Embodied Intelligence (EI) is propelling industrial manufacturing systems toward the Industry 5.0 paradigm, yet a core challenge remains in systematically integrating physical entities with advanced cognitive capabilities. This article aims to address this challenge by proposing a novel ecosystem framework to guide the EI empowerment of smart manufacturing characterized by autonomy, adaptability, and collaboration. Our primary contribution is the presentation of the Embodied Smart Manufacturing Empowerment (ESME) ecosystem that dissolves the boundaries among heterogeneous production elements inherent in conventional manufacturing, thereby enabling dynamic empowerment across the entire lifecycle of the manufacturing system. The ESME ecosystem uniquely structures the manufacturing lifecycle into three interconnected stages: (1) Design Stage EI, focusing on comprehensive virtual optimization; (2) Manufacturing Stage EI, for adaptive and high-precision physical execution; and (3) Running Stage EI, dedicated to continuous autonomous evolution in operation. Through a comprehensive review and analysis of empowerment processes associated with diverse EI-related enablers and an examination of their mutual coupling relationships across these stages, a closed-loop “perception-analysis-decision-execution-evolution” cycle can be established. This work provides a clear roadmap and theoretical foundation for realizing highly intelligent, ubiquitous, and autonomous manufacturing systems, offering significant implications for future academic research and industrial practice.

Diesel engines in open-pit mines frequently experience compound faults due to long-term operation under harsh environments and high-power conditions, posing severe challenges to autonomous perception, dynamic decision-making, and intelligent operation and maintenance. To address the issues of severe feature redundancy, low model interpretability, and the incomplete embodied intelligence system construction in existing methods, a dynamic feature selection method that fuses hierarchical feature pre-screening with a Deep Q-Network (DQN) from the perspective of embodied intelligence is proposed. In the proposed method, a multi-domain fusion feature library is established, and a dual-screening mechanism based on Analysis of Variance (ANOVA) and correlation analysis is introduced to effectively eliminate redundant interference. The DQN is employed to reformulate the feature selection as a sequential decision-making process, adaptively searching for the optimal feature subset under a predefined feature budget constraint. Building on this, a closed-loop embodied intelligence system integrating dynamic diagnosis, risk assessment, and digital twin validation is established, creating a complete “perception-cognition-decision-execution” operation and maintenance workflow. Experimental results demonstrate that the proposed method achieves a diagnostic accuracy of 89.17% on the test dataset. Ablation experiments reveal that the proposed method exhibits superior performance in terms of accuracy, computational cost, and inference time compared to baseline schemes, including “using only the DQN without hierarchical pre-screening” and “employing only pre-screening without DQN-based dynamic decision-making”. The experimental results validate the effectiveness of the proposed method for the intelligent diagnosis of compound faults in diesel engines.

Fault diagnosis based on bearing vibration signals is an important approach to improving the operational safety of complex equipment. However, existing zero-shot fault diagnosis methods across operating conditions typically require numerous learning samples and fail to model and analyze the physical mechanisms of bearing failure and the noisy sensor signal characteristics, limiting the domain adaptability of the diagnostic model. Therefore, this article proposes a zero-shot bearing fault diagnosis method based on physics-informed machine learning to improve the model's cross-domain diagnostic performance under noisy sensor data. By integrating correlation analysis operations into neural networks, an automatic physical information learning module for bearing fault feature decoupling is established to achieve fault attribute separation guided by physical information. Then, leveraging the physical characteristics of periodic signal pulses in faulty bearings, a deep encoding-parsing network is designed to automatically generate fault attributes and analyze failure features for rolling bearing fault diagnosis. Comparisons with state-of-the-art methods demonstrate that the proposed method outperforms other methods in unknown domains, even under noisy sensor data. It also demonstrates great potential for cross-equipment implementation, potentially addressing the problem of data-free, cross-equipment adaptive fault diagnosis.

