System reliability analysis aims to evaluate the failure probability of engineering systems with uncertain inputs and multiple failure modes. Kriging-based active learning methods have shown high efficiency for component-level reliability but face challenges at the system level due to interactions among limit state functions. In this work, in line with the principle of the one-step look-ahead policy, a novel Kriging-based active learning method is proposed for system reliability analysis. Two key measures are derived: an upper bound of the sample misclassification probability, which quantifies the probability of incorrectly classifying the signs of sample responses, and its integral over the input space, which quantifies the uncertainty in the estimated failure probability. Building on these metrics, an Expected Uncertainty Reduction (EUR) function is developed to evaluate the potential contribution of each candidate sample-component pair in improving the accuracy of failure probability estimation. Additionally, a global convergence criterion is proposed to ensure the accuracy of the estimated failure probability before stopping the learning process. Finally, the effectiveness of the proposed method is demonstrated through several numerical and engineering examples. The results show that the proposed method achieves high accuracy while requiring fewer performance function evaluations compared with existing approaches.

PurposeAchieving high-fidelity modeling of physical dynamic behavior distinguishes digital twin (DT) from simulation. Physical manufacturing systems or processes are complex aggregates of information composed of different elements. The DT behavioral model (DT_BM) constructed based on single-model makes research fall into the dilemma of "model islands." Therefore, this article aims to propose a system-level DT_BM modeling paradigm with multi-model integration to enable more intelligent DT applications in the manufacturing.Design/methodology/approachA modeling reference architecture for DT_BM is proposed to accelerate technical standardization and promote consensus among stakeholders. A modeling method of DT_BM is proposed based on systems modeling language (SysML), which is used to realize the description of information interaction and dynamic collaboration between sub-models in DT_BM. Based on timed automaton (TA), a formal transformation framework and toolchain for DT_BM verification are designed to ensure the correctness and reliability of modelling.FindingsThe proposed modeling paradigm shows good ability in expressing complex dynamic interactions between sub-models. A case study of processing quality control proves its practicability. The results show that the proposed method can accurately, effectively and verifiably characterize the behavior of complex physical manufacturing systems or processes.Originality/valueThis article innovatively proposes a DT_BM modeling paradigm that covers the entire process of architecture, modeling and verification. Unlike traditional single-model method, it supports multi-model collaboration, dynamic interaction and correctness assurance, providing theoretical and technical support for the realization of intelligent and reliable DT systems.

In the process of material processing, the wear state of the tool is the key factor affecting the processing quality. However, the existing wear state judgment methods rely on manual experience and have low accuracy, resulting in a decline in the processing quality of the workpiece. Aiming at this problem, this paper proposes a tool wear trend prediction method based on deep residual shrinkage network and auxiliary particle filter. This method takes the vibration signal of the tool base as the input, and the surface roughness of the workpiece as the wear label. Finally, the real-time prediction value of the tool wear trend and the future multi-step prediction value can be obtained. In the comparative verification experiment, the R^2 values of real-time prediction results and future multi-step prediction results are 0.9762 and 0.72, respectively. Both are superior to the results of other classical models. The results show that compared with the traditional method, this method can predict the wear trend of the tool more accurately in a complex machining environment. This method provides a new solution for intelligent operation and maintenance in product processing, and has high application value.

Active learning Kriging is widely used in structural reliability analysis for its computational efficiency and accuracy. While numerous learning functions exist to accelerate Kriging convergence, their performance varies across problems, with no single function universally dominating. In this study, a learning function selection strategy based on the Markov Decision Process (MDP) is proposed. Specifically, the selection of learning functions is modeled as an MDP, with actions corresponding to several representative learning functions, thereby avoiding reliance on a fixed sample selection preferences. An accuracy measure for failure probability is developed and designed as the MDP reward, shifting the focus of sample selection from the state of single samples to overall model improvement. Guided by the Bellman optimality principle, the proposed method selects the learning function that maximizes the expected long-term gain in model accuracy at each iteration, thereby achieving a theoretically optimal selection strategy. Several numerical and engineering examples are adopted to validate the effectiveness of the proposed method. The results show that it effectively overcomes the limitation of blindly selecting learning functions and can even outperform the optimal learning function in the action space.

