As the most basic mechanical components, bearing troubleshooting is essential to ensure the safe and reliable operation of rotating machinery. Bearing fault diagnosis is challenging due to the scarcity of bearing fault diagnosis samples and the susceptibility of fault signals to external noise. To address these issues, a ResNet-CACNN-BiGRU-SDPA bearing fault diagnosis method based on time–frequency bi-domain and feature fusion is proposed. First, the model takes the augmented time-domain signals as inputs and reconstructs them into frequency-domain signals using FFT, which gives the signals a bi-directional time–frequency domain receptive field. Second, the long sequence time-domain signal is processed by a ResNet residual block structure, and a CACNN method is proposed to realize local feature extraction of the frequency-domain signal. Then, the extracted time–frequency domain long sequence features are fed into a two-layer BiGRU for bidirectional deep global feature mining. Finally, the long-range feature dependencies are dynamically captured by SDPA, while the global dual-domain features are spliced and passed into Softmax to obtain the model output. In order to verify the model performance, experiments were carried out on the CWRU and JNU bearing datasets, and the results showed that the method had high accuracy under both small sample size and noise perturbation conditions, which verified the model’s good fault-feature-learning capability and noise immunity performance.
This work offers a latent angular feature (LAF)-industrial adaptive smoothing (IAS) intelligent diagnostic model to tackle challenges such the lack of physical restrictions, inadequate adaptation of activation functions, and insufficient depth of multimodal fusion in fault diagnosis for process industries. Utilizing engineering heuristic constraints to eliminate excessive noise, coupled with Gram angular field computations, the model produces a LAF feature map that improves temporal correlations and weak features. An intensity-adaptive activation function (IAS) is implemented for slope adaptation, and a Residual recurrent auto-regressive neural network, which integrates multiple layers of IAS activation functions, is developed to amplify weak fault responses through differential activation on the positive and negative half-axes. The model employs time-series signals and LAF image features as multimodal inputs, integrating dynamic weight-adaptive fusion with a SE-channel attention mechanism to facilitate the profound extraction of multimodal data and the amplification of critical aspects. Experiments on the TE dataset and empirical data indicate that the model attains elevated fault diagnosis accuracy and generalization ability, while improving the interpretability and resilience of diagnostic outcomes.
To address the common challenges of feature alignment and sparse samples in unsupervised fault diagnosis tasks, this paper proposes a novel domain-robust feature-aware and sparse-sample-gain FPEAW unsupervised fault diagnosis network. Its core components include the FFE activation function, PC-EAE generation theory, and AWCA loss function, which enhance the ability to characterize fault essence under complex cross-domain conditions and improve generalization performance for complex coupled data. First, to address the sparsity of frequency-domain peak representations and the widespread distribution of background noise, we introduce an asymmetric FFE activation function based on frequency-domain entropy correction. This function amplifies fault peaks in the positive interval while mildly suppressing noise in the negative interval, effectively enhancing peak representation and noise sup-pression capabilities. Second, the PC-EAE generation theory employs multi-scale encoding and physically con-strained decoding strategies to significantly enhance data generation quality and feature fitting under sparse sample conditions. The AWCA loss function optimizes feature alignment and generated data adaptation between source and target domains through a dynamic weight adjustment mechanism. Experiments validate its superior alignment ac-curacy and domain adaptation capabilities in complex cross-domain tasks. Experimental results on the JNU, SEU, and real-world Composite fault datasets demonstrate that the FPEAW model, leveraging the strengths of its proposed modules, achieves superior diagnostic accuracy compared to existing methods in cross-domain unsupervised diag-nosis, alignment tasks, and disturbance resistance tasks. This validates the effectiveness and feasibility of the pro-posed approach in practical engineering applications.