In practical conditions such as escalators, vibration signals collected from rolling bearings often exhibit strong non-stationarity and significant noise interference, making early compound fault feature extraction challenging for traditional methods. To address this, this paper proposes an intelligent compound fault diagnosis method integrating adaptive feature mode decomposition and improved multipoint optimal minimum entropy deconvolution adjusted (AFMD-IMOMEDA). First, a novel objective function, correlation ensemble kurtosis (CEK), is designed to comprehensively evaluate the impact intensity of modal components. Based on CEK, an improved crested porcupine optimizer (ICPO) algorithm is proposed and validated through six benchmark functions and comparison with five optimization algorithms. Subsequently, ICPO jointly optimizes the key parameters of feature mode decomposition (FMD) (i.e., the mode number M and filtering length L_1 ) to realize an adaptive decomposition strategy. After decomposition, the improved Gini index (IGI) is used to select effective modal components, which are then reconstructed to enhance fault-related transient features. To separate different fault components in the reconstructed compound fault signals, these signals are input to IMOMEDA, where a new fitness function, fault characteristic energy ratio (FCER), is introduced to assess the prominence of fault impact features. ICPO is used to jointly optimize MOMEDA parameters (i.e., the filter length L_2 and fault impact period T ), enabling parameter tuning and filtering for distinct fault types with characteristic impact periods, thus achieving precise separation of compound fault components. Finally, Hilbert envelope analysis extracts characteristic frequencies for accurate diagnosis of rolling bearing compound faults. Simulation results and experimental analyses based on the publicly available XJTU-SY bearing degradation dataset demonstrate that the proposed AFMD-IMOMEDA method significantly outperforms traditional FMD and maximum correlated kurtosis deconvolution (MCKD) in fault feature recognition accuracy and fault separation capability.

Against the backdrop of intelligent manufacturing, precise prediction of tool life is crucial for ensuring machining accuracy and efficiency. Aiming at the problem that traditional methods have limited accuracy in tool remaining useful life (RUL) prediction, this paper proposes an algorithm model integrating time-domain and frequency-domain features with a TCN-Informer cascade for tool RUL prediction. First, eight time-domain features (such as mean and standard deviation) and four frequency-domain features (such as amplitude spectrum mean and variance) are extracted from the PHM tool dataset. Features with a Spearman correlation coefficient higher than 0.8 with the maximum tool wear are screened through Spearman correlation analysis to construct an input dataset with high representational capability. Second, the causal convolution and dilated convolution of TCN are used to capture the local temporal dependencies and long-distance correlations of signals. Combined with the sparse self-attention mechanism of Informer to solve the computational complexity problem in long-sequence processing, it realizes global modeling of the dynamic features of tool wear. Experimental results show that on the C1, C4, and C6 tool datasets, the coefficient of determination (R2) of the model exceeds 0.997, and the root mean square error (RMSE) and mean absolute error (MAE) are significantly lower than comparative methods such as CNN-LSTM and Informer, demonstrating remarkable advantages in prediction accuracy and generalization ability. This method provides great help for tool health management. It can assist in formulating scientific tool replacement strategies by monitoring feature changes in real time, which is of great significance for improving machining efficiency and accuracy. In future research, multimodal data fusion and lightweight model design can be further explored to adapt to the needs of complex industrial scenarios.

The massive resources of intelligent instruments provide data support for cross enterprise regional network collaborative manufacturing. However, the discrete distribution of data, a large amount of redundancy, and weak correlation limit system knowledge learning. This paper proposes an automatic construction method of a knowledge graph for the supply and demand matching of products and the relationship between upstream and downstream enterprises. Firstly, Stanford CoreNLP tool is used to achieve enterprise entity recognition. Secondly, considering the potential relationship between boundary words and product keywords, an improved CRF model is proposed to achieve product entity recognition. Then, a method for extracting enterprise products supply relationships based on relationship indicator lexicon and dependency syntax analysis is proposed to obtain the upstream and downstream relationships of enterprise products. Finally, the Neo4j database is used to store the knowledge graph of cross enterprise product supply relationship of intelligent instruments, and it is published on a third-party service platform to promote efficient resource sharing and collaborative manufacturing of intelligent instruments across enterprises.

