With the general rise of globally distributed enterprise layout and assembly-to-order production mode, the multi-factory collaborative production mode is able to coordinate multi-factory and cross-region production problems, and has become an important development direction for enterprise management. In this study, a bilevel interactive optimization model for low-carbon product family design and multi-factory collaborative productionconfiguration is developed, aiming to maximize profit while controlling carbon emissions. In this model, we consider both the potential benefits and low-carbon costs of the products. In addition, the uncertainty in the process is expressed as interval numbers and reformulated as a multi-objective optimization problem to optimize both the average and fluctuation levels of the model simultaneously. In order to balance the solution quality and computational efficiency, a nested genetic algorithm with branch-and-bound is proposed to solve the problem. Numerical experiments were conducted using a laptop product family as an industrial case study: Compared with the traditional two-stage method, our model achieved a 47.14% increase in profit and an 11.23% reduction in carbon emissions. Compared with the conventional nested bilevel interactive genetic algorithm, the proposed algorithm reduced total costs by approximately 35.5% and improved computational efficiency by approximately 177 times.Sensitivity analysis further investigates the adaptive strategies of the proposed model at different number of variants, low carbon levels and uncertainty levels, highlighting the practical applicability of the proposed approach.

The wear conditions of honeycomb sealing rings in aerospace engines are often complex. Traditional human operations based on sample paste exhibit poor adaptability and are inefficient. This paper proposes an automated detection method for the geometric features of wear marks on honeycomb sealing structures based on depth ratio features. Adaptive identification and quantification of cellular wear through point cloud data analysis. First, the point cloud data is cropped, followed by a least-squares fit iterative method to compute a reference line at a specified cross-section, which serves as a standard for computing the width and depth of wear marks while denoising the point cloud data. Subsequently, N-neighborhood sets and the depth ratio features within these sets are introduced, transforming the task of detecting wear marks' start and end points into a peak detection problem. An improved automatic multiscale-based peak detection (AMPD) algorithm with a masking mechanism is utilized to determine the extent of each wear mark. Finally, the geometric features are calculated for each wear mark. Experimental results demonstrate that the proposed method can robustly identify wear areas with varying depths and distributions, measurement time reduced by more than 90%, and fulfilling the requirements for identifying and measuring honeycomb wear marks.

Nomex honeycomb composites (NHCs) are highly sensitive to the abnormal wear state of disc tools during cutting, leading to poor product quality. This paper proposes a real-time anomaly detection method combining a novel CNN–GRU–Attention (CGA) deep learning model with an Exponentially Weighted Moving Average (EWMA) control chart to monitor sensor data from the disc tool. The CGA model integrates an improved CNN layer to extract multidimensional local features, a GRU layer to capture long-term temporal dependencies, and a multi-head attention mechanism to highlight key information and reduce error accumulation. Trained solely on normal operation data to address the scarcity of abnormal samples, the model predicts cutting force time series with an RMSE of 0.5012, MAE of 0.3942, and R2 of 0.9128, outperforming mainstream time series data prediction models. The EWMA control chart applied to the prediction residuals detects abnormal tool wear trends promptly and accurately. Experiments on real NHC cutting datasets demonstrate that the proposed method effectively identifies abnormal machining conditions, enabling timely tool replacement and significantly enhancing product quality assurance.

Predicting the time series energy consumption data of manufacturing processes can optimize energy management efficiency and reduce maintenance costs for enterprises. Using deep learning algorithms to establish prediction models for sensor data is an effective approach; however, the performance of these models is significantly influenced by the quantity and quality of the training data. In real production environments, the amount of time series data that can be collected during the manufacturing process is limited, which can lead to a decline in model performance. In this paper, we use an improved TimeGAN model for the augmentation of energy consumption data, which incorporates a multi-head self-attention mechanism layer into the recovery model to enhance prediction accuracy. A hybrid CNN-GRU model is used to predict the energy consumption data from the operational processes of manufacturing equipment. After data augmentation, the prediction model exhibits significant reductions in RMSE and MAE along with an increase in the R2 value. The prediction accuracy of the model is maximized when the amount of generated synthetic data is approximately twice that of the original data.

As global climate change intensifies, assessing product carbon footprints serves as a foundational step for quantifying greenhouse gas emissions throughout a product’s lifecycle, forming the basis for achieving sustainability and emission reduction goals. Traditional lifecycle assessment methods face challenges such as subjective boundary definitions and time-consuming inventory construction. This study introduces PCF-RWKV, a novel model based on the RWKV architecture with task-specialized low-rank adaptations (LoRAs). Trained on carbon footprint datasets, the model minimizes memory use and data interference, enabling efficient deployment on consumer-grade GPUs without relying on cloud computing. By integrating multi-agent technology, PCF-RWKV automates the creation of lifecycle inventories and aligns production processes with emission factors to calculate carbon footprints. This approach significantly improves the efficiency and security of corporate carbon footprint assessments, providing a potential alternative to traditional methods.

