In complex operational scenarios involving variable speeds, burdens, and noise, rotating machinery necessitates extended maintenance. Extracting stable and effective fault-sensitive features under such intricate conditions presents a significant challenge. To tackle this issue, the paper introduces the Multi-Scale Convolutional Neural Network (MSCNN) model tailored specifically for such complexities. The approach in this paper simultaneously captures multi-scale vibration signal features using the innovative Multi-Scale Bifurcation (MSB) module and subsequently aggregates them through a multi-scale fusion layer. This model effectively addresses the common problem of low CNN accuracy in fault diagnosis. The paper validates methodology using a bearing dataset provided by Case Western Reserve University (CWRU) and demonstrates superior performance compared to classical CNN models and other alternatives, achieving an impressive 99.56% classification accuracy for normal signals.

Rolling bearings serve as the most widely utilized general components in drive systems for rotating machinery, and they are susceptible to regular malfunctions. To address the performance degradation encountered by current convolutional neural network-based rolling-bearing-fault diagnosis methods due to significant noise interference and variable working conditions in industrial settings, we propose a rolling-bearing-fault diagnosis method based on dual multi-scale mechanism applicable to noisy-variable operating conditions. The suggested approach begins with the implementation of Variational Mode Decomposition (VMD) on the initial vibration signal. This is succeeded by a denoising process that utilizes the goodness-of-fit test based on the Anderson–Darling (AD) distance for enhanced accuracy. This approach targets the intrinsic mode functions (IMFs), which capture information across multiple scales, to obtain the most precise denoised signal possible. Subsequently, we introduce the Dynamic Weighted Multi-Scale Feature Convolutional Neural Network (DWMFCNN) model, which integrates two structures: multi-scale feature extraction and dynamic weighting of these features. Ultimately, the signal that has been denoised is utilized as input for the DWMFCNN model to recognize different kinds of rolling-bearing faults. Results from the experiments show that the suggested approach shows an improved denoising performance and a greater adaptability to changing working conditions.

This paper addresses the issue of noise interference and variable working conditions in the production and machining environment, which lead to weak tool life features and reduced prediction accuracy. A tool lifetime prediction method based on AVMD-DAMResNet is proposed. The method first adapts the parameters of the variational modal noise reduction algorithm using an improved sparrow optimization algorithm, and then reconstructs the original vibration signal with noise reduction. Second, the residual module of the deep residual network is enhanced using a two-dimensional attention mechanism. A parallel residual network tool prediction model (DAMResNet) was constructed to optimize the model’s weight allocation to different features, achieving multi-channel and multi-dimensional feature fusion. Finally, the noise-reduced signal was input into the DAMResNet model to accurately predict tool lifetime. The experimental results show that, compared with the original ResNet model, the proposed AVMD-DAMResNet model improves the coefficient of determination (R²) by 5.8%, reduces the root mean square error (RMSE) by 31.2%, and decreases the mean absolute percentage error (MAPE) by 31.4%. These results demonstrate that the AVMD-DAMResNet-based tool lifetime prediction method effectively reduces noise and achieves high prediction accuracy.

Freeform machining of impellers involves extended cycle times, leading to high energy consumption and costs necessitating efficient process optimization. This study develops a CAD/CAM-integrated hybrid Taguchi-Artificial Neural Network (ANN) model to optimize machining parameters for a freeform impeller. Four controllable factors, namely cutting feed ($C_f$), feed Z ($F_z$), retract feed ($R_f$), and cutter diameter ($C_D$), were investigated at five levels using an L25 orthogonal array, with machining time as the response. Taguchi analysis identified cutting feed as the most dominant factor, while retract feed was insignificant, and a first-order regression model yielded an R² of 95.88%. A two-layer feedforward neural network with six hidden neurons achieved an R² of 0.9999 and a mean absolute error of 0.0976 min. To rigorously validate generalization, leave-one-out cross-validation was employed, identifying three hidden neurons as optimal with a cross-validated R² of 0.9823, RMSE of 0.5350 min, and MAE of 0.3429 min. The final model trained on all samples achieved an R² of 0.9996. Comparison with a quadratic regression model on the same test set confirmed the superior predictive capability of the ANN ($R^2=0.9992$ vs. $0.9983$). Optimal parameters ($C_f=\mathrm{12,000} \mathrm{m}\mathrm{m}/\mathrm{m}\mathrm{i}\mathrm{n}$, $F_z=600 \mathrm{m}\mathrm{m}/\mathrm{m}\mathrm{i}\mathrm{n}$, $R_f=4000 \mathrm{m}\mathrm{m}/\mathrm{m}\mathrm{i}\mathrm{n}$, $C_D=6 \mathrm{m}\mathrm{m}$) were validated through simulation, yielding a machining time of 11.05 min, representing a 52.6% reduction from 23.32 min. The hybrid Taguchi–ANN framework effectively optimizes freeform impeller machining, significantly enhancing productivity while maintaining process reliability.

The reliability of urban rail train doors is critical for passenger safety and operational efficiency under high-frequency service. However, the coexistence of sudden faults (e.g. obstruction jamming) and progressive degradations (e.g. component wear) presents significant maintenance challenges. Existing studies often focus on a single fault type, limiting both real-time response and long-term health assessment. To address this issue, a self-consistent multi-scenario decision method is proposed for fault diagnosis and health assessment. Based on urgency and evolutionary characteristics, operation scenarios are classified into healthy, emergency, and fault-tolerant modes, forming a unified mechanism throughout the door’s life cycle. During door motion, the 3-sigma rule combined with long short-term memory network enables real-time detection of sudden faults in emergency scenarios. After each motion, multi-domain motor current features are used to construct a health indicator (HI) via a denoising autoencoder, further corrected using the local outlier factor and least squares method. Thresholds are dynamically adjusted through the Chebyshev’s inequality for multi-level health assessment in fault-tolerant scenarios. Validation on a digital twin platform demonstrates a self-consistency score of 0.9826, surpassing comparison methods. The proposed method provides an integrated approach for fault diagnosis, degradation monitoring, and preventive maintenance of urban rail train doors.

Domain generalization methods for bearing fault diagnosis aim to enhance model generalization under unseen working conditions. However, distribution discrepancies between source and target domains often degrade diagnostic accuracy. While current research predominantly focuses on model optimization, the critical role of source domain composition remains underexplored. To address this gap, this paper introduces MMD-VD—a novel dual-criterion source domain selection method based on Maximum Mean Discrepancy (MMD) and Variance Discrepancy (VD). The proposed approach innovatively repurposes these metrics from traditional roles as loss functions to a proactive selection mechanism, dynamically optimizing both the quantity and combination of source domains to significantly improve cross-condition generalization. Extensive validation on publicly available CWRU and HUST bearing datasets demonstrates the effectiveness of the proposed method. The optimal source domain combinations selected by MMD-VD achieved diagnostic accuracies remarkably close to the empirically optimal combinations, with deviations of only 0.14% on the CWRU dataset and 0.64% on the HUST dataset under cross-condition scenarios.