Despite advances in Convolutional Neural Networks (CNNs) for intelligent fault diagnosis in CNC machine tools, bearing fault diagnosis in CNC feed systems remains challenging, particularly in multi-scale feature extraction and generalization across operating conditions. This study introduces an enhanced multi-scale feature network (MSFN) that addresses these limitations through three integrated modules designed to extract critical fault features from vibration signals. First, a Soft-Scale Denoising (S2D) module forms the backbone of the MSFN, capturing multi-scale fault features from input signals. Second, a Multi-Scale Adaptive Feature Enhancement (MS-AFE) module based on long-range weighting mechanisms is developed to enhance the extraction of periodic fault features. Third, a Dynamic Sequence–Channel Attention (DSCA) module is incorporated to improve feature representation across channel and sequence dimensions. Experimental results on two datasets demonstrate that the proposed MSFN achieves high diagnostic accuracy and exhibits robust generalization across diverse operating conditions. Moreover, ablation studies validate the effectiveness and contributions of each module.

To address the issues of insufficient multi-scale feature extraction, key information attenuation, and weak model generalization in train axle box bearing fault diagnosis, we proposed a Multimodal Feature Collaborative Convolutional-Attention Network (MFCCAN). The model employs three parallel CNN branches to process raw signals, max-pooled, and average-pooled vibration data, thereby fully capturing fault information across the entire frequency domain. To enhance feature representation, the model incorporates an SE channel attention mechanism to strengthen key frequency band information and integrates a Transformer encoder to model long-range dependencies. Additionally, the synergy of SE and ECA attention modules effectively focuses on fault impulse features. Experimental results on a self-built train axle box bearing dataset demonstrate that the proposed method achieves a high accuracy of 99.37%. Further ablation studies validate the critical roles of the multi-scale structure and dual attention mechanisms, whose collaboration significantly improves model performance. The experimental results indicate that MFCCAN exhibits good adaptability in both theoretical design and engineering applications, providing a feasible solution for intelligent diagnosis and safe operation and maintenance of train bearings.

Modern subway platforms are generally equipped with platform screen door systems to enhance safety, but the gap between the platform screen doors and train doors may cause passengers or objects to become trapped, leading to accidents. Addressing the issues of excessive parameter counts and computational complexity in existing foreign object intrusion detection algorithms, as well as false positives and false negatives for small objects, this article introduces a lightweight deep learning model based on YOLOv11n, named GA-YOLOv11. First, a lightweight GhostConv convolution module is introduced into the backbone network to reduce computational resource waste in irrelevant areas, thereby lowering model complexity and computational load. Additionally, the GAM attention mechanism is incorporated into the head network to enhance the model’s ability to distinguish features, enabling precise identification of object location and category, and significantly reducing the probability of false positives and false negatives. Experimental results demonstrate that in comparison to the original YOLOv11n model, the improved model achieves 3.3%, 3.2%, 1.2%, and 3.5% improvements in precision, recall, mAP@0.5, and mAP@0.5: 0.95, respectively. In contrast to the original YOLOv11n model, the number of parameters and GFLOPs were reduced by 18% and 7.9%, respectfully, while maintaining the same model size. The improved model is more lightweight while ensuring real-time performance and accuracy, designed for detecting foreign objects in subway platform gaps.

This study proposes a planetary gearbox fault diagnosis method for variable operating conditions based on knowledge distillation and edge computing. Vibration features are extracted via Welch transform, with WDCNN as the teacher network and a small-scale 1D-CNN as the student network. Knowledge distillation improves the accuracy of the student network, which is then deployed on a Raspberry Pi for efficient, accurate, and resource-saving diagnosis.