This study proposes a hybrid process method based on Cryogenic Minimum Quantity Lubrication and Ultrasonic Rolling Strengthening Process (CMQL-URSP) to improve the surface integrity and enhance the mechanical properties of blade surface. First, the principle of the CMQL-URSP hybrid process was described. Second, the TEM experiments were conducted to clarify the influence of the CMQL-URSP hybrid process on aircraft engine blades. Finally, the blade processing experiments were performed to demonstrate the effectiveness of the CMQL-URSP hybrid process on the blade surface micro-morphology, micro-hardness and residual stress. The results indicate that the CMQL-URSP hybrid process effectively overcomes the limitations inherent to the solitary application of URSP. The CMQL process inhibits work hardening, thereby allowing the URSP to achieve better low-plastic surface improvement. Grain refinement governed by the continuous dynamic recrystallization mechanism leads to the formation of nanocrystals, which significantly improves mechanical properties of blade surface. The CMQL-URSP hybrid process results in a reduction in surface roughness from Ra 2.644 μm to Ra 0.055 μm, an increase in surface residual compressive stress from 321 MPa to 657 MPa, and the formation of a reinforced layer with a depth of 3.7 mm on the blade, thereby enhancing the surface integrity of the blade.

Abstract Archaeological illustration is a graphic recording technique that delineates the shape, structure, and ornamentation of cultural artifacts using lines, serving as vital material in archaeological work and scholarly research. Aiming at the problems of low line accuracy in the results of current mainstream image generation algorithms and interference caused by severe mural damage, this paper proposes a mural archaeological illustration generation algorithm based on multi-branch feature cross fusion (U2FGAN). The algorithm optimizes skip connections in U2Net through a channel attention mechanism, constructing a multi-branch generator consisting of a line extractor and an edge detector, which separately identify line features and edge information in artifact images before fusing them to generate accurate, high-resolution illustrations. Additionally, a multi-scale conditional discriminator is incorporated to guide the generator in outputting high-quality illustrations with clear details and intact structures. Experiments conducted on the Dunhuang mural illustration datasets demonstrate that compared to mainstream counterparts, U2FGAN reduced the Mean Absolute Error (MAE) by 10.8% to 26.2%, while also showing substantial improvements in Precision (by 9.8% to 32.3%), Fβ-Score (by 5.1% to 32%), and PSNR (by 0.4 to 2.2 dB). The experimental results show that the proposed method outperforms other mainstream algorithms in archaeological illustration generation.

To improve robot relocalization accuracy in both static and dynamic environments, we introduce a novel network, FusedNet, which incorporates a cross-attention to fuse global and local image features for end-to-end relocalization. This approach relies solely on a monocular camera sensor that is fixed on the mobile robot, and directly predicts the absolute pose from the input RGB image. Additionally, we have collected a mobile robot relocalization dataset, termed moBotReloc, consisting of dynamic large-scale scenes, using the Unity 3D simulation platform and a real mobile robot. Through extensive experiments on 7Scenes and moBotReloc, we demonstrate that FusedNet achieves significant accuracy in 6-DoF camera relocalization in static scenes, and exhibits superior relocalization performance in dynamic large-scale scenes for mobile robot applications, outperforming existing end-to-end methods that rely solely on a single global or local feature.

Objective: Infection by Demodex mites is a significant factor in causing blepharitis. The collected superficial eyelid tissue images currently have low resolution and strong noise. High-quality reconstruction of low-quality images is needed to assist ophthalmologists in accurately and quickly diagnosing mite infestations. Methods: We propose a novel GAN-based super-resolution reconstruction model. We utilize Residual-in Residual Dense Blocks (RRDB) as the fundamental building units and employ high-order degradation to simulate real-world degradation. Additionally, we introduce a new attention mechanism, All Attention Mechanism (AAM), which effectively captures and utilizes crucial image features, expanding the scope of utilized pixels. This significantly enhances the quality of reconstructed images. Results: Compared to existing super-resolution reconstruction models, our model demonstrates the strongest generalization ability and robustness. On the CIDMG and FMD datasets, we achieve the lowest NIQE and the highest SSIM values. Our model outperforms the SOTA SwinIR model, reducing reconstruction time by 76% and improving the image quality by 6.6% for low-noise images and 15.3% for high-noise images. Conclusion: By combining GAN with AAM, we have significantly improved the model's generalization ability, robustness, and real-time performance. Significance: This is the first time that domain generalization has been used for reconstructing superficial eyelid tissue images. The stable and fast reconstruction performance reduces the impact of noise and low resolution on image quality, significantly reducing examination time and improving diagnostic accuracy. This greatly enhances the clinical utility of confocal microscopy for eyelid demodex examination.

