Rolling bearings are crucial in mechanical systems, where failures pose safety risks and economic losses. However, their rarity makes timely fault detection and data collection challenging. To address this, a parallel dual-branch fault diagnosis model was proposed. One branch applied Variational Mode Decomposition (VMD) with the Hybrid Particle Swarm Optimization (HPSO) algorithm to automatically and optimally decompose the raw signals. Then, Continuous Wavelet Transform (CWT) was used on the decomposed modes to create stacked feature maps, which was followed by the normalization with a 1x1 convolution to accomplish the feature engineering. Next, the improved Convolutional Neural Networks (CNNs) and the Bidirectional Long Short-Term Memory (BiLSTM) models were used in sequence to extract features from both the frequency and temporal domains. In the other branch, the original signal was downsampled at three different step sizes to capture multi-scale temporal representations, and the Multi-Layer Perceptrons (MLPs) were employed to learn temporal features from different time scale, which were fused into one feature output thereafter. Eventually, the time-frequency and the temporal features from both branches were concatenated and passed to the output for fault diagnosis. The proposed feature engineering method and the parallel dual-branch fault diagnosis model were evaluated on small datasets containing 20 samples from Case Western Reserve University (CWRU), achieving an impressive 100% accuracy. Moreover, the model was compared with recent mainstream models on the Paderborn (PU) dataset. The results demonstrated that it excelled at learning key features from small samples, highlighting its potential for small-sample rolling bearing fault diagnosis.

Bearing fault diagnosis under multiple operating conditions is challenging due to the complexity of changing environments and the limited availability of training data. To address these issues, this paper presents an advanced diagnosis method using a hybrid Grey Wolf Algorithm (HGWA)-optimized convolutional neural network (CNN) and Bidirectional long short-term memory (BiLSTM) architecture. The proposed model leverages CNN for extracting spatial features and BiLSTM for capturing temporal dependencies. Through HGWA, hyperparameters are efficiently optimized, achieving 100% diagnostic accuracy across four operating conditions with the CWRU dataset. Additionally, the optimized CNN–BiLSTM model demonstrated high diagnostic accuracy when applied as a pre-trained model in new environments, even with minimal training data. The proposed model not only improves diagnostic performance but also enhances optimization efficiency, achieving faster results within the same time frame. This approach mitigates the challenges of manually tuning neural network hyperparameters and effectively addresses bearing fault diagnosis under constrained sample conditions, representing a meaningful contribution to the field of rolling bearing fault diagnostics.

The fault symptoms of rolling bearings are subject to various interferences in complex industrial environments, so achieving accurate, robust, and generalized fault diagnosis has become a key research direction. This article proposes a rolling bearing fault diagnosis method based on 1D-Inception-SE, which combines the 1D-Inception network model with Squeeze and Excitation Attention and can directly use the original vibration signals for fault diagnosis. The method incorporates the Adaptive Batch Normalization algorithm to enhance the model’s generalization performance in the presence of noise interference and cross-load diagnostics. Performance tests on Paderborn University Bearing and Case Western Reserve University datasets show that our approach achieves superior recognition accuracy compared to other models under similar and varied loads, as well as different signal to noise ratio. Ablation and visualization tests confirm the rationality and effectiveness of the model structure.

Accurate prediction of the remaining useful life (RUL) of industrial equipment is critical to operational safety and efficiency, particularly in complex multi-condition environments. However, existing methods often struggle to capture degradation patterns effectively due to the use of isolated time-series sampling strategies. To address this limitation, this study proposes a novel RUL prediction framework based on a window-fused data construction strategy, enhancing degradation modeling through systematically integrating historical operational data. A multi-layer sliding window approach is first employed to construct temporal samples, enabling alignment of operational phases and capturing cross-stage dependencies. Principal component analysis (PCA) is then used to fuse window-level features, thereby mitigating redundancy and noise while retaining key information about degradation. Finally, the PCA-fused window-level features are fed into the multi-dilated causal convolution block (MDCCB)-improved squeeze-and-excitation former model for RUL prediction. This hybrid deep learning model combines a MDCCB for extracting multi-scale temporal features with ISEformer, an improved transformer architecture incorporating adaptive multi-head weighting inspired by the squeeze-and-excitation mechanism. This design allows the model to dynamically recalibrate the importance of individual attention heads, thereby improving its ability to model complex degradation trajectories and temporal dependencies. Experimental results on benchmark datasets, including commercial modular aero-propulsion system simulation and a milling tool wear dataset, validate the superior performance of the proposed framework, demonstrating its impressive diagnosis capability across diverse industrial scenarios.