In this paper, a novel fractal model for the contact resistance based on axisymmetric sinusoidal asperity is proposed, which focuses on the resistance characteristics of the rough interface at a microscopic scale. By introducing the unique geometric shape of axisymmetric sinusoidal asperity, and combining it with a three-dimensional fractal theory, the micro-morphology characteristics of the rough interface can be characterized more precisely. Subsequently, by conducting a theoretical analysis and numerically solving the deformation mechanisms of asperities on the rough interface, a refined model for contact resistance is constructed. This research comprehensively employs theoretical analysis, numerical simulation, and experimental testing methods to deeply explore the current transmission mechanisms during the contact process of the rough interface. The findings suggest that the proposed model is capable of precisely capturing the intricate interplay of various factors, including contact area, contact load, and material properties, with the contact resistance. Compared to the existing models, the presented model demonstrates significant advantages in terms of prediction accuracy and practicality. This research provides an important theoretical basis and design guidance for optimizing the electrical performance of the rough interface, which has great significance for engineering applications.

Tool wear prediction can ensure product quality and production efficiency during manufacturing. Although traditional methods have achieved some success, they often face accuracy and real-time performance limitations. The current study combines multi-channel 1D convolutional neural networks (1D-CNNs) with temporal convolutional networks (TCNs) to enhance the precision and efficiency of tool wear prediction. A multi-channel 1D-CNN architecture is constructed to extract features from multi-source data. Additionally, a TCN is utilized for time series analysis to establish long-term dependencies and achieve more accurate predictions. Moreover, considering the parallel computation of the designed architecture, the computational efficiency is significantly improved. The experimental results reveal the performance of the established model in forecasting tool wear and its superiority to the existing studies in all relevant evaluation indices.

In order to obtain the contact resistance of relay contacts more accurately, a novel contact resistance model for the spherical–planar joint interface is constructed based on the three-dimensional fractal theory. In this model, three-dimensional fractal theory is adopted to generate a rough surface at microscopic scale. Then, using contact mechanics theory, the deformation mechanism of asperities on rough surfaces is explored. Combined with the distribution of asperities, a contact resistance model for the planar joint interface is established. Furthermore, by introducing the surface contact coefficient, cross-scale coupling between the macro-geometric configuration and micro-surface topography is achieved, and a contact resistance model for the spherical–planar joint interface is constructed. After that, experiments are conducted to verify the accuracy of the proposed model, and the maximum relative error of the proposed model is 8.44%. Ultimately, combining numerical simulation analysis, the patterns of variation in contact resistance influenced by factors such as macroscopic configuration and microscopic topography are discussed, thereby revealing the influence mechanism of the contact resistance for the spherical–planar joint interface. The proposed model provides a solid theoretical foundation for the optimization of relay contact structures and improvements in manufacturing processes, which is of great significance for ensuring the safe and stable operation of power systems and electronic equipment.

A novel analytical model based on the generalized ubiquitiformal Sierpinski carpet is proposed which can more accurately obtain the normal contact stiffness of the grinding joint surface. Firstly, the profile and the distribution of asperities on the grinding surface are characterized. Then, based on the generalized ubiquitiformal Sierpinski carpet, the contact characterization of the grinding joint surface is realized. Secondly, a contact mechanics analysis of the asperities on the grinding surface is carried out. The analytical expressions for contact stiffness in various deformation stages are derived, culminating in the establishment of a comprehensive analytical model for the grinding joint surface. Subsequently, a comparative analysis is conducted between the outcomes of the presented model, the KE model, and experimental data. The findings reveal that, under identical contact pressure conditions, the results obtained from the presented model exhibit a closer alignment with experimental observations compared to the KE model. With an increase in contact pressure, the relative error of the presented model shows a trend of first increasing and then decreasing, while the KE model has a trend of increasing. For the relative error values of the four surfaces under different contact pressures, the maximum relative error of the presented model is 5.44%, while the KE model is 22.99%. The presented model can lay a solid theoretical foundation for the optimization design of high-precision machine tools and provide a scientific theoretical basis for the performance analysis of machine tool systems.

