Advanced Engineering Informatics最新文献

{"title":"Relative position contrast learning updated temporal convolutional network encoder multivariate time series large language model for the fault detection of wind turbines","authors":"Tian Wang,&nbsp;Linfei Yin","doi":"10.1016/j.aei.2025.103706","DOIUrl":"10.1016/j.aei.2025.103706","url":null,"abstract":"<div><div>Wind turbine fault detection is essential for reliable wind farm operation. Existing methods often need extensive training data to ensure model performance. Recently, pretrained large language model (LLM) has been employed to represent time-series data. However, the different compositions of temporal and linguistic data challenge the ability of LLM to represent time-series. The study proposes a relative position contrast learning updated temporal convolutional network encoder multivariate time series large language model (RPCL-updated TCNE-MTS-LLM) algorithm for the fault detection of wind turbines. The proposed RPCL-updated TCNE-MTS-LLM algorithm primarily consists of a proposed RPCL-updated TCNE and a fine-tuning process. Specifically, the proposed RPCL-updated TCNE-MTS-LLM algorithm tokenizes preprocessed MTS through a sliding window. The RPCL-updated TCNE encodes the tokenized data and embeds into fine-tuning process. Meanwhile, the encoded features are max pooling to obtain the pooling features and embedded into the proposed RPCL; the proposed RPCL updates the proposed TCNE to enhance the capture of time-series features. Finally, fine-tuning process optimizes the proposed RPCL-updated TCNE-MTS-LLM algorithm for downstream fault detection tasks. Experimental results on real datasets show that the RPCL-updated TCNE-MTS-LLM algorithm outperforms existing methods in wind turbine fault detection.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"68 ","pages":"Article 103706"},"PeriodicalIF":9.9,"publicationDate":"2025-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144721234","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Machine learning in chemical–mechanical planarization: A comprehensive review of trends, applications, and challenges","authors":"Georg Winkler ,&nbsp;Tom Rothe ,&nbsp;Mudassir Ali Sayyed ,&nbsp;Linda Jäckel ,&nbsp;Jan Langer ,&nbsp;Harald Kuhn ,&nbsp;Martin Stoll","doi":"10.1016/j.aei.2025.103663","DOIUrl":"10.1016/j.aei.2025.103663","url":null,"abstract":"<div><div>Chemical–mechanical planarization (CMP) is a critical and complex process in semiconductor manufacturing, where high precision and tight tolerances demand sophisticated process control. Machine learning (ML) has emerged as a promising tool to support this need. However, the literature is fragmented across domains and often lacks concrete guidance on deploying ML methods in real-world CMP environments. This review addresses that gap and provides actionable insights for both academic researchers and industrial practitioners. This work systematically analyzes peer-reviewed publications at the intersection of ML and CMP. It offers three significant contributions: First, it provides a detailed assessment of the types and sources of data used in CMP-related ML studies, distinguishing between equipment-level data, process-level data, and product-level data. Second, it introduces a structured classification of ML applications across four key areas: Research and Development, Real-Time Process Control, Equipment Health Monitoring, and Post-Process Feedback and Analysis. Third, it synthesizes commonly reported challenges along the data science pipeline. Where available, practical solutions and implementation strategies are discussed. Across these areas, existing studies illustrate ML’s potential to, among other things, enable more adaptive control, reduce metrology demands, enhance defect detection, and support layout- and process-aware design optimization, highlighting its growing role in CMP process innovation. The review concludes by outlining open research gaps and proposing future directions, including the development of benchmark datasets, integration of domain knowledge, and extension of modeling efforts to additional targets such as dishing and within-wafer non-uniformity. While focused on CMP, many challenges discussed, such as data leakage, sparse target labeling, and hybrid modeling, are relevant across engineering informatics. Addressing them in CMP can inform the broader development of robust and adaptable ML systems in intelligent manufacturing.