Axial piston pump in the hydraulic system is a key power component of high-end equipment. As a key component of the axial piston pump, the wear behavior of the slipper has an important impact on the stable operation of the equipment. The microstructure, wear change curve, vibration characteristics analysis, and wear mechanisms of the slipper during the wear process are investigated. The results indicate that the friction contact between the surface of the slipper and the swash plate aggravates the degree of wear, resulting in increased friction. The friction heat increases and the stability of the oil film decreases. Slipper coating and substrate material wear seriously. A vibration characteristic analysis system based on time-frequency-phase trajectory coupling is constructed, and the dynamic correlation law between nonlinear vibration response and phase space trajectory evolution during wear is revealed. During slipper wear, adhesive wear, fatigue cracking, scratching, crack formation, and fracture phenomena are primarily observed. The main wear mechanisms of the slipper include abrasive wear, adhesive wear, and oxidative wear. By observing the changes in friction and wear on the slipper-swash plate surfaces, as well as through multi-physics field coupling simulations, the changes in the friction performance and wear mechanism of the slipper-swash plate are verified. The research findings provide important technical support for improving the stability of hydraulic systems.

Industrial manufacturing environments are inherently unpredictable, leading to diverse defect manifestations influenced by factors such as ambient lighting, oil contamination, and varying damage levels. The limited diversity of defect samples constrains the effectiveness of deep learning models in defect detection. To address this challenge, this study introduces a low-shot defect detection method for metal multi-surface defects under dynamic uncertain manufacturing conditions through data decoupling augmentation. The proposed framework features a progressive upsampling decoder (Conv_ResUp) that captures fine-grained features in small-scale feature maps using standard convolution, enhancing decoding residual connections and linear interpolation. Additionally, an adaptive discriminator disturbance augmentation (ADDA) module dynamically adjusts disturbance strength based on the discriminator’s overfitting level, ensuring stable adversarial training. A defect decoupling loss module further enables unsupervised attribute separation, improving defect generation quality. Experimental results demonstrate that the proposed method outperforms conventional generative models, producing high-fidelity, diverse defect samples that better align with real manufacturing conditions. When integrated with YOLOv5 and EfficientNet, the augmented training dataset enhances defect detection performance, improving mean average precision (mAP) and F1-score by 5.95% and 5.80% for end-face and tooth-face defect detection, respectively.

High-fidelity digital twin modeling is the core of digital twin machine tool (DTMT) to achieve accurate mapping and deliver functional services. Model fusion is a key modeling technology to promote the integrity and system connectivity of DTMT. However, current model fusion lacks attention to the multi-scenario characteristics of DTMT, which hinders the effective application of DTMT. Therefore, this paper proposes a multi-scenario model fusion and verification method for DTMT to eliminate information islands, improve model collaboration and respond to dynamic application requirements. Firstly, an S3C2 architecture is proposed to guide the multi-scenario model fusion of DTMT. The S3C2 architecture helps clarify the structural relationships of multi-scenario models and mask their heterogeneity, thus enabling DTMT to fuse the right models at the right time and provide the desired digital twin service. In addition, the fusion mechanism with different topologies is also considered to support the information exchange in the multi-scenario model fusion process of DTMT. Then, a method combining SysML and it-calculus is proposed to describe the fusion behavior and verify the fusion process. Verifying the correctness of interactive behaviors and semantic consistency in the model fusion process is helpful to ensure the stability of the digital twin system and improve the utilization rate of resources. Finally, the effectiveness and operability of the proposed method is proved by a case study.

Axial piston pump is a complex and typical thermal-fluid-structural coupled system. Its reliability directly affects the operational stability of the complex hydraulic system. It faces challenges including scarce fault samples and data distribution discrepancies across operating conditions. Regarding the problem that traditional methods fail to effectively integrate and utilize multi-source information, resulting in incomplete description of fault information, this paper proposes an intelligent cross-domain industrial information integration fault diagnosis method that integrates Digital Twin and adversarial transfer. Firstly, a multi-domain coupled Digital Twin model is constructed to generate multi-source fault simulation information data. The model employs co-simulation of multi-body dynamics and hydraulic systems to ensure the physical fidelity of fault information. Multi-source fused Gramian Angular Summation Fields feature encoding is designed to map multidimensional signals into two-dimensional spatiotemporal correlation images, thereby integrating and enhancing the representation of information. Secondly, an improved Auxiliary Classifier Generative Adversarial Network with multiple generators is adopted to align the distributions of simulated and measured data, with a dynamic optimization strategy employed to enhance generation quality. Finally, a Multi-scale Attention Domain Adversarial Transfer Network is constructed, combining a Gradient Reversal Layer and Conditional Maximum Mean Discrepancy to suppress the cross-domain distribution differences between the simulation and the experimental data. The experiment shows that by integrating experimental and simulation data, the proposed method achieves an average accuracy of over 98 % in cross-condition fault diagnosis tasks under unknown conditions, showing significant improvement over traditional transfer learning methods. Ablation studies validate the effectiveness of each module, providing a novel approach for complex hydraulic system fault diagnosis under small-sample scenarios.

