Grinding chatter is the critical factor limiting high-efficiency and high-quality machining in the centerless grinding of ceramic tubes with high aspect ratio. To suppress chatter in through-feed centerless grinding, a multi-modal dynamic model of the centerless grinding system is established, incorporating multiple regenerative mechanisms, such as workpiece regeneration, grinding wheel regeneration, and geometric layout effect. The stability boundaries of the grinding system are predicted. Multi-sensor experiments integrating vibration, acoustic emission, and grinding power were conducted to investigate the dynamic stability characteristics of the complex grinding system. The results demonstrate that the proposed model effectively predicts grinding stability. Stability lobe diagrams indicate that 60–80 μm represents the critical stability threshold. A dynamic parameter identification method for the workpiece-support system based on follow-up support was proposed in this study. Increasing the feed distances of the regulating wheel reduces the amplitude of the frequency response function, achieving an average suppression rate exceeding 60 %. Considering the stiffness improvement, the optimal range for feed distances of the regulating wheel is 0.4 mm∼0.8 mm. Batch machining of silicon carbide ceramic tubes with a high aspect ratio (L/D≈50) and a wall thickness of 0.395 mm was achieved by the grinding strategy based on stability analysis. Compared with the conventional process, grinding efficiency increased by 62.5 %, while outer diameter fluctuation decreased by approximately 97 %. This study provides theoretical and technical support for the precision centerless grinding of ceramic tubes with high aspect ratio.
To investigate the motion-force coupling mechanism of cylindrical roller bearing cages under different whirl modes, a dynamic model-based on the dynamic Reynolds equation and the short bearing assumption-for simulating the cage motion trajectory was established on the basis of existing modeling work. The combined friction-damping effect at the cage-guide ring interface is further considered. The capability of the proposed model to predict the motion states and principal motion characteristics of the cage was validated by comparing the simulation results with existing literature and experimental data. The average and the maximum relative error between simulated and experimental dominant frequency components were 8.91% and 19.26%, respectively, and the predicted cage whirl direction and guide clearance phase relationships were generally consistent with the experimental results. On this basis, the motion-force coupling mechanism of cages under three typical whirl modes was further investigated, and the results indicate that: In the stable whirl mode, the oil film force Fco serves as the primary force sustaining the cage whirl, the frictional force Fct determines the whirl direction and the resultant collision force Fcc induces interference. In the random whirl mode, the irregular whirl of the cage is predominantly induced by the resultant collision force Fcc. In the local periodic whirl mode, the oil film force Fco primarily contributes to the eccentric motion of the cage, whereas the resultant collision force Fcc induces the tangential reciprocating motion of the cage.
Vibration measurements and Experimental Modal Analysis (EMA) are fundamental tools for validating complex mechanical systems. While full-field optical techniques, such as 3D Digital Image Correlation (DIC), overcome the spatial limitations of pointwise sensors, they are inherently constrained by the resolution-frame rate trade-off, restricting low-cost cameras to low-frequency dynamics. This paper proposes a comprehensive framework for blind, broadband full-field EMA using low-speed cameras, effectively bypassing the Nyquist-Shannon limit. The core novelty is a coherence-based bandwidth merging strategy, supported by an optimized Design of Experiments (DoE), which seamlessly assembles a global Frequency Response Function (FRF) from multiple band-limited measurements without prior knowledge of the structure’s resonances. To efficiently process the massive spatial datasets typical of DIC, a novel frequency-domain Repetition Method is introduced for signal reconstruction. Numerical and experimental validations on a multi-DoF model and a cantilever beam demonstrate high reconstruction accuracy with Frequency Response Assurance Criterion (FRAC) values reaching 0.99. Furthermore, the proposed frequency-domain algorithm reduces computational time by 59% compared to existing time-domain approaches. Ultimately, the framework successfully identified five structural modes up to 689 Hz using camera frame rates as low as 78 fps, offering a scalable, automated, and cost-effective alternative for high-frequency optical modal testing.
Surface-mounted arrayed piezoelectric laminated structures have potential applications in active vibration and acoustic control of fluid-loaded structures. However, the distributed actuator layout and the feedback effect of heavy-fluid loading increase the modelling complexity. This study focuses on a representative rectangular plate structure, conducting research on vibro-acoustic coupling modeling and active control. The distributed piezoelectric laminated plate is discretized into multiple control units, with coupling conditions between units and structural boundaries modeled using virtual springs. Unknown field variables are expanded using Chebyshev polynomials. The element mass matrix, stiffness matrix, and vibro-acoustic coupling matrix are derived based on Hamilton’s principle. The underdetermined global vibro-acoustic coupling equation is assembled from these element matrices. Furthermore, a discrete Chebyshev spectral boundary element method is developed, ultimately establishing a complete set of vibro-acoustic coupling equations. Building upon this foundation, an active control model incorporating both minimum mean square velocity control and minimum radiation sound power control is established, and the control mechanisms and effectiveness of different strategies are systematically analyzed. The underwater heavy-fluid vibro-acoustic formulation is verified numerically through comparison with a fully coupled finite element model and through light-/heavy-fluid response analysis. An active vibration control experiment is further conducted to validate the structural-electromechanical control submodel, including piezoelectric actuation, structural response prediction, and vibration suppression. This research provides a theoretical and numerical basis for the preliminary design and mechanism analysis of distributed piezoelectric smart structures under fluid loading.
Rotating machinery, which is a key component of industrial robots and precision equipment, is frequently susceptible to compound faults. However, diagnosing such faults remains challenging because of the coupling interference among fault features and the scarcity of labeled fault samples. To address this issue, this study proposes a semi‑supervised intelligent fault diagnosis method that integrates hierarchical attention and self‑supervised graph learning. The proposed method employs a three-layered progressive framework. At the node level, a multi-head self-attention mechanism screens and enhances the local features of each sample. At the edge level, a self-supervised graph autoencoder based on graph attention is constructed to model the correlations between samples from different sensors at the same time steps, thereby extracting and enhancing the features of various fault types. At the feature level, global similarity matching is applied to all extracted features to uncover the intrinsic relationships between compound and single faults. Finally, only a small set of labeled samples is used to train a multi‑label classifier to achieve an accurate diagnosis and effective decoupling of fault categories. Experimental results show that the proposed method achieves an average accuracy of over 99.0% in diagnosing compound faults, maintaining excellent generalization and stability, even with limited training samples. This study not only presents an effective technical solution for the compound fault diagnosis of rotating machinery but also provides substantial theoretical support for the development of predictive maintenance systems in precision machinery applications.