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"68 ","pages":"Article 103663"},"PeriodicalIF":9.9,"publicationDate":"2025-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144721238","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Intelligent integrity detection and damage localization of pile from low-strain test using deep learning with accelerated training via variable-order gradient descent","authors":"Guangyao Chen ,&nbsp;Chan Ghee Koh ,&nbsp;Sihao Li ,&nbsp;Ziyang Jiang ,&nbsp;Yangze Liang ,&nbsp;Zhao Xu ,&nbsp;Daguo Wu","doi":"10.1016/j.aei.2025.103690","DOIUrl":"10.1016/j.aei.2025.103690","url":null,"abstract":"<div><div>Low-Strain Testing (LST) is widely used for pile integrity assessment, where wave velocity signals enable engineers to qualitatively evaluate pile’s integrity and locate damage. However, this method heavily relies on the expertise of inspectors, involving subjectivity, uncertainty, and inefficiency. To enhance the accuracy, efficiency, and automation of pile quality inspection, this study proposes a novel two-stage deep learning framework for pile integrity recognition and damage localization, using ShuffleNet and U-Net1D. In the first stage, the framework addresses the low energy and narrow frequency characteristics of one-dimensional LST signals by applying Gramian Angular Field, Continuous Wavelet Transform, and Short-Time Fourier Transform to generate three feature images, which are combined into an RGB image and input into ShuffleNet for pile integrity detection. In the second stage, U-Net1D is employed for damage localization. To overcome challenges posed by sparse, discontinuous and discrete binary damage labels, a novel damage label construction strategy is introduced to more precise guidance for localization, based on Gaussian membership functions and the probability sum approach. Additionally, to improve model training speed and deployment, a variable-order gradient descent algorithm called λ-FAdaMax is proposed, combining Caputo fractional derivatives with the Adam optimizer to balance convergence speed and precision. Experimental results show that, under λ-FAdaMax, the framework achieves 99.17 % accuracy on the training set and 100 % on the test set for pile integrity detection, with median localization errors of 0.034 m, 0.033 m, and 0.073 m for three damage levels, respectively. The proposed algorithm λ-FAdaMax outperforms traditional optimizers, demonstrating significant potential for engineering applications.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"68 ","pages":"Article 103690"},"PeriodicalIF":9.9,"publicationDate":"2025-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144721356","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A hybrid data-physics framework with conformal GNN for enhanced damage identification","authors":"Armin Dadras Eslamlou ,&nbsp;Arshia Ghasemlou ,&nbsp;Brais Barros ,&nbsp;Belén Riveiro","doi":"10.1016/j.aei.2025.103718","DOIUrl":"10.1016/j.aei.2025.103718","url":null,"abstract":"<div><div>Structural damage identification is crucial for ensuring safety, yet existing data-driven and physics-based methods often suffer from accuracy and computational limitations. To address these issues, we propose a hybrid framework that integrates Graph Neural Networks (GNNs) with a physics-based Finite Element (FE) model updating approach. The first module employs a GNN trained on modal data from FE simulations to estimate the location and severity of structural damage, with an evolutionary AutoML framework optimizing the GNN’s architecture and hyperparameters. In the second module, a conformal prediction technique quantifies uncertainty in the GNN’s predictions, ensuring robust confidence bounds for damage estimations. These uncertainty-aware predictions initialize a warm-started FE model updating workflow, where the Water Strider Algorithm (WSA) efficiently minimizes a cost function based on limited modal data. The proposed methodology has been validated on benchmark structures, including the Louisville bridge, IASC-ASCE building and a dome structure, demonstrating a remarkable increase in damage identification accuracy compared to conventional approaches. Unlike pure data-driven and physics-based methods, this hybrid framework leverages their strengths while integrating uncertainty quantification, enhancing their efficiency. This hybrid approach is scalable to various structural configurations, making it a promising solution for enhanced structural health monitoring.