Defect recognition is an effective quality control measure for the key manufacturing process nodes of industrial products. However, current defect recognition models, which rely on a large amount of supervised data, cannot quickly adapt and identify new classes of defects that appear in the production process. Moreover, automated industrial defect recognition is plagued by the problems of high expert annotation costs, scarce samples, and a wide variety of defects. To address these issues, this study explored a few-shot defect recognition method for the multidomain industry using attention embedding and fine-grained feature enhancement. To enhance the feature representation of highly discriminative regions in images, a spatial multidirectional group attention convolutional backbone network (Res12_SMGA) was proposed. A multi-directional feature-enhanced representation of the group channel space was achieved by grouping the channels and generating direction-sensitive feature mappings. In addition, a lightweight feature enhancement branch (LFEB) was designed to select more discriminative pixel-level features from the enhanced high-dimensional features, while aiming to improve the generalization ability for multidomain industrial few-shot defect recognition tasks. Subsequently, the different granularity feature outputs of Res12_SMGA and LFEB were fused, and defect recognition was completed using the Nearest Class Mean classifier (NCM). Experiments showed that the proposed method achieved optimal performance in identifying 10 types of defects in five industrial product domains. For the 10way-5shot and 5way-5shot settings, the classification accuracies were 91.71% and 95.97%, respectively. Thus, this study demonstrated that new industrial defects can be recognized using only a few labelled samples.

The collected vibration signals of rotating machinery contain pulses, missing, and other low-quality anomalous data due to environmental noise interference, unstable data transmission, and data acquisition instrument failure. These low-quality data obstruct the analysis of the healthy operation condition of rotating machinery. This paper proposes a method for anomalous vibration signal detection and recovery based on the local outlier factor algorithm and the modified sparsity adaptive matching pursuit algorithm. The method combines the local outlier factor algorithm and compressive sensing theory to realize anomalous vibration signal detection and recovery. This paper evaluates the recovery performance both qualitatively and quantitatively and discusses how the proposed method's hyperparameter selection affects the recovery results. A set of simulated signal and measured hob base signal are used to verify the proposed method. The results indicate that, when compared to the other seven reconstruction algorithms, the proposed method's recovered signal has the lower error level and the higher waveform similarity which reaches more than 98% to the original signal, effectively improving data quality.

Reasonable deployment of temperature sensors is the key to accurately monitoring the temperature field of machine tools and improving the accuracy of thermal error prediction and compensation models. To determine the optimal deployment location of sensors, this paper proposes a temperature-sensitive points selection method tightly coupled with rough set and multi-objective optimization. Firstly, the importance of each temperature measurement point to the thermal error is calculated based on the rough set, and information entropy is introduced to amplify the importance difference among adjacent measurement points at the same heat source. Then, with the temperature measurement points groups as the variables, the number of temperature measurement points in the group, and the information importance of the group as the objectives, a multi-objective attribute reduction model is established, which transforms the temperature-sensitive points selection problem into a discrete multi-objective optimization problem. Finally, a multi-objective adaptive hybrid evolutionary algorithm is proposed, which designs a population initialization method based on mutual information and interval probability, and dynamic adaptive evolutionary parameters to achieve optimal temperature-sensitive points selection. Experiments on the high-speed dry hobbing machine verify the superiority and effectiveness of the proposed method.

Current mainstream wear diagnosis approaches rely on the entire lifecycle wear-labeled datasets, which is costly, delayed, and low-universality. To achieve the monitoring goal of online, real-time, and strong generalization for hob wear and get rid of the dependence on the wear datasets, an online unsupervised monitoring method is explored based on the multi-domain features extraction and improved Q-statistic control chart, which is progressive. First, the monitoring feature matrix of each hob shifting period is transformed into the principal component subspace, the wear subspace, and the interference subspace based on the principal component analysis (PCA). Then, features in the wear subspace of the previous shifting period are used to get the control limit and those of the next shifting period are used to calculate the Q-statistic. The hob wear status is determined by comparing the control limit and the Q-statistic. The gear hobbing experiment verifies the feasibility of the proposed method.