Currently, fault diagnosis for rotating machinery addresses the data scarcity problem across scenarios by incorporating Federated Learning (FL) technology. However, in flexible customized production, frequent process scheduling changes and the collaborative operation of various rotating machinery can lead to the challenge of domain shifts influenced by dual spatiotemporal features across processes and machinery, resulting in the negative aggregation in the FL central model. This challenge encompasses spatial feature domain shifts caused by individual and type variations in rotating machinery, as well as temporal feature domain shifts induced by production scheduling. Consequently, we propose a contrastive prototype guided federated learning approach for rotating machinery fault diagnosis (FedCPG). A global prototype generator (GPG) based on the Kolmogorov-Arnold Network (KAN) is trained on the server, leveraging its nonlinear high-dimensional mapping ability to generate personalized inverse mappings and its computation is optimized. The aggregation layer adopts a Momentum Contrastive Loss Fusion Mechanism (MCLFM) to enhance adaptive feature separability across fault domains. In the edge learning layer, a Hierarchical Adaptive Contrastive Learning Strategy (HACLS) is constructed to extract domain-invariant features, achieving tight temporal alignment of prototypes within the same domain while preserving spatial semantic information. The proposed approach is evaluated using a cross-process and cross-machine fault dataset, with multiple sets of excellent diagnostic results fully illustrating the generalization and superiority of the FedCPG.

With the continued advancement of Artificial Intelligence (AI) technology in Customized Manufacturing (CM), the current intelligence model, which separates ‘perception’ from ‘execution,’ lacks adaptability and generalizability. Embodied Intelligence (EI), an emerging technology emphasizing real-time environmental interaction and feedback, is expected to enable integrated intelligent manufacturing systems characterized by ‘perception-cognition-execution-feedback,’ enhancing the performance and intelligence of multifactor systems like the Internet of Manufacturing Things (IoMT), equipment management, and resource scheduling. However, current AI systems in CM remain isolated and lack methods for environmental interaction with multi-source perception and feedback, hindering the development of autonomous, evolving intelligent systems. To address the environmental interaction bottlenecks among diverse production elements in CM, this paper proposes a Circular Embodied Intelligence Manufacturing (CEIM) architecture aimed at enabling the fusion of heterogeneous production-element information. The objective is to enhance environmental perception and establish autonomous system evolution mechanisms, thereby optimizing production decision-making. By integrating sensor data and AI model outputs, and leveraging the advanced reasoning capabilities of large language models (LLMs), CEIM facilitates the semantic fusion of multiple production elements—including IoMT, intelligent devices, and manufacturing resources—to enable the deployment of EI applications in CM. The implementation of CEIM is illustrated and validated through a case study on a customized gift box packaging platform. Finally, this paper discusses the opportunities and challenges of applying EI in manufacturing and aims to provide insights into the future development of EI-oriented manufacturing systems.

Aiming at the problems of low precision and slow detection of daylily maturity, an intelligent detection method based on YOLOv8 daylily maturity is proposed. This intelligent detection method incorporates the SimAM into the C2f module to enhance the model's attention towards daylily. The utilization of the CARAFE module aims to enhance the extraction of crucial features. Furthermore, the study employs the Inner-loU as a loss function to enhance the network's bounding box regression performance. The findings from the experiment indicate that the YOLOv8 Daily algorithm enhances both detection precision and recall by 1.55

The stiffness of the main bearings of a diesel engine is the research that has been focused on due to the fact that traditional evaluation indexes can hardly accurately characterize the coordinated deformation of the crankshaft and the bearing shell. Regarding the clearance margin that could judge whether motion interference occurs between the main bearing shell and the crankshaft journal after deformation as the evaluation index, the characterization method of for diesel engine main bearings based on the evaluation index and the numerical calculation method for the main bearing clearance are proposed. Experiments are designed and conducted to verify the feasibility and accuracy of the evaluation index. The results show that it is feasible to take the main bearing clearance as the evaluation index of the coordinated deformation. The relative errors between the calculation results and the measurement data of the centripetal radial deformation of the main bearing shell in the preload condition are less than 7

A method for characterizing coordinated deformation of diesel engine main bearing is proposed to address the problem that traditional evaluation indexes are difficult to accurately characterize the coordinated deformations of the main bearing shell and the crankshaft. The margin between the main bearing clearance and the assembly clearance is employed to judge whether motion interference occurs between the main bearing shell and the crankshaft journal after deformation. The numerical calculation method for the main bearing clearance is proposed. Experiments are designed and conducted to verify the feasibility and accuracy of the evaluation index. The influencing mechanisms of sensitive parameters on the coordinated deformation characteristics of the main bearing are explored based on the orthogonal test. The results show that it is feasible to take the main bearing clearance as the evaluation index of the coordinated deformation. The relative errors between the calculation results and the measurement data of the centripetal radial deformation of the main bearing shell in the preload condition and the main bearing loading condition are less than 7% and 9.7%, respectively, indicating the accuracy of the computational method. The main bearing clearances tend to decrease with the increase of horizontal and vertical bolt loads until they remain basically constant. The main bearing clearance decreases with the increase of the interference followed by a slight increase, while it is positively correlated with the side clearance. This study provides theoretical basis for main bearing stiffness reliable design of HPD diesel engine.