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"68 ","pages":"Article 103718"},"PeriodicalIF":9.9,"publicationDate":"2025-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144721236","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"3D reconstruction and optimization system for underwater dam surface based on modified SfM and neural radiation fields","authors":"Huadu Tang ,&nbsp;Fei Kang ,&nbsp;Zaiming Geng ,&nbsp;Zhe Li","doi":"10.1016/j.aei.2025.103683","DOIUrl":"10.1016/j.aei.2025.103683","url":null,"abstract":"<div><div>Regular inspections are essential for dam operation and maintenance. However, conventional underwater inspections, typically relying on manual visual assessments, suffer from low efficiency and limited spatial coverage, which hinder comprehensive detection of dams. To address these limitations, this paper proposes a novel three-dimensional (3D) reconstruction and optimization system using modified structure from motion and neural radiation fields. The rapid generation of 3D point clouds from underwater imagery was achieved through the combination of DISK-NN and COLLISION-MAPping (COLMAP). The SeaThru-NeRF is employed for accurate 3D reconstruction, integrated with multiresolution hash encoding and fully fused neural networks to expedite the process. Bayes’ rays are introduced for efficient artifact removal, facilitating optimization for enhanced 3D reconstruction quality. Experimental results demonstrate that the proposed method achieves a 3.31% improvement in peak signal to noise ratio (PSNR) and a 24.15% acceleration in reconstruction speed. Post-processing optimization of the 3D reconstruction can be effectively controlled by adjusting the uncertainty threshold. The system’s applicability and scalability are validated through multi scenes experiments in engineering projects. This study provides a novel technical pathway for the intelligent inspection of water conservancy facilities, demonstrating significant practical value.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"68 ","pages":"Article 103683"},"PeriodicalIF":9.9,"publicationDate":"2025-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144723629","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A knowledge graphs construction method enhanced by multimodal large language model for industrial equipment operation and maintenance","authors":"Zhengping Zhang ,&nbsp;Junyuan Yu ,&nbsp;Bo Yang ,&nbsp;Kaze Du ,&nbsp;Shilong Wang ,&nbsp;Xing Qi","doi":"10.1016/j.aei.2025.103705","DOIUrl":"10.1016/j.aei.2025.103705","url":null,"abstract":"&lt;div&gt;&lt;div&gt;The industrial equipment Operation and Maintenance (O&amp;M) is a core component in ensuring production safety and efficiency, urgently requiring the support of intelligent technologies. Knowledge graphs, which represent equipment and faults in graph structures, are widely utilized to enable efficient association and rapid retrieval of maintenance knowledge, thereby being extensively applied in intelligent decision-making for the equipment O&amp;M. Traditional knowledge graph construction methods, which rely on a single textual modality, are confronted with challenges such as the scarcity of annotated samples, difficulties in dynamically associating old and new equipment, and insufficient parsing of complex equipment relationships. As a result, issues like missing graph entities and broken causal chains are often encountered, thereby negatively impacting the quality of maintenance decision-making. Therefore, a dual-attention model enhanced by multimodal large language models (MllmDA-KGC) is proposed in this paper. Multimodal large language model(MLLM) is introduced to fully utilize multi-modal knowledge from the O&amp;M domain, thereby enabling a more effective understanding and modeling of complex O&amp;M knowledge. As a result, the quality of knowledge graph construction is significantly improved. In the MllmDA-KGC framework, first, QWEN2-VL is introduced into a dual-stream Transformer architecture to achieve dynamic alignment between images and text while supplementing semantics. As a result, the precision of identifying relationships between parts and problem entities in the O&amp;M domain is significantly enhanced; second, the MT-Transformer module is proposed, which integrates causal convolution, dilated convolution, and Memory_Bank mechanisms to achieve cross-modal temporal embedding fusion. As a result, the continuity of associations between new and old parts, as well as the precision of causal chain embeddings, is significantly improved; third, a multimodal dynamic weight attention-guided module is designed, in which weighted key-value guided attention mechanisms are introduced to focus on critical aspects. Schematic diagrams and textual features are fused to enhance the precision of entity relationship modeling between parts and faults; finally, to fully leverage the multimodal understanding capabilities of the MLLM, the image embeddings and positional embeddings generated and marked by MLLM are integrated into the feature embeddings of RoBERTa. Subsequently, CRF and Softmax are combined to accomplish the MNER (Multimodal Named Entity Recognition) and MRE (Multimodal Relation Extraction) tasks, thereby enabling the construction of a multimodal equipment O&amp;M