Abstract Purpose The secondary prevention strategy for cardiovascular disease (CVD) does not include anti-inflammatory treatment, which may lead to some patients being in a high inflammatory state for a long time. The aim of this study was to assess the association between the residual inflammatory risk based on Glasgow Outcome Score (GPS) and long-term mortality in patients with CVD. Methods This study included 3833 patients (≥ 20 years old) with CVD in the National Health and Nutrition Survey from 1999 to 2010. The death result is determined by the correlation with the national death index on December 31, 2019. GPS consists of serum C-reactive protein and albumin. The main outcome was all-cause death, including cardiac death and non-cardiac death. The Cox proportional hazards adjusted for demographic factors and traditional cardiovascular risk factors were used to test the impact of GPS level on mortality. The sensitivity analysis included components of CVD, heart failure, coronary heart disease, angina, heart attack, and stroke. Results Among 3833 CVD patients with a median follow-up of 9.6 years, 2431 all-cause deaths, 822 cardiac deaths, and 1609 non-cardiac deaths were recorded. After full model adjustment, compared with the GPS (0) group, the risk ratio (HR) of all-cause death for GPS (1) and GPS (2) were 1.667 (95% confidence interval (CI), 1.490–1.865) and 2.835 (95% CI, 2.077–3.869), respectively (P for trend < 0.001). Compared with the GPS (0) group, the HR of cardiac death for GPS (1) and GPS (2) were 1.693 (95% CI, 1.395–2.053) and 2.268 (95% CI, 1.264–4.070), respectively (P for trend < 0.001). Compared with the GPS (0) group, the HR of non-cardiac death for GPS (1) and GPS (2) were 1.656 (95% CI, 1.443–1.901) and 3.136 (95% CI, 2.171–4.530), respectively (P for trend < 0.001). The results of the sensitivity analysis were almost consistent with the overall cohort. Conclusions Using the US national database, and adjusting for a large number of potential confounders through flexible modeling, we found that residual inflammatory risk based on GPS was strongly associated with a increased risk of death in patients with CVD and that the higher GPS level was associated with an increased risk of death, and this score, which consists of readily available biomarkers, may in the future be used for risk stratification and potentially for improving patient outcomes.

Automatic manufacturing feature recognition (AFR) is a critical technology for realizing CAD/CAPP/CAM integration in the era of intelligent manufacturing. Despite the numerous feature recognition approaches that have been proposed, the recognition of intersecting features remains a challenge. The most important reason is that the boundaries of features and their geometric topological information are altered or destroyed when interacting with other features. To address this problem, this paper proposes a manufacturing feature recognition method based on graph and minimum non-intersection feature volume suppression. Firstly, the geometric and topological information of a part is extracted and represented as an attributed adjacency graph. Then, a subgraph isomorphism algorithm is designed to recognize the features corresponding to each subgraph. After that, a method to construct and suppress the minimum non-intersection volume of the recognized features is introduced to repair the boundaries of intersecting features. On this basis, intersecting features are separated and recognized at different stages. The experimental results indicate that the proposed approach is effective in recognizing intersection features. Moreover, the recognition result of the proposed method incorporates the surface and volume of a feature, providing enriched feature information for engineering applications.

The correct understanding of the effective frequency components distribution in different states is essential for high precision hob fault diagnosis and wear identification. To explore the unclear vibration frequencies, a hob dynamic differential equation is established. The frequency components of gear hobbing under healthy states, steady faults, impact faults, and composite faults are summarized theoretically, especially the higher harmonics and the corresponding modulation sidebands, which are mainly the results of the convolution, overlapping, and the nonlinear feedback of the hobbing frequency and its harmonics, meshing frequency and its harmonics, rotating frequency and its harmonics. Moreover, the influence of the hob second response function on the amplitudes of the vibration components are analyzed, which has a relatively negative effect on the amplitudes of the lower frequency vibration components but markedly increases the vibration components near the natural frequency. The above theoretical analysis is verified by the acquired hob vibration signal.

