机械专业英语

The problem of industrial bearing health monitoring and fault diagnosis has recently been a popular research topic. Extracting sufficient features from the input raw vibration signals and mapping them to the most likely fault labels is the essence of bearing fault diagnosis. This study proposes a novel framework for bearing defect diagnostics by merging dilated residual convolutional neural networks and attention mechanisms. In this framework, multiple parallel dilated convolutional networks can automatically learn rich fault features at each scale from vibration signals. Simultaneously, the attention approach boosts fault-related features and suppresses irrelevant ones, improving fault detection performance and generalization. According to the experimental results of two different bearing datasets, the framework achieves a higher accuracy and can accurately identify various types of faults.

Intelligent fault diagnosis techniques have replaced time-consuming and unreliable human analysis, increasing the efficiency of fault diagnosis. Deep learning models can improve the accuracy of intelligent fault diagnosis with the help of their multilayer nonlinear mapping ability. This paper proposes a novel method named Deep Convolutional Neural Networks with Wide First-layer Kernels (WDCNN). The proposed method uses raw vibration signals as input (data augmentation is used to generate more inputs), and uses the wide kernels in the first convolutional layer for extracting features and suppressing high frequency noise. Small convolutional kernels in the preceding layers are used for multilayer nonlinear mapping. AdaBN is implemented to improve the domain adaptation ability of the model. The proposed model addresses the problem that currently, the accuracy of CNN applied to fault diagnosis is not very high. WDCNN can not only achieve 100% classification accuracy on normal signals, but also outperform the state-of-the-art DNN model which is based on frequency features under different working load and noisy environment conditions.

The whole life cycle degradation data set of rolling bearings has the characteristics of large capacity, diversity, and non-stationarity. As a powerful tool for processing such time series data in deep learning algorithms, LSTM is prone to the loss of important time series information in the process of the life prediction of rolling bearings, which leads to a decline in prediction accuracy. Therefore, a method for predicting the remaining useful life (RUL) of rolling bearings based on the combination of temporal pattern attention mechanism (TPA) and LSTM is proposed. The method firstly combines hierarchical clustering and principal component analysis (PCA) to construct a multi-faceted and multi-scale preferred feature set reflecting the degradation information of rolling bearings, then strengthens the information correlation between hidden layers of the LSTM model through TPA and optimates the parameters of the fusion model of TPA and LSTM by using the gazelle optimization algorithm (GOA). Finally, the model is applied to the experimental data set of rolling bearing degradation. The results show that, compared with the traditional model, this method is more suitable for the remaining life prediction of rolling bearings.

Surface defect inspection is a key technique in industrial product assessments. Compared with other visual applications, industrial defect inspection suffers from a small sample problem and a lack of labeled data. Therefore, conventional deep-learning methods depending on huge supervised samples cannot be directly generalized to this task. To deal with the lack of labeled data, unsupervised subspace learning provides more clues for the task of defect inspection. However, conventional subspace learning methods focus on studying the linear subspace structure. In order to explore the nonlinear manifold structure, a novel neural subspace learning algorithm is proposed by substituting linear operators with nonlinear neural networks. The low-rank property of the latent space is approximated by limiting the dimensions of the encoded feature, and the sparse coding property is simulated by quantized autoencoding. To overcome the small sample problem, a novel data augmentation strategy called thin-plate-spline deformation is proposed. Compared with the rigid transformation methods used in previous literature, our strategy could generate more reliable training samples. Experiments on real-world datasets demonstrate that our method achieves state-of-the-art performance compared with unsupervised methods. More importantly, the proposed method is competitive and has a better generalization capability compared with supervised methods based on deep learning techniques.

With the continuous development of intelligent manufacturing technology, the application of intelligent feed systems in modern machine tools is becoming increasingly widespread. Digital twin technology achieves the monitoring and optimization of the entire life cycle of a physical system by constructing a virtual image of the system, while neural network controllers, with their powerful nonlinear fitting ability, can accurately capture and simulate the dynamic behavior of complex systems, providing strong support for the optimization control of intelligent feed systems. This article discusses the design and implementation of an intelligent feed system based on digital twins and neural network controllers. Firstly, this article establishes a mathematical model based on the traditional ball screw structure and analyzes the dynamic characteristics and operating mechanism of the system. Subsequently, the mathematical model is fitted using a neural network controller to improve control accuracy and system response speed. The experimental results demonstrate that the neural network controller shows good consistency in fitting traditional mathematical models, not only effectively capturing the nonlinear characteristics of the system but also maintaining stable control performance under complex operating conditions.

Reinforcement learning plays a crucial role in the field of robotic arm grasping, providing a promising approach for the development of intelligent and adaptive grasping strategies. Due to distribution shift and local optimum in action, traditional online reinforcement learning is difficult to use existing grasping datasets, leading to low sample efficiency. This study proposes an improved QT-Opt algorithm for robotic arm grasping based on offline reinforcement learning. This improved algorithm proposes the Particle Swarm Optimization (PSO) to identify the action with the highest value within the robotic arm’s action space. Furthermore, a regularization term is proposed during the value iteration process to facilitate the learning of a conservative Q-function, enabling precise estimation of the robotic arm’s action values. Experimental results indicate that the improved QT-Opt algorithm achieves higher average grasping success rates when trained on multiple offline grasping datasets and demonstrates improved stability throughout the training process.