While Industry 4.0 improves human productivity, it also raises sustainability and social challenges. Industry 5.0, as a supplement and logical continuation of the Industry 4.0 paradigm, focuses on the development of a human-centric, sustainable, and resilient manufacturing system. This paper reviews the existing literature. First, it discusses the definition and implementation framework for Industry 5.0. Then, it expounds the application status of Industry 5.0 in the field of intelligent manufacturing from four perspectives: digital manufacturing and intelligence, human-centric intelligent manufacturing and production process management, decentralized and resilient production, and sustainable production. It summarizes the role of Industry 5.0 technology in various intelligent manufacturing scenarios, as well as the challenges it faces, and concludes by analyzing Industry 5.0’s potential development direction and future technologies. This paper believes that the application of Industry 5.0 technology will effectively improve the production capacity of intelligent manufacturing systems and promote the development of intelligent manufacturing systems in a safe, efficient, sustainable and resilient direction; the development of Industry 5.0 will focus on giving full play to human creativity, avoiding repetitive labor through human-robot collaboration, and thereby realizing human value.

In real industrial scenarios, fault data are characterized by class imbalance, a major challenge for data-driven intelligent fault diagnosis. This article proposes a novel three-level regularization framework that integrates data, models, and labels to diagnose the imbalanced fault. First, a signal to image (S2I) module is introduced, which converts 1-D signals into 2-D images to conduct research and reduce model development workload and model-specific dependencies. Then, a regularization framework is proposed consisting of three submodules, inner local feature regularization (ILFR), outer local feature regularization (OLFR), and class balance margin loss (CBML), improving the faulty health state recognition accuracy without degrading the normal health state recognition performance. Finally, adequate experiments are carried out on four mechanical fault datasets. The results show that under the extremely imbalanced conditions, the proposed framework can improve the accuracy of the baseline method by 28%, 38%, and 25% on the three datasets (including PU, JNU, and UoC), respectively. Moreover, the proposed framework outperforms the SOTA method on the CWRU dataset, which validates the effectiveness and superiority of the proposed framework.

Presently, resource-constrained devices in manufacturing Internet of Things (MIoT), such as sensors and radio frequency identification (RFID) devices, collect a large amount of privacy-sensitive data. However, weak passwords and vulnerable encryption capabilities in MIoT have often become loopholes of security risks. To this end, this paper proposes a novel secure data-sharing scheme based on the integration of blockchain and fusion of both real and fake data to address the data security requirements of resource-constrained MIoT devices. First, a computing resource collaboration architecture is designed, thereby enabling these devices to interface with multiple devices with full resource to implement flexible resource scheduling. Then, a resource-assistance mechanism is devised through blockchain-based smart contracts by utilizing the idle resources of full nodes to complement the computational tasks launched by resource-constrained nodes. In addition, polygon semantic rules are proposed to improve the security of private data. Subsequently, real data artifacts are generated by data tampering to achieve privacy cover, avoiding the consumption of computing resources in traditional encryption algorithms. Finally, the feasibility of the proposed scheme is verified on a customized candy production line. The experimental results validate that data protection using polygon semantic rules can prevent actual data from being peeped. Moreover, the results also indicate that the proposed method can obtain resource assistance from other nodes through the resource compensation mechanism.

Due to its large size of coal and high mining output, lump coal is one of the hidden risks in mining conveyor damage. Typically, lump coal can cause jamming and even damage to the conveyor belt during the coal mining and transportation process. This study proposes a novel real-time detection method for lump coal on a conveyor belt. The space-to-depth Conv (SPD-Conv) module is introduced into the feature extraction network to extract the features of the mine's low-resolution lump coal. To enhance the feature extraction capability of the model, the normalization-based attention module (NAM) is combined to adjust weight sparsity. After loss function optimization using the Wise-IoU v3 (WIoU v3) module, the SPD-Conv-NAM-WIoU v3 YOLOv8 (SNW YOLO v8) model is proposed. The experimental results show that the SNW YOLOv8 model outperforms the widely used model (YOLOv8) in terms of precision and recall by 15.82% and 11.71%, respectively. Significantly, the real-time detection speed of the SNW YOLOv8 model is increased to 192.93 f s-1. Compared to normal models, the SNW YOLO v8 model overcomes the disadvantages of normal models, such as being overweight, and the parameters of SNW YOLO v8 are reduced to only 6.04 million with a small model volume of 12.3 MB. Meanwhile, the floating point of SNW YOLOv8 is significantly reduced. Consequently, it demonstrates excellent lump coal detection performance, which may open up a new window for coal mining optimization.

