtain a suitable bandwidth parameter. (b) A cycle-one-step estimation frame is constructed to extract synchronous modes from the envelope waveform of the vibration signal. Meanwhile, a synchronous mode spectrum (SMS) is generated using the information of the extracted synchronous modes, which is a novel REBs fault diagnosis technique with tacholess and resampling-free. In contrast to the current TFD algorithms, the proposed method needs fewer input parameters and owns a well anti-noise performance because there is no iterative optimization in the procedure of construction of SMS. As a result, the health conditions of REBs are evaluated by detecting the exhibited features in the SMS. Simulations and experiments are conducted to validate the effectiveness of the proposed method in terms of REB fault diagnosis. Analysis results demonstrate that the proposed method outperforms the current TFD algorithm and the conventional order tracking technique for fault diagnosis of REB under varying speed conditions.