To improve the workpiece localization accuracy, the paper proposes an improved image alignment optimization method by random sample consensus (RANSAC) and selecting some point pairs with high confidence from them, and then calculating the affine transformation matrix separately using the combination method to filter the optimal affine transformation matrix algorithm. The image is preprocessed. First, the grayscale image is filtered with a median to remove the noise, and the scale-invariant feature transform (SIFT) detection algorithm is used to obtain the image feature points and perform feature point matching; the more decisive matching points are coarsely selected by the distance threshold between the matched feature point pairs, and the more decisive matching points are chosen from Several teams of issues are selected from the more decisive matching points. The base image transformation matrix is calculated respectively, specific methods filter the better matrix, and the better rotation angle is calculated. The accuracy and stability of the algorithm are verified through experiments, and the experimental results show that the error of the rotation angle of the workpiece is within 0.1° and the algorithm's running time is within 1 ms.

Tool wear (TW) is the gradual deterioration and loss of cutting edges due to continuous cutting operations in real production scenarios. This wear can affect the quality of the cut, increase production costs, reduce workpiece accuracy, and lead to sudden tool breakage, affecting productivity and safety. Nevertheless, since conventional tool wear monitoring (TWM) approaches often employ complex physical models and empirical rules, their application to complex and non-linear manufacturing processes is challenging. As a result, this study presents a TWM model using a convolutional neural network (CNN), an Informer encoder, and bidirectional long short-term memory (BiLSTM). First, local feature extraction is performed on the input multi-sensor signals using CNN. Then, the Informer encoder deals with long-term time dependencies and captures global time features. Finally, BiLSTM captures the time dependency in the data and outputs the predicted tool wear state through the fully connected layer. The experimental results show that the proposed TWM model achieves a prediction accuracy of 99%. It is able to meet the TWM accuracy requirements of real production needs. Moreover, this method also has good interpretability, which can help to understand the critical tool wear factors.

Machine vision has numerous uses and is currently at the forefront of research in this field. In this paper, for monocular vision systems, we mainly compare the monocular camera hand-eye calibration results and errors of two vision software platforms, and analyze the possible causes of errors to provide an idea for obtaining accurate hand-eye calibration results. And this paper derives the accuracy chain of hand-eye calibration from the whole process of hand-eye calibration, and analyzes the causes of errors in the hand-eye calibration results from each link in the accuracy chain. The results show that the accuracy of the hand-eye calibration results of Halcon software platform is higher than that of OpenCV calibration results, and the error from the robot arm can be reduced by using the relative coordinate system of the robot arm considering the direction of the hand-eye calibration accuracy chain analysis.

Herein, to accurately predict tool wear, we proposed a new deep learning network—that is, the IE-Bi-LSTM—based on an informer encoder and bi-directional long short-term memory. The IE-Bi-LSTM uses the encoder part of the informer model to capture connections globally and to extract long feature sequences with rich information from multichannel sensors. In contrast to methods using CNN and RNN, this model could achieve remote feature extraction and the parallel computation of long-sequence-dependent features. The informer encoder adopts the attention distillation layer to increase computational efficiency, thereby lowering the attention computational overhead in comparison to that of a transformer encoder. To better collect location information while maintaining serialization properties, a bi-directional long short-term memory (Bi-LSTM) network was employed. After the fully connected layer, the tool-wear prediction value was generated. After data augmentation, the PHM2010 basic dataset was used to check the effectiveness of the model. A comparison test revealed that the model could learn more full features and had a strong prediction accuracy after hyperparameter tweaking. An ablation experiment was also carried out to demonstrate the efficacy of the improved model module.

Flame recognition is an important technique in firefighting, but existing image flame-detection methods are slow, low in accuracy, and cannot accurately identify small flame areas. Current detection technology struggles to satisfy the real-time detection requirements of firefighting drones at fire scenes. To improve this situation, we developed a YOLOv5-based real-time flame-detection algorithm. This algorithm can detect flames quickly and accurately. The main improvements are: (1) The embedded coordinate attention mechanism helps the model more precisely find and detect the target of interest. (2) We advanced the detection layer for small targets to enhance the model’s associated identification ability. (3) We introduced a novel loss function, α-IoU, and improved the accuracy of the regression results. (4) We combined the model with transfer learning to improve its accuracy. The experimental results indicate that the enhanced YOLOv5′s mAP can reach 96.6%, 5.4% higher than the original. The model needed 0.0177 s to identify a single image, demonstrating its efficiency. In summary, the enhanced YOLOv5 network model’s overall efficiency is superior to that of the original algorithm and existing mainstream identification approaches.

