Rolling-element bearings in rotating machinery operate in a scenario where multiple vibration sources are coupled so that the fault-related components of the monitored bearing are easily masked by the vibration of neighboring components. To address this problem, this paper proposes a fault diagnosis framework that integrates a variational mode decomposition optimized by the Bayesian information criterion (BIC) (OVMD) with an Elman network optimized by an improved genetic algorithm (IGA-Elman). First, the BIC adaptively estimates the optimal mode number of VMD from the measured vibration signal. Second, six time-domain statistics with clear physical interpretations, namely, energy, standard deviation, kurtosis, crest factor, impulse factor, and sample entropy, are extracted uniformly from each decomposed mode, and the resulting 42-dimensional feature vector is fed into the IGA-Elman classifier to identify the fault pattern. On simulated multicomponent signals, OVMD reduces the maximum amplitude deviation rate to 1.167%, against a minimum of 1.911% for conventional VMD, and its energy leakage is smaller than that of EMD, conventional VMD, and other mainstream decomposition methods. On the CWRU bearing dataset, 30 repeated comparative experiments were carried out under a recordwise 5:3 split: With the optimal mode number k∗ = 7, the overall accuracy reaches 99.47% ± 0.50%, and both the mean accuracy and the stability are better than those of the other six compared models. Under strong noise, the accuracy remains 98.3% at a signal-to-noise ratio of 10 dB, which is markedly better than that of conventional VMD features and raw-signal features, and cross-sensor few-shot fine-tuning raises the accuracy to 85.59% ± 3.40%. The framework thus provides a highly interpretable and engineering-oriented solution for intelligent bearing fault diagnosis in strongly disturbed industrial scenarios.
Blast waves generated by moving charges exhibit pronounced directional asymmetry, which cannot be adequately represented by stationary‐charge blast formulations. In this study, a two‐dimensional axisymmetric multimaterial Eulerian model was established to investigate the free‐field detonation of a bare spherical TNT charge moving at Mach numbers from 0 to 4. Pressure‐time histories, arrival times, peak overpressures, and wavefront evolution were extracted from the numerical results. The simulations show that charge motion causes forward enhancement, rearward attenuation, and an overall migration of the reconstructed wavefront. Based on these characteristics, an engineering model was developed by combining a stationary‐detonation overpressure baseline, a geometric‐center migration state variable, an effective scaled distance, Mach‐number‐dependent directional corrections, and a logistic transition function across the transonic regime. A velocity‐based holdout strategy was adopted for model identification and testing. For the 140 testing samples, the complete model achieved a mean relative error of 3.14%, an MAE of 75.52 kPa, an RMSE of 186.65 kPa, and a maximum relative error of 16.52%. Ablation analysis confirmed the contributions of geometric‐center migration, directional corrections, and the logistic transition to prediction accuracy. For 24 publicly available BRL moving‐charge samples, three‐fold cross‐validated recalibration of the high‐Mach‐number directional correction yielded an MAE of 27.41 kPa, an RMSE of 34.48 kPa, a mean relative error of 5.89%, and a maximum relative error of 20.92%. Within the investigated charge‐velocity, scaled‐distance, and geometric conditions, the proposed formulation provides a physically interpretable engineering method for predicting the peak overpressure generated by free‐field moving detonation of bare spherical TNT charges.
High length-to-diameter ratio boring bars are prone to regenerative chatter during deep hole boring, which seriously restricts machining accuracy and stability. Passive damping boring bars, due to their reliable structure and strong engineering adaptability, have become the core technical direction for suppressing boring vibration. This paper systematically reviews the research progress of passive damping boring bars. Based on the vibration damping mechanism, the existing methods are classified into three major categories: energy dissipation type, dynamic vibration absorption type, and stiffness enhancement type. The structural features, applicable working conditions, and vibration damping performance of these methods are compared and analyzed. The research achievements and engineering performance of emerging technologies such as particle damping, constrained layer damping, nonlinear and adaptive dynamic vibration absorbers, as well as composite materials and lattice structures in wideband chatter suppression are mainly reviewed. Based on this, the common problems of the current passive damping boring bar in terms of frequency domain adaptability, parameter sensitivity, and multimechanism coupling modeling are summarized. The development direction of wideband adaptive, lightweight, and digital design is proposed, providing a systematic theoretical reference for the structural design and performance optimization of the passive damping boring bar.
This paper investigates the dynamic response of a fractional order viscoelastic foundation and a medium-thick shell tunnel lining system under harmonic loads. To describe the frequency-dependent mechanical behavior and memory characteristics of the foundation, the foundation is simulated as a fractional-order viscoelastic medium, considering the influence of shear deformation, and the tunnel lining is modeled using the medium-thick shell theory. Based on Helmholtz decomposition, the foundation wave equation is decoupled into compression wave and shear wave components. Subsequently, the general solution form of the wavefield displacement potential function was derived and expressed as a Bessel function to satisfy the radiation conditions at infinity in the entire space. The dynamic response of the tunnel lining and surrounding foundation under harmonic loads was obtained by utilizing the continuous conditions at the interface between the foundation and tunnel lining and inverse Fourier transform. The influence of coupled system parameters on displacement response was discussed through numerical examples, and the spatial distribution law of displacement was obtained. Research has shown that fractional-order viscoelastic models are more accurate in describing the energy dissipation characteristics of vibration than integer-order viscoelastic models.
Compared with constant operating conditions, the vibration signals of rolling bearings under variable load conditions are more susceptible to the influence of load fluctuations, resulting in complex and weakened fault characteristics, making it difficult for traditional models to extract effective features and possess sufficient generalization ability, thereby affecting the accuracy of diagnosis. To address these issues, this paper proposes a wide kernel convolutional neural network (WLGCNN) for fault diagnosis under different load conditions. This model uses a first-layer wide convolution kernel to capture the fault influence time-series patterns within the rotation period, and the subsequent part uses 1 × 1 convolution to complete channel feature fusion and nonlinear transformation. Combined with data preprocessing strategies such as signal segmentation and Z-score normalization, the sample data are effectively purified and the feature recognition ability is improved. Experiments are conducted using the standard dataset of rolling bearing faults provided by Case Western Reserve University (CWRU) and the Mechanical Fault Prevention Technology (MFPT) dataset. Ablation experiments and cross-load comparison experiments verify that the model constructed in this paper has high diagnostic accuracy and excellent generalization performance.
