The design of single-degree-of-freedom spatial mechanisms tracing a given path is challenging due to the highly non-linear relationships between coupler curves and mechanism parameters. This work introduces an innovative application of deep learning to the spatial path synthesis of one-degree-of-freedom spatial revolute–spherical–cylindrical–revolute (RSCR) mechanisms, aiming to find the non-linear mapping between coupler curve and mechanism parameters and generate diverse solutions to the path synthesis problem. Several deep learning models are explored, including multi-layer perceptron (MLP), variational autoencoder (VAE) plus MLP, and a novel model using conditional
Continuous fiber-reinforced metal matrix composites (CFMMCs), reinforced by ceramic fibers (e.g., Al2O3 and SiC fibers) in a tough metal matrix, are extensively utilized in aerospace applications, such as engine casings and piston rods, because of their excellent high-temperature resistance and creep resistance. However, their heterogeneous composition presents machining challenges, including fiber pull-out, matrix adhesion, and increased tool wear. Ultrasonic vibration-assisted grinding (UVAG) effectively reduces grinding forces and abrasive wear. However, research on abrasive machining of specific CFMMCs is lacking. This study conducted single-grain cubic boron nitride grinding on SiCf/TC17 with UVAG and compared the material removal mechanisms along two different directions (longitudinal fiber [LF] and transverse fiber [TF]). A simulation model was proposed to reveal the stress distribution and its propagation. Results showed that UVAG could effectively reduce grinding forces along both directions, with an average reduction of about 17.8% compared with conventional grinding. SiC fibers were removed in three models: micro-fractures, macro-fractures, and pull-outs. The introduction of ultrasonic energy mitigated fiber damage. The simulation model was consistent with Removal Model 1. The matrix’s surface stress during grinding along LF was more concentrated than that during grinding along TF under the action of the abrasive grain. The proposed model helps understand the removal behavior of CFMMCs. This research is expected to enhance the comprehension of abrasive machining of CFMMCs and facilitate their application in the aerospace field.
Kinematic analysis and synthesis are two key topics for the study of mechanisms, and they are also important foundations of the practical application of mechanical design and control. Machine learning (ML), as a data-driven approach, enables the kinematic analysis and synthesis of different types of mechanisms while avoiding complex analytical and numerical methods. In this review, we summarize the various applications of ML algorithms and different types of data representations in the kinematic analysis and synthesis of mechanisms. A comprehensive literature review and brief analysis of current advances in ML-based approaches for the kinematic analysis of serial and parallel mechanisms, as well as their kinematic synthesis, are presented. The advantages of applying single, modular, and hybrid neural networks in the kinematics of mechanisms are discussed and compared. The future integration of ML and the kinematics of mechanisms is proposed, and the potential challenges involved are addressed.
Deployable parabolic cylindrical antennas with lightweight and high deploy/fold ratio are a research hotspot in aerospace. Most of the deployable structures of parabolic cylindrical antennas are double-layer truss structures, which are heavy and oversized in folded volume. The 2D origami-inspired structure is a typical single-layer deployable structure, including multiple origami configurations that provide various strategies for designing single-layer deployable structures. This study proposes a design method for origami-inspired single-layer truss structures applied to deployable parabolic cylindrical mesh reflector antennas. Unlike the widely researched thick-panel origami structure, we adopt the strategy of equating the creases in the origami model as links with constant length, and the vertices are regarded as hinges. The design criteria for an origami-inspired single-layer truss structure are researched and summarized by analyzing the engineering issues during design. Based on this design method, a single-layer deployable truss applied to a parabolic cylindrical antenna is presented. An optimization model of the antenna driving components is established to ensure that the antenna can deploy appropriately on the basis of the co-simulation of MATLAB and finite element software Abaqus. The optimization results are validated through software simulation and prototype test. The work presented in this paper can broaden the application of origami-inspired structures and provide a reference for the design of parabolic cylindrical antennas or curved surface mechanisms.