Abstract Archaeology drawing is a graphic recording technique that delineates the shape, structure, and ornamentation of cultural relics with lines, serving as a vital material in archaeological work and scholarly research. Aiming at the problems of low line accuracy and serious disease interference in the results of current mainstream image generation algorithms, this paper proposed an archaeology drawing generation algorithm based on multi-branch feature cross fusion (U2FGAN). The algorithm optimized skip connections in U2Net through channel attention mechanism, thereby constructing a multi-branch generator consisting of a line extractor and an edge detector, which separately identified line features and edge information within relic images before fusing them to generate accurate high-resolution line drawings. Additionally, a multi-scale conditional discriminator was incorporated to guide the generator towards outputting high-quality line drawings with clear details and intact structures. Experiments conducted on the Dunhuang mural line drawing datasets demonstrate that compared to mainstream counterparts, U2FGAN achieved a reduction in Mean Absolute Error (MAE) by 10.8–26.2%, while also exhibiting substantial improvements in Precision (by 9.8–32.3%), Fβ-Score (by 5.1–32%), and PSNR (by 0.4 to 2.2 dB). Experimental results show that the proposed method outperforms other mainstream algorithms in the task of archaeological line drawing generation.

As global climate change intensifies, assessing product carbon footprints has become essential for measuring greenhouse gas emissions throughout a product’s lifecycle, a key factor in achieving sustainable development and emission reduction goals. Traditional lifecycle assessment methods encounter challenges such as subjective product boundary determination and lengthy lifecycle inventory construction. Recent advancements in large language models offer new opportunities for rapid, knowledge-based content generation, yet these models often lack domain-specific training, face data security concerns, and have complex deployment requirements. This study introduces the PCF-RWKV model, which utilizes the RWKV architecture with multiple stacked residual blocks and three task-specialized Low-Rank Adaptations (LoRAs). Trained on a carbon footprint assessment-specific dataset using low-rank adaptive techniques, the model minimizes data interference and memory waste, enabling efficient deployment on consumer-grade single GPUs without relying on cloud storage.By integrating Multi-Agents technology, PCF-RWKV automates the construction of LCI for production processes, aligns production processes with emission factors to calculate carbon footprints, thereby enhancing the efficiency and security of enterprise carbon footprint evaluations and addressing the limitations of traditional methods.

Most existing robotic datasets capture static scene data and thus are limited in evaluating robots' dynamic performance. To address this, we present a mobile robot oriented large-scale indoor dataset, denoted as THUD (Tsinghua University Dynamic) robotic dataset, for training and evaluating their dynamic scene understanding algorithms. Specifically, the THUD dataset construction is first detailed, including organization, acquisition, and annotation methods. It comprises both real-world and synthetic data, collected with a real robot platform and a physical simulation platform, respectively. Our current dataset includes 13 larges-scale dynamic scenarios, 90K image frames, 20M 2D/3D bounding boxes of static and dynamic objects, camera poses, and IMU. The dataset is still continuously expanding. Then, the performance of mainstream indoor scene understanding tasks, e.g. 3D object detection, semantic segmentation, and robot relocalization, is evaluated on our THUD dataset. These experiments reveal serious challenges for some robot scene understanding tasks in dynamic scenes. By sharing this dataset, we aim to foster and iterate new mobile robot algorithms quickly for robot actual working dynamic environment, i.e. complex crowded dynamic scenes.

To improve the efficiency and controllability of our previous proposed cross-category product assembly line, we design an assembly language (A-code) to express a product's assembly process. We further develop an IDE to describe assembly processes as statements, which are organized according to predefined syntax as an A-code file. The interpreter translates statements into executable low-level commands. In addition, a four-layer architecture of the A-code assembly system (ACAS) is proposed to implement this language, thus an A-code files can be run on the assembly line physically. The proposed ACAS can reconfigure each unit for specific products, and control their assembly processes. The expressivity of this language is validated in two assembly cases, a simple shuttle valve and a complex relief valve. The functionality and feasibility of this system are tested by assembling 50 relief valves. The results demonstrate our ACAS can perform complex assembly tasks in a more efficient way.