Due to the high thermal conductivity and large thermal expansion coefficient of aluminum alloy materials, significant cutting forces are easily generated during metal processing, which affects the quality of the finished product. In this paper, by deeply exploring the mechanism of micro-textures on tool surfaces, we designed and prepared three types of micro-textured tools with identical morphologies but differing distributions. Cutting tests were conducted to analyze the specific impact of micro-textures on the rake and flank faces of the tool on cutting performance. To further optimize the cutting efficiency of these micro-textured tools, a coating treatment was applied to their surfaces, and an in-depth study was conducted on the changes in cutting performance under the synergistic effect of micro-textures and coating materials. The research results indicate that tools with micro-textures on both the rake and the flank faces exhibit optimal cutting performance. Furthermore, such tools with coated micro-textures can further reduce cutting forces by 21.57 % and cutting temperatures by 7.56 %, effectively facilitating the smooth discharge of chips and preventing the accumulation of cutting heat on the tool surface. This paper explores the interaction between micro-textures and coating materials, as well as their specific impacts on cutting performance, providing a theoretical foundation for subsequent research on coated micro-textured tools.

Previous research has found that the peripheral and end cutting edges of the cutter had different cutting mechanisms in the micro end cutting process considering the size effect. This investigation is a further study on this point considering the cutting edge radius of the cutter and the material of the workpiece based on the methods of finite element simulation and the micro end cutting experiment. This study adopts a combination of simulation and experiment research methods and the cutting edge radius and the workpiece material as two variables. Considering the cutting mechanisms of the peripheral cutting edge and the end cutting edge are different, the peripheral cutting edge and the end cutting edge are studied respectively. Meanwhile, the minimum undeformed chip thickness (MUCT) value is determined in three ways, chip morphology, cutting force, and surface roughness, so the final result obtained by comparing three kinds of results has a very important reference value. Not only are the chip morphology obtained by finite element simulation and the surface roughness obtained by the micro end cutting experiment used to identify the MUCT value, but also the cutting force. The simulation and experimental results show that the cutting force can be used to identify the MUCT value for the peripheral cutting edge, but it cannot be used for the end cutting edge. The MUCT value increases with the increase of the cutting edge radius, no matter which process it is. The material property has some effects on the MUCT value; even the cutting parameters and the cutting edge radius remain unchanged for the peripheral cutting edge. However, the material property has no effect on the MUCT value for the end cutting edge. In this study, the influence of important variables on MUCT is studied as much as possible to reflect a real application situation.

There are considerably rigorous requirements for accuracy and stability of the mechanism to accomplish large-scale and complex surface machining tasks in the aerospace field. In order to improve the stiffness performance of the parallel mechanism, this paper proposes a novel three degrees of freedom (DOF) redundantly actuated 2RPU-2SPR (where R, P, U and S stand for revolute, prismatic, universal and spherical joints, respectively) parallel mechanism. Firstly, the kinematics position inverse solution is derived and a dimensionless generalized Jacobian matrix is established through the driving Jacobian matrix and constraint Jacobian matrix. Secondly, the stiffness model of the parallel mechanism is deduced and the accuracy of the stiffness model is verified through finite-element analysis. Using eigenscrew decomposition to illustrate the physical interpretation of the stiffness matrix, the stiffness matrix is equivalent to six simple screw springs. Finally, the simulation experiment results demonstrate that redundantly actuated parallel mechanism has better stiffness performance compared to the traditional 2RPU-SPR parallel mechanism.

To address the unstructured on-site work requirements in shipyards and large-steel-structure manufacturing plants, this paper develops a 4SRRR (where S is spherical and R is rotational) quadruped wall-climbing robot, establishes a dynamic analytical model, and analyzes its natural frequencies. First, Timoshenko beam elements, which consider shear deformation, are used to replace Euler–Bernoulli beam elements, and the dynamic control equations for each element are established. The Lagrangian equation is then used to derive the dynamic control equations for the rods. Second, based on the theory of multipoint constraint elements and linear algebra, a set of independent displacement coordinates is established for the connection points between rods and the moving platform, rods and rods, and rods and the fixed platform. The global independent generalized displacement coordinates of the mechanism are obtained by combining these independent displacement coordinates with the internal node displacement coordinates. Third, the overall dynamic control equations of the mechanism are obtained by combining the Lagrangian equation with the global independent generalized displacement coordinates. Comparing the results with the finite element method (FEM) established using Ansys software, it is found that even when the rods are considered single elements the error in the first three natural frequencies does not exceed 3.5 %. When the rods are divided into three elements, the error in the first six natural frequencies does not exceed 5 %. Further increasing the number of rod divisions results in diminishing reductions in the error.

Lightweight construction is important in the design of railway rolling stock. Mild steels are often used for dynamically loaded damper brackets. This investigation was conducted to see if it is possible to redesign such a dynamic loaded damper bracket for additive manufacturing, especially for selective laser melting (SLM) technology, using a mild steel powder. A dynamic loaded part of a high-speed train was redesigned for SLM manufacturing technology. The design was created and topologically optimized with computer-aided design (CAD) tools. Validation of design and manufacturing (inner stress) was performed by simulation (Ansys) and by physical tests (tensile test, V-notch impact test, and fatigue test). Both the results of the simulation and of the physical tests are presented here. The design is approved by the design authority. A complete train of nine cars was equipped with 16 pieces of the redesigned part and was in validation phase. Considerable weight savings could be achieved. The mild steel powder that was used (short name E185) fulfilled the expectations according to strength and ductility. The described way of working for approval process can be understood as an example for redesign of other components. An Alstom standard for manufacturing and purchasing AM (additive manufactured) parts has been written.