knowledge graph. In this paper, the vehicle O&amp;M dataset from an automobile company was utilized to validate the proposed method. The experimental results showed that the F1 scores of the model in the MNER and MRE tasks reached 88.40% and 93.79%, respectively, demons","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"68 ","pages":"Article 103705"},"PeriodicalIF":9.9,"publicationDate":"2025-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144723628","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhancing PCB assemblies reliability: An adaptive CatBoost-MOA approach integrated with Hu-Washizu variational principle for fatigue life prediction","authors":"Akashjyoti Barman ,&nbsp;Konkimalla Venkata Sai Varun ,&nbsp;Sudha Radhika ,&nbsp;GR Sabareesh ,&nbsp;Veerabhadra Reddy","doi":"10.1016/j.aei.2025.103720","DOIUrl":"10.1016/j.aei.2025.103720","url":null,"abstract":"<div><div>Fatigue life prediction is crucial in ensuring the reliability and longevity of electronic components, particularly in high-stakes industries such as aerospace and defence, where premature failures can lead to catastrophic consequences, safety risks, and high operational costs. The current research introduces a novel adaptive model for predicting the fatigue lifetime of printed circuit board assemblies (PCBAs) under vibrational loading conditions. The model integrates adaptive versions of the Categorical Boosting (CatBoost) model and the Mother Optimization Algorithm (MOA) with the Hu-Washizu variational principle, addressing the limitations of conventional machine learning models by combining both data-driven and physics-informed components, improving accuracy and efficiency. The adaptive CatBoost-MOA-HuWashizu model significantly outperforms traditional methods, achieving a coefficient of determination (R²) of 0.9982 and a mean absolute percentage error (MAPE) of 0.0832 for single-component PCBs, and an R² of 0.9839 and MAPE of 0.0995 for multicomponent PCBs, demonstrating superior predictive capability for non-linear degradation behaviours under vibrational loading conditions. The integration of MOA optimises hyperparameters by balancing exploration and exploitation, reducing computational load and accelerating convergence, while the Hu-Washizu principle ensures efficient parameter tuning and resource allocation. Comparative results show reduced computational time, with faster convergence and no loss in prediction accuracy. The model’s robustness was validated through cross-validation, yielding an average R² of 0.9001 and an average MAPE of 0.3495 for single-component PCBs, and an average R² of 0.9084 and an average MAPE of 0.3466 for multicomponent PCBs, confirming its generalizability across varied operational scenarios. The proposed novel adaptive model achieves near-perfect alignment between predicted and numerically estimated fatigue lifetimes, with minimal residual errors. The research offers a promising solution for real-time fatigue life estimation in aerospace, defence, and related fields. Future research will expand the model’s scalability to other complex loading conditions and component types to enhance its applicability in broader reliability assessment frameworks.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"68 ","pages":"Article 103720"},"PeriodicalIF":9.9,"publicationDate":"2025-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144721235","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Distributed design optimisation in collaborative product development by integrating analytical target cascading with kriging","authors":"Kai-Wen Tien ,&nbsp;Chih-Hsing Chu","doi":"10.1016/j.aei.2025.103708","DOIUrl":"10.1016/j.aei.2025.103708","url":null,"abstract":"<div><div>In the current era of economic globalization, small and medium-sized enterprises have increasingly recognized the imperative of inter-company collaboration across the supply chain to enhance competitiveness. The effective utilization of distributed design resources has become crucial to address product complexity and shorter life cycles. Collaborative product development (CPD) has thus emerged as a common practice to achieve this goal, in which design teams, possibly dispersed across different companies, negotiate engineering solutions that not only fulfil the overall product development goal but also align with their own interests. This research proposes a novel approach by integrating Analytical Target Cascading (ATC) with Kriging to solve distributed optimal design problems in the context of CPD. The focus is to improve the iterative process of the ATC coordination strategy by incorporating the Kriging model to address the situation of limited information disclosure. Case studies of real-world engineering problems validate the effectiveness of the proposed approach. Test results show that it contributes not only to increasing the efficiency of the design process but also to improving the overall design quality in CPD.