为了提高轴承故障诊断预测的准确率，本文提出了一种基于自适应柯西变异粒子群(ACMPSO)算法对长短时记忆神经网络(LSTM)神经网络优化的轴承故障诊断预测模型(ACMPSO-LSTM)。ACMPSO算法利用非线性变化惯性权重和基于遗传算法的变异操作来提高PSO算法的全局寻优能力和收敛速度。实验结果表明，这种模型能够有效提高轴承故障诊断的准确率和稳定性。ACMPSO-LSTM模型是一种有效的轴承故障诊断预测方法，可以解决LSTM模型参数较难选取的问题，同时提高了轴承故障诊断的性能。

In this paper, we propose a novel method developed for detecting incomplete ship targets under cloud interference and low-contrast ship targets in thin fog based on superpixel segmentation, and outline its application to optical remote sensing images. The detection of ship targets often requires the target to be complete, and the overall features of the ship are used for detection and recognition. When the ship target is obscured by clouds, or the contrast between the ship target and the sea-clutter background is low, there may be incomplete targets, which reduce the effectiveness of recognition. Here, we propose a new method combining constant false alarm rate (CFAR) and superpixel segmentation with feature points (SFCFAR) to solve the above problems. Our newly developed SFCFAR utilizes superpixel segmentation to divide large scenes into many small regions which include target regions and background regions. In remote sensing images, the target occupies a small proportion of pixels in the entire image. In our method, we use superpixel segmentation to divide remote sensing images into meaningful blocks. The target regions are identified using the characteristics of clusters of ship texture features and the texture differences between the target and background regions. This step not only detects the ship target quickly, but also detects ships with low contrast and under cloud cover. In optical remote sensing, ships at sea under thin clouds are not common in practice, and the sample size generated is relatively small, so this problem is not applicable to deep learning algorithms for training, while the SFCFAR algorithm does not require data training to complete the detection task. Experiments show that the proposed SFCFAR algorithm enhances the detection of obscured ship targets under clouds and low-contrast targets in thin fog, compared with traditional target detection methods and as deep learning algorithms, further complementing existing ship detection methods.

This paper presents a multi-link magnetic wheeled pipeline robot (PR-I), which is a flexible and modular robotic mechanism. It is designed for inspection, cleaning or disinfection of central air conditioning ventilation duct. PR-I can adapt to complex pipeline terrain, and move in the ventilation ducts freely. A novel magnetic wheel is proposed for wall climbing, which is circumferentially embedded with rectangular permanent magnets. Firstly, the wall climbing model is established to obtain the adsorption conditions, then the effects of the magnetizing direction of the magnets, magnetic wall thickness and gap on the magnetic force are analyzed by finite element method. Finally, the optimal magnetic wheel structure is obtained through dynamic simulation analysis. The motion control of the robot is carried out in an embedded system. Two fuzzy controllers based on ranging sensor and IMU are proposed for straight motion and turning motion, and gait sequences are designed for obstacle surmounting. The performance of the robot was evaluated in a real ventilation duct. The experimental results show that PR-I has good trafficability and control performance in various terrains.

为增强学生运用理论知识解决实际工程问题的能力,促进实验中多学科知识的交叉融合,设计了基于视觉识别的智能垃圾分类设备,开发了一套全新的垃圾分类视觉与图像处理检测方法,并将该设备应用于机械类专业本科生的图像处理技术与应用课程实验教学.介绍了该设备的整体方案设计,以及该设备利用机器视觉模块任务的不同环节开发的6个子实验项目.这些实验项目能够让学生充分做到理论与实践相结合,锻炼学生的自主开发与科研能力,为以后的毕业工作和深造打下基础.