Abstract Response coordination design becomes an inevitable trend based on the multidisciplinary optimization design (MOD) in the development of the design for complex mechanical structures. Aiming at the problem of unclear mechanism of response coordination for main loaded structures, a response coordination design technology characterized by a response coordination evaluation system is proposed based on an analytic hierarchy process. In addition, a nondominated sorting elephant herding optimization (NSEHO) algorithm is proposed by introducing the Pareto optimization theory into elephant herding optimization and combining fixed‐sized candidate set adaptive random testing algorithm to overcome inherent limitations of the traditional algorithm in response coordination design. Taking the main bearing assembly structure as an example, the response coordination design is conducted based on the proposed technology and NSEHO algorithm. The response coordination design scheme of the main bearing assembly structure is experimentally verified. The results show that the Pareto frontier of NSEHO is smoother and is closer to the real Pareto frontier compared to NSGA‐II in condition of small population size and restricted iteration generations, revealing the advancement of the solving ability of NSEHO. The coordination coefficients of strength, stiffness, contact strength, and mass of the main bearing assembly structure are increased by 13.91%, 14.96%, 2.63%, and 0.07%, respectively, and the overall coordination coefficient are increased by 11.10%. The indicators of strength, stiffness, and contact strength of the response coordination scheme are better than those of the original scheme by experimental verification, demonstrating the effectiveness of the response coordination design technology. The proposed technology reveals the response coordination mechanism of the main loaded structures of internal combustion engines and provides an important guiding significance to MOD for complex mechanical structures.

During the transfer process for coal with the utilization of mining conveyor belts, large pieces of coal often affect the safety of transportation, so real-time monitoring of the transport process in lump coal is essential. Therefore, a real-time monitoring method GSB YOLOv5 is proposed. Firstly, the dataset's contrast is enhanced by adaptive histogram equalization, while its richness is improved by combining Mosaic multi-data enhancement. Secondly, it proposes the utilization of Ghost Net, which is a neural network with low computational requirements for lightweight extraction and fusion of features. This approach effectively minimizes the model computation. In addition, the combination of Squeeze-Excitation mechanism improves the extraction of model's feature capabilities. Finally, a feature pyramid with weighted bidirectional is employed to accomplish multi-source information fusion by effectively merging features at varying resolutions. The experimental findings demonstrate that the enhanced GSB YOLOv5 algorithm achieves a 35.256% significant reduction in network layers, while it has substantial reductions of 63.023% and 68.582% in parameters and floating point operations, respectively. Furthermore, the compressed model size is decreased from 92.7MB to 34.4MB. In addition, there are improvements of 1.421% and 1.460% in the detection precision and recall rate of the model, respectively, while the detection efficiency of the real-time has a significant boost from 68.34 FPS to 107.91 FPS. Automatic, fast and high precision monitoring of lump coal objects on conveyor belts in underground coal mines can be realized.

Terahertz and higher frequency band wireless communication technologies represent the most promising spectrums for 6G networks. Compared to 4G/5G networks, 6G networks operate in a higher frequency band with greater propagation and penetration losses, which may bring serious challenges to network planning and green communications. The digital twin (DT) technology is able to model and simulate wireless networks to improve network performance and deployment efficiency. The artificial intelligence (AI) algorithms provide strong self-evolution and self-optimization capabilities, enabling the generation of an intelligent network. To achieve intelligent, energy-efficient, and cost-effective deployment of 6G networks in smart factories, this article proposes a DT-based system architecture and a mobile-enhanced edge computing-cloud collaborative mechanism for handling diverse and complex data. Moreover, a DT and AI-based method is developed to enable intelligent planning and deployment of 6G networks in factories, which improves network performance while reducing operational costs.

To address the problems of low precision and poor real-time performance in the process of part identification and positioning of production line assembly robotic arm, Ghost-SE YOLOv5, an assembly part identification and positioning algorithm integrating lightweight network and attention mechanism is proposed. First, the redundancy of feature map convolution is utilized, which solves the problems of large number of model parameters and floating point operations by using Ghost convolution and Ghost Bottleneck modules. Second, the attention mechanism SE Module is introduced in the backbone network to increase the propensity of feature extraction. Last, the loss function is optimized to speed up the convergence of the model. The results shows that the number of parameters, float operation per second and train time of the proposed algorithm are reduced by 45.98%, 55.99% and 24.07%, respectively. And GPU use was reduced from 7.61G to 6.43G. Furthermore, during the test the precision reached 98.6%, and the recall rate realized 95.3%. The real-time detection performance achieved 97.59 FPS, with an improvement of 34.53%. It can be seen that Ghost-SE YOLOv5 algorithm has better practicality in the part identification and positioning of robotic arm for production line assembly.

