Fused silica is extensively deployed in high-end optical components, semiconductor lithography systems, and aerospace payloads owing to its exceptional deep-ultraviolet transmittance, ultra-low thermal expansion, superior chemical stability, and high laser-induced damage threshold. Nano-grinding and polishing serves as the core enabling technology for realizing high-precision, low-damage machining of fused silica components. Nevertheless, the multi-scale evolution mechanism of surface damage and the cross-scale interaction laws governing material removal and defect formation during nano-grinding and polishing remain poorly elucidated. To fill this critical knowledge gap, a multi-scale methodology combining variable-load nano-scratch experiments, finite element simulation based on the JH-2 constitutive model, and molecular dynamics simulation with the Tersoff potential was employed to systematically investigate the material removal modes, dynamic stress field evolution, and intrinsic formation mechanism of surface densification damage in fused silica. The results reveal that with increasing abrasive scratching load and depth, the material removal behavior of fused silica transits sequentially through three stages: ductile removal, ductile-brittle mixed removal, and brittle removal, accompanied by continuous elevation of instantaneous surface stress and grinding heat, which markedly aggravates surface damage. Increasing the scratching speed effectively reduces the stress amplitude and internal energy accumulation of the machining system, thereby alleviating surface machining damage. Molecular dynamics simulations validate that octahedral shear stress drives irreversible atomic plastic flow, which acts as the core inducement of permanent surface densification damage, whereas hydrostatic stress only generates instantaneous volume strain without residual retention after unloading. This work clarifies the multi-scale evolution law of surface damage in fused silica during nano-grinding and polishing, and unravels the atomic-scale formation mechanism of surface densification. The findings provide fundamental theoretical support for the optimization of low-damage process parameters and precise control of machining quality in the high-performance atomic-scale manufacturing of fused silica optical components.
Diamond compound refractive lenses (CRLs) are key focusing components for the fourth-generation High Energy Photon Source (HEPS) in Huairou, Beijing. However, existing lens centering methods suffer from an inherent drawback because they rely on “relative alignment” rather than “absolute alignment.” This strategy depends heavily on the perpendicular relationship between the target and the V‑groove fixture, and the centering accuracy becomes increasingly unstable as the number of lenses grows. To meet the stringent requirements of HEPS for diamond CRLs, this study introduces machine vision into the fabrication of diamond CRLs for the first time, aiming to achieve “absolute alignment” of the lenses and fundamentally circumvent the intrinsic deficiency of the “relative alignment” approach. To cope with the overexposure and severe Poisson noise caused by the harsh imaging conditions in the femtosecond laser system, a three‑stage centering algorithm is constructed. This algorithm comprises a novel illumination error compensation model, a lightweight deep learning semantic segmentation network, and the RRCanny–Zernike edge detection operator, all proposed in this work. Laser processing experiments were performed on ten diamond lens blanks using the proposed centering method. The results show that the mean absolute error (MAE) of the lens center deviation is 5.21 μm, the root mean square error (RMSE) is 5.24 μm, and the mean square error (MSE) is 27.43 μm². The centering accuracy surpasses that of the current world‑leading lenses fabricated by Antipov’s team and satisfies the precision requirements of the Beijing Huairou HEPS for diamond CRLs. A gray-level-based exposure error compensation model is proposed: This model resolves the degradation of edge detection caused by overexposure in femtosecond laser machining environments. Experiments verify that the Radial Basis Function neural network optimized by the Whale Optimization Algorithm (WOA-RBF) achieves the best compensation performance. An enhanced DeepLabV3+ segmentation model is developed: MobileNetV3 replaces the conventional ResNet backbone, and a newly invented Efficient-ASPP module is introduced to replace the standard ASPP, resulting in significant improvements in segmentation efficiency and accuracy. A RRZ-Canny subpixel edge detection operator is proposed:By integrating ROI extraction, RANSAC, Zernike polynomials, and an improved adaptive Canny algorithm, the method substantially enhances edge extraction accuracy and robustness, enabling high-precision center localization.
Through quasi-static experimental testing of a hybrid energy absorber combining shrinking-by-extrusion and groove-splitting, the energy absorption mechanism was qualitatively analyzed. Based on this experimental work, an implicit elastoplastic finite element method (FEM) was employed to precisely analyze and understand the groove splitting phenomenon observed after extrusion-induced shrinking during the test. By leveraging the symmetry of the test process, the minimal analysis region was modeled as a finite element analysis, thereby avoiding complex self-contact issues and maximizing both the reliability of the results and computational efficiency. The flow curve was obtained from the tensile test, and a rational method for determining the friction coefficient and critical damage in the normalized Cockcroft-Latham damage model was proposed, considering that the test process comprises friction-dominated and fracture-dominated parts that are almost independent. Using a high-precision finite element analysis model under optimized conditions, the energy absorption mechanism of the hybrid energy absorber during testing was elucidated quantitatively into six distinct phases.
This study proposes a machine vision-based tracking vibrometry system for moving targets. A closed-loop architecture integrating visual tracking, optical path control, and vibration detection is established to achieve synchronous tracking and vibration signature extraction of dynamic targets. Firstly, a coaxial vision unit comprising a galvanometer scanner, camera, and laser is constructed. An imaging model based on Radial Basis Function Neural Network is developed to map the relationship between pixel coordinates and spatial positions under varying galvanometer deflection angles, enabling high-precision pixel-to-world coordinate transformation. Secondly, addressing galvanometer scanning errors, a bivariate polynomial curve fitting method constructs an error compensation model that rectifies galvanometer commands in real time, ensuring accurate laser targeting. Furthermore, after target localization via integrated image processing and the imaging model, kinematic modeling of the error-compensated scanning system generates galvanometer control commands. These commands drive laser-target synchronization, forming a real-time closed loop: image acquisition-target localization-optical path adjustment. Finally, an experimental platform was built to conduct tracking vibration measurement experiments. Two vibration measurement schemes, fixed-point and dynamic, were designed, and constant-frequency excitation was applied to collect verification signals. The experimental results show that the acquired signals are highly consistent with the constant-frequency excitation, verifying the system’s synchronous tracking capability and vibration measurement accuracy. This study addresses the key technical challenges of vibration measurement for moving targets through imaging modeling, error compensation, and multi-module coordination. The feasibility of the proposed system is experimentally validated, providing a novel methodological reference for the design of related systems and offering effective technical support for industrial applications such as condition monitoring of moving machinery.
This study utilized ANSYS software to simulate the riveting and tensile testing processes of countersunk rivets and performed experimental validation. It analyzed the relationship between the stress distribution around the rivet hole after hammer riveting and the riveting parameters, as well as their influence on the maximum tensile load. The results indicate that when the rivet diameter increases from 4.83 mm to 6.35 mm and the rivet length increases from 5.08 mm to 8.13 mm, the stress around the rivet gap first decreases and then increases. An increase in the friction coefficient results in an increase in equivalent stress (EQV) but a reduction in the absolute values of directional stresses. The optimal parameter combination for tensile performance is a rivet diameter of 5.08 mm, a rivet length of 8.13 mm, and a friction coefficient of 0.1. The parameter-stress fitting results indicate that rivet diameter dominates the variation in EQV, rivet length regulates the stress distribution, and friction coefficient affects the stress in the vertical direction. The stress variations are caused by the load direction and frictional effects. The relationship between stress and tensile load indicates that the maximum tensile load is nonlinearly related to the equivalent stress. The established model demonstrates high accuracy in predicting tensile performance, with an error of only 5.08