The acquired hob vibration signals are inevitably contaminated by noise in the industrial environment, which changes the vibration signal frequency distribution and reduces the accuracy of feature extraction and hob wear identification. To solve this problem, a novel hob vibration signal denoising and effective feature enhancement method, CEEMDAN-FRS, is proposed based on the improved complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and fuzzy rough sets (FRS). First, the effective frequency distributions of gear hobbing, particularly the modulation sidebands, were obtained as prior knowledge by analyzing the hob vibration response mechanism. Then, the two key parameters of CEEMDAN (i.e., noise standard deviation and ensemble size) were adaptively determined based on signal characteristics to achieve improved decomposition compared with the use of fixed values. The evaluation and selection of intrinsic mode functions (IMFs) based on a single feature such as kurtosis or root mean square, only focus on the partial characteristics, leading to an identification bias. Thus, 11 features with different sensitivities of IMFs are weighted based on FRS, and fused as a unified feature to conduct a comprehensive IMF evaluation. Finally, a reweighted IMF reconstruction strategy is proposed. The comparisons of the proposed method and related approaches to hob vibration signals show that the proposed method achieves better performance in terms of effective feature enhancement, noise removal, and signal-to-noise ratio improvement

Augmented reality(AR) plays an important role in geography teaching for creating interactive and immersive experiences. Combining object detection algorithms with AR can identify the specified content quickly and thus overlay digital content to the real world. However, finger occlusion in AR interactions has a bad influence on object detection, which will affect the users' experience. In this paper, we focus on the detection of country regions on a globe, and aim to improve the performance of object detection in practical AR development. Firstly, we propose a geographic region recognition approach based on region missing-completion. Specifically, we design a supplementary algorithm PCCNet to infer the obscured country by utilizing the invariance of relative position between countries. Moreover, to reduce manual annotation and enrich the virtual dataset, we design a scalable automatic annotation system based on the Unreal Engine and construct a virtual globe dataset named DGAR. Finally, we build an AR geography-assisted teaching system to recognize the area pointed and play multi-media materials. Experiment results show that our proposed approach effectively improves the recognition accuracy from 88.5% to 94%. The practical significance and value of the proposed recognition approach have been confirmed based on the positive user experience with the AR system, highlighting its efficacy in real world scenarios.

To improve the fault diagnosis performance of rotating machinery under harsh conditions, a weighted average selective ensemble strategy of deep convolutional models based on the grey wolf optimizer (GWO) is proposed. Firstly, two datasets in the time domain, two in the frequency domain, and one in the time–frequency domain are respectively constructed to guarantee the diversity and comprehensiveness of input expression. Secondly, two deeper Resnet18 models, which are sensitive to the comprehensive and abstract features, and two relatively shallow CNN models, which focus on the details more, are built. Thus, the fault differential features are diverse. Moreover, the improved convolutional block attention module (CBAM) is used to enhance the diagnosis performance. Then, a total of ten individual models are trained. Thirdly, F1 Score is used to evaluate the diagnostic performance of each individual model on different faults, and the fault class-specific thresholds are set. For each fault, individual models with F1 Score lower than the corresponding thresholds are regarded as negative and need to be discarded. Especially, the fault class-specific thresholds are optimized by GWO. Finally, the weighted average selectively ensemble strategy is implemented based on the threshold-treated weights. Experiment results indicate that the proposed ensemble model significantly improves the diagnostic accuracy and stability of the individual models, which is also verified by other compared ensemble strategies and ensemble models.

Defect detection plays an important role in implementing zero-defect manufacturing (ZDM) and improving the sustainability of manufacturing systems. The remarkable diversity of gear types, the inhomogeneity of end-face structure, as well as the small size and multi-scale of defects, are the common problems confronted during the metal gear end-face defect detection, which leads to poor performance of existing detection methods in terms of detection rate and accuracy. To address the problems above, this study proposes a cascaded combination method SR-ResNetYOLO to automatically detect the defects by region extraction and multi-scale fusion of sampled features under 16X. To obtain more effective features, this study proposes the visual-saliency-based method to extract the machined area image, eliminating the interference between the invalid features of non-machined areas and edge burrs and reducing thereby the image complexity. Subsequently, establish a 16X down-sampled feature extraction backbone network (ResNet-21), to efficiently obtain the high-resolution features of the defects by using the machined area images as input. With the multi-scale fusion module, the min-scale feature map, output by the ResNet-21, fuses at the medium- and large-scales. Finally, the three-fused-scale feature maps are classified and located by the location and classification module. The proposed method achieves satisfactory performance in terms of the mAP and recall rate, which are respectively 96.66% and 97.07%, and the average computation time of the detection for per image is 0.12 s, which can effectively detect small size and multiple scale defects of metal gear end-face.