Objective. Training data with annotations are scarce in the intelligent diagnosis of retinopathy of prematurity (ROP), and existing typical data augmentation methods cannot generate data with a high degree of diversity. In order to increase the sample size and the generalization ability of the classification model, we propose a method called ROP-GAN for image synthesis of ROP based on a generative adversarial network. Approach. To generate a binary vascular network from color fundus images, we first design an image segmentation model based on U2-Net that can extract multi-scale features without reducing the resolution of the feature map. The vascular network is then fed into an adversarial autoencoder for reconstruction, which increases the diversity of the vascular network diagram. Then, we design an ROP image synthesis algorithm based on a generative adversarial network, in which paired color fundus images and binarized vascular networks are input into the image generation model to train the generator and discriminator, and attention mechanism modules are added to the generator to improve its detail synthesis ability. Main results. Qualitative and quantitative evaluation indicators are applied to evaluate the proposed method, and experiments demonstrate that the proposed method is superior to the existing ROP image synthesis methods, as it can synthesize realistic ROP fundus images. Significance. Our method effectively alleviates the problem of data imbalance in ROP intelligent diagnosis, contributes to the implementation of ROP staging tasks, and lays the foundation for further research. In addition to classification tasks, our synthesized images can facilitate tasks that require large amounts of medical data, such as detecting lesions and segmenting medical images.

Abstract The main bearing assembly structure is one of the most important main load‐bearing structures of a diesel engine. The working loads of this structure increases dramatically with the increase of the power density of diesel engines, resulting in the problem of unsatisfactory reliability coordination design in restricted design space, while the mechanisms for coordinated design are still at the blank stage. Aiming at the problem, a multilevel multiobjective coordination matching design technique is innovatively proposed in this paper. This technique is characterized by a multilevel multiobjective coordination evaluation system for the assembly structure and its components based on the improved analytic hierarchy process. The finite element model and mathematical model of the main bearing assembly structure are established to realize the joint of finite element technology and optimization technology for coordination matching design. By carrying out verification experiments, the strength coordination, deformation coordination and contact strength coordination of the assembly structure increase by 13.91%, 14.96%, and 2.63%, respectively, after matching design, while the mass coordination remains almost constant meeting the lightweight design requirements. The overall coordination of the main bearing assembly structure is improved by 11.10%, achieving the goal of matching design of the main bearing assembly structure. The results show that the coordination evaluation system can quantitatively characterize the coordination relationship of the assembly structure and the multi‐reliability of components, and it is a feasible coordination evaluation method. The demonstrated coordination evaluation system and coordination matching design modeling approach provide important theoretical guidance for the matching design of complex assembly structures.

Abstract During the transfer of coal using mining conveyor belts, broken coal and lump coal are typically present, along with a few larger pieces of coal. These larger pieces can pose a safety risk during transportation, making it crucial to monitor the lump coal in real-time throughout the process. To solve the above problems, this paper proposes the real-time monitoring method Ghost-SE-Bi FPN (GSB) YOLOv5. First, for data pre-processing, we enhance the contrast of the dataset using adaptive histogram equalization. To further enrich the dataset, we combine Mosaic multi-data enhancement techniques. Second, in this paper, we introduce Ghost Net, a lightweight neural network that performs feature extraction and fusion. This reduces the computational complexity of the model while maintaining its performance. We also incorporate the Squeeze-Excitation attention mechanism to improve feature extraction and weight adjustment, which accelerates model convergence. Finally, in order to efficiently complete multi-source information fusion, a weighted bidirectional feature pyramid is employed to fuse features of varying resolutions. According to the experimental results, the GSB YOLOv5 algorithm has been improved to reduce the number of network layers by 35.256%. Furthermore, the number of parameters and floating point operations have been reduced by 63.023% and 68.582%, respectively. The model size is compressed from 92.7M to 34.4M. In terms of detection performance, the precision and recall rate of the model detection improved by 1.421% and 1.460%, respectively. Additionally, the real-time detection efficiency increased from 68.34 FPS to 107.91 FPS. These findings demonstrate that GSB YOLOv5 not only achieves lightweight, but also effectively enhances various detection performance of the model.