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"68 ","pages":"Article 103708"},"PeriodicalIF":8.0,"publicationDate":"2025-07-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144714504","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Process mechanisms fusion enhanced spatially scalable convolution network for multi-indicator prediction in process industries","authors":"Ming Wang ,&nbsp;Jie Zhang ,&nbsp;Youlong Lyu ,&nbsp;Peng Zhang ,&nbsp;Cheng Li","doi":"10.1016/j.aei.2025.103684","DOIUrl":"10.1016/j.aei.2025.103684","url":null,"abstract":"<div><div>In process industries, the detection of interrelated production indicators, including throughput, quality, etc., is often delay due to the inherent continuity, causing production disruptions and quality issues. Accurate prediction of multi-indicator is crucial, but intricate nonlinear correlations among parameters and indicators pose significant challenges. Data-driven prediction can overcome the challenge of accurately constructing process mechanism models covering the entire production process and all elements, but struggle to infer cross-space migration patterns of parameters’ impacts on indicators. To address this issue, this study proposes a process mechanism fusion enhanced intelligent multi-indicator prediction method for process industries, using the polyester fiber polymerization process as an illustrative case. Firstly, a process mechanism model is established to generate mechanism data encapsulating process mechanisms like coupled relationships and spatial correlations, and these mechanisms are extracted as mechanism features, which are fused with data features to enhance the model’s performance. Secondly, a spatially scalable convolutional neural network is raised, which extracts the implicit deep data features and mechanism features between parameters and indicators from both within-process and cross-process dimensions, utilizing both real and mechanism data. Furthermore, a multi-head self-attention mechanism is employed to adaptively adjust the self-attention weights of the fused features, guiding the model to learn the complex relationships between fused features and enhancing the ability to learn complex coupled relationships and spatial correlations. Finally, the proposed prediction method is validated using polymerization process data and demonstrated superior performance in achieving accurate multi-indicator prediction compared to both efficient machine learning and advanced deep learning methods.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"68 ","pages":"Article 103684"},"PeriodicalIF":8.0,"publicationDate":"2025-07-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144714505","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A safety risk formation model for intelligent human–machine interaction based on computational neuroscience","authors":"Yining Zeng ,&nbsp;Youchao Sun ,&nbsp;Xia Zhang ,&nbsp;Yuanyuan Guo ,&nbsp;Heming Wu","doi":"10.1016/j.aei.2025.103668","DOIUrl":"10.1016/j.aei.2025.103668","url":null,"abstract":"<div><div>The widespread application of intelligent technologies in cockpits enhances the efficiency of human–machine systems, introduces new challenges for safety risk analysis. Safety risk analysis focuses on explaining the safety risk formation process from a macro perspective. However, the neurocognitive mechanisms of pilots during this process are still unclear. This paper proposes a safety risk formation model for intelligent human–machine interaction based on computational neuroscience. Specifically, the safety risk factors affecting pilot behavior during intelligent human–machine interaction are identified. The regulatory mechanisms involving dopamine, acetylcholine, and the thalamus corresponding to different safety risk factors are elucidated. The structure and function of the Cortico-Basal Ganglia-Thalamus-Cortical (CBTC) neural circuit are introduced. To quantitatively describe the dynamic characteristics of neurons, a computational model of CBTC is established. Experimental validation of the proposed model is conducted using a cockpit intelligent interaction platform. The results indicate that the established computational model of CBTC effectively simulates the impact of varying levels of safety risk factors on pilots.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"68 ","pages":"Article 103668"},"PeriodicalIF":8.0,"publicationDate":"2025-07-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144714506","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}