GH4169D superalloy exhibits exceptional service performance, enhancing the capabilities of aero-engines. However, it also poses challenges to component machining. Ultrasonic vibration-assisted machining has demonstrated advantages in enhancing material machinability. However, comprehensive analyses that pertain to the tool cutting edge path, material removal mechanism, and surface texture in longitudinal ultrasonic vibration-assisted side-milling (LUVM) are rare. In this study, GH4169D superalloy was subjected to LUVM and conventional milling (CM) to investigate the material removal mechanism and surface texture generation. Furthermore, a noncutting time ratio model was proposed to predict the reduction in maximum milling force achieved by LUVM. Results indicated that compared with that of CM, the machining of LUVM was divided into milling and noncutting. The inclusion of noncutting contributed to a reduction in the maximum milling force during LUVM. However, as milling speed increased, noncutting time ratio decreased and subsequently diminished the advantage of LUVM. The chip morphology formed using LUVM exhibited a greater degree of curliness compared with that obtained using CM, facilitating chip breaking. The utilization of LUVM resulted in the formation of a thinner lamellar structure on the free surface of chips compared with the use of CM. The machined surface exhibited a distinct ultrasonic vibration texture in LUVM, which was characterized by a physics formula. The utilization of LUVM demonstrated a reduction in machined surface roughness Ra compared with the use of CM at a low milling speed. The findings of this study contribute to the prediction of the effects of LUVM on reducing maximum milling force and achieving control over chip morphologies and machined surface texture.
In the era of intelligent revolution, pneumatic artificial muscle (PAM) actuators have gained significance in robotics, particularly for tasks demanding high safety and flexibility. Despite their inherent flexibility, PAMs encounter challenges in practical applications because of their complex material properties, including hysteresis, nonlinearity, and low response frequencies, which hinder precise modeling and motion control, limiting their widespread adoption. This study focuses on fuzzy logic dynamic surface control (DSC) for PAMs, addressing full-state constraints and unknown disturbances. We propose an improved neural DSC method, combining enhanced DSC techniques with fuzzy logic system approximation and parameter minimization for PAM systems. The introduction of a novel barrier Lyapunov function during system design effectively resolves full-state constraint issues. A key feature of this control approach is its single online estimation parameter update while maintaining stability characteristics akin to the conventional backstepping method. Importantly, it ensures constraint adherence even in the presence of disturbances. Lyapunov stability analysis confirms signal boundedness within the closed-loop system. Experimental results validate the algorithm’s effectiveness in enhancing control precision and response speed.
The continuous pursuit of extremely lightweight and multi-functional integrated designs in modern industries requires that structural materials are not limited to ensuring the structural load-bearing function of lightweight designs; rather, they must have high mechanical properties and high damping capabilities. Self-healing materials are becoming popular because of their attractive repairability and reprocessability. Dynamic reversible bonds, which are included in self-healing polymer networks, have been extensively studied with respect to different chemical mechanisms. Nevertheless, the ability to reach high stiffness and high damping performance is crucial. In this review, different types of self-healing materials are introduced, and their complex and contradictory relationships with stiffness, damping, and self-healing properties are explained. This review combines intrinsic damping sources and extrinsic deformation driving modes as a holistic concept of material–structure–performance integrated design methodology to address the extensive challenges of increasing specific damping performance. Specifically, the sources of damping at the nanolevel and the deformation-driving modes at different levels of structural hierarchy are explained in depth to reveal the cross-scale coordination between intrinsic damping sources and extrinsic deformation-driven modes originating from extremely different length scales in the microstructural architecture of a material. The material–structure–performance integrated design methodology is expected to become a key strategy for the sustainable development of breakthrough and transformative damping composite structures for aerospace, terrestrial, and marine transportation.
The change of working conditions not only makes the data distribution inconsistent, but also increases the diagnosis difficulty of fuzzy samples at the fault boundary. The traditional distance-based deep metric learning cannot effectively classify the fuzzy samples at the fault boundary. In the traditional transfer learning models, the maximum mean discrepancy (MMD) and joint maximum mean discrepancy only increase the transferability of same-class samples, and neglect the discriminability of different-class samples across different domains. The discriminative joint probability MMD (DJP-MMD) increases the transferability of same-class samples and the discriminability of different-class samples across different domains, but it only considers the global transferability of all fault classes, ignoring the different transferability of each same fault class. Therefore, a Yu norm-based deep transfer metric learning based on weighted DJP-MMD is proposed to further improve the diagnosis accuracy of bearings under variable working conditions. The deep transfer metric learning model adopts the Yu norm-based similarity instead of the distance-based similarity to effectively classify the data samples, especially those at the fault boundary, and uses the weighted DJP-MMD to measure the data distribution discrepancy between the source and target domains to increase the transferability of each same-class samples and discriminability of different-class samples across different domains. Through the fault diagnosis analysis on bearings under variable working conditions, the diagnosis results demonstrate that the proposed deep transfer metric learning model can diagnose bearing faults with higher accuracy, stronger generalization and anti-noise capabilities compared with other fault diagnosis methods based on transfer learning.