The assembly and service of thin-walled parts are usually threatened by machining deformation. In this study, phase change material (Sn-Bi-Pb-Cd alloy with a melting point of 70 degrees C) was used to develop an adaptive clamp system, which could adjust the clamping position when the material was liquid to follow the stress-releasing deformation after rough machining. The material's solidification could re-clamp the workpiece, whose internal stress had been minimized. Afterwards, the workpiece deformation could be removed in finishing machining. Comparing to the counterpart using traditional vacuum fixture, the maximum deformation and the flatness tolerance of the web using the adaptive clamp system reduced by 82.0 % and 72.9 %, respectively, thus effectively improving the accuracy of thin-walled parts manufacturing.

The aero-engine casing, with its thin-walled complex structure, is a critical component that significantly influences machining quality due to its low-stiffness dynamic characteristics. In this study, we propose a multi-point flexible adaptive clamping technology to enhance the local stiffness of large-scale aero-engine casings. This approach aims to mitigate deformation during milling and drilling processes and improve precision throughout the multi-process machining procedure. Firstly, we analyze the milling and drilling processes involved in multi-process machining of aero-engine casings and construct a comprehensive error transfer model that considers both geometric errors and coupling effects caused by machining deformation. Furthermore, we elucidate the principles behind positioning using multi-point flexible clamping fixtures and controlling machining deformation. Finally, through simulation analysis of machining errors as well as actual machining experiments, we verify the effectiveness of our proposed multi-point flexible clamping fixture in suppressing deformation during milling and drilling processes. Our results demonstrate that this method effectively controls casing deformation during machining: it reduces flatness error at the casing mounting edge by 38.3%, while decreasing verticity error and position error at the casing mounting hole by 40.2% and 33.1%, respectively.

The on-machine inspection occupies certain manufacturing time, so it is important to improve the inspection accuracy as much as possible on the premise of an acceptable sampling scale. We proposed efficient adaptive sampling methods for NURBS curve and surface based on deviation analysis. The deviation is defined as the difference between the theoretical and reconstructed curves. For curve sampling, the less significant points are removed iteratively from initial dense on-curve points. In addition, we derived a closed solution for curve deviation, thus it is superior to existing methods in terms of reconstruction accuracy and time consumption. The curve sampling algorithm is further extended to surface sampling by simplifying it into curve sampling in two directions. Our methods are compared with classic sampling strategies and the results show that the curve and surface reconstruction errors of our method are reduced by 62% and 71% respectively.

为了准确掌握滚动轴承剩余寿命信息,评估轴承的退化状态,提出了一种基于深度学习理论的卷积神经网络模型,对轴承剩余使用寿命进行预测.通过选取最新的ResNeXt作为网络骨干,设计卷积神经网络模型.该网络模型可以堆叠大量的卷积层从而抽取到丰富的语义特征,即使在训练数据较少时仍然具有很好的泛化能力.最后在公开数据集上对算法进行了训练和验证,表明该方法可以根据滚动轴承的振动信号较为准确地对轴承的剩余使用寿命进行预测.

The on-machine inspection technique requires a certain manufacturing time, so it is important for a sampling approach to achieve high precision for a fixed number of inspection points. This study designs an efficient adaptive sampling method for the non-uniform rational basis spline (NURBS) curves and surfaces based on deviation analysis. For the free-form curves, it is an iterative method that is used to remove points that are less significant to the reconstruction error from the dense points on the curve. That is, the points are ranked by their maximum deviation from the theoretical curves. Different from the existing methods, a closed-form is derived to approximate the maximum deviation by analyzing the curve reconstruction method, i.e., piecewise cubic spline interpolation. The proposed method is compared with recent curve sampling methods, and the comparison results have shown that, under the same number of inspection points, the reconstruction error of the proposed method is reduced by 82%. The proposed curve sampling algorithm is then further extended to surface sampling, where the global characteristics of a surface are extracted as a series of curves on the surface. Thus, surface sampling is simplified to curve sampling in two directions. The proposed surface sampling strategy is compared with classic surface sampling methods using three representative surfaces. The results show that by using the proposed surface sampling strategy, the reconstruction error is reduced significantly. By applying our sampling method to the on-machine inspection system, the inspection accuracy can be greatly improved.

With the rising demand for customized and frequently updated products, the current assembly systems require a higher level of reconfigurability and flexibility. Thus we proposed a scheme of reconfigurable assembly line and hierarchical control system. To increase the flexibility of control systems, a communication method and control strategy is presented in this paper. This approach will endow the reconfigurable assembly system with flexibility by encapsulating the ladder diagram program and providing an external interface for the control software to call. This approach is validated with one assembly unit assembling shuttle valve, which takes 40s during the process. The control system can not only monitor the assembling process, but also control the action of the actuator via control panel or control commands.