Engineering Applications of Artificial Intelligence最新文献

{"title":"Overcoming challenges in leveraging blockchain technology: Entropy-based q-rung orthopair fuzzy model for benchmarking application barriers","authors":"Sana Shahab ,&nbsp;Naoufel Kraiem ,&nbsp;Ashit Kumar Dutta ,&nbsp;Mohd Anjum ,&nbsp;Vladimir Simic ,&nbsp;Dragan Pamucar","doi":"10.1016/j.engappai.2025.112433","DOIUrl":"10.1016/j.engappai.2025.112433","url":null,"abstract":"<div><div>Blockchain technology has emerged as a transformative solution across industries, delivering enhanced transparency, security, and operational efficiency. Nevertheless, its adoption remains hindered by significant challenges, especially in complex, data-intensive domains such as logistics. This study introduces a novel integration of the entropy-based q-rung orthopair fuzzy compromise ranking of alternatives from distance to ideal solution (CRADIS) approach to systematically evaluate and prioritize key barriers to blockchain adoption. The innovation of this work lies in applying q-rung orthopair fuzzy sets which are particularly capable of handling higher degrees of uncertainty and hesitancy, and then integrated with entropy for objective criterion weighting and CRADIS for robust decision-making. A real-world case study is presented, involving five critical barriers, lack of legal and regulatory frameworks, high implementation costs, technological scalability issues, data privacy and security concerns, and cultural resistance to change evaluated against eight decision criteria. The entropy weighting revealed regulatory clarity (0.168) and security (0.154) as the most influential factors, while the CRADIS ranking identified a lack of legal frameworks as the top barrier. This framework provides a transparent, data-driven method for decision-makers to identify and prioritize adoption challenges, particularly in uncertain and multi-faceted environments. By demonstrating the model’s applicability and precision, the study contributes to the emerging body of literature on blockchain integration and supports organizations in navigating the transition towards decentralized technologies.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112433"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159636","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Adaptive detection method for driver fatigue using facial multisource dynamic behavior fusion","authors":"Guoxin Zhang ,&nbsp;Fei Yang ,&nbsp;Xin Fang ,&nbsp;Lili Wang ,&nbsp;Lei Zhao ,&nbsp;Chaoning Yu","doi":"10.1016/j.engappai.2025.112482","DOIUrl":"10.1016/j.engappai.2025.112482","url":null,"abstract":"<div><div>Driving while fatigued is a leading cause of traffic accidents. This study proposed an adaptive detection model to recognize driver fatigue based on the dynamic facial behavior information of drivers. First, drivers’ facial fatigue features were extracted to establish a general feature space, including pupil movement, eye state, and fatigue expression parameters. A differentiated feature space was then built based on individual drivers, taking into account the homogeneity, regularity, and individual variances in drivers' facial behavior at various states. A complete adaptive fatigue feature space was built by integrating the general feature space and differentiated feature space. Finally, a driver adaptive fatigue discrimination model was constructed to classify the general and adaptive fatigue feature space to detect driver fatigue states adaptively. A driver fatigue detection dataset from real scenarios had been established to validate the performance of the proposed model. Experimental results demonstrated that the proposed method significantly improved the detection accuracy of driver fatigue. In terms of artificial intelligence, this study contributes a novel adaptive feature space construction method based on multimodal dynamic feature fusion for facial fatigue recognition; in engineering application, it develops an adaptive driver fatigue detection system grounded in multimodal dynamic behaviors, which provides real-time alerts upon detecting driver fatigue and ensures driving safety.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112482"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159652","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Dual angle magnetic arc blow estimation in keyhole tungsten inert gas welding using high dynamic range imaging and a lightweight vision transformer network with coordinate attention and multiple auxiliary branches","authors":"Xiyin Chen ,&nbsp;Xiaohu Zhang ,&nbsp;Yonghua Shi ,&nbsp;Yuxiang Huang ,&nbsp;Junjie Pang","doi":"10.1016/j.engappai.2025.112432","DOIUrl":"10.1016/j.engappai.2025.112432","url":null,"abstract":"<div><div>In high current Keyhole Tungsten Inert Gas (K-TIG) welding, magnetic arc blow frequently causes severe defects such as lack of fusion and undercut, which seriously affect weld formation quality. Conventional visual sensing systems are limited by dynamic range, making it difficult to capture arc morphology, while single angle descriptors fail to represent nonlinear deflection and lightweight convolutional models struggle with long range dependencies. To address these challenges, this study employs a High Dynamic Range (HDR, 120 decibel [dB]) imaging system to capture detailed arc variations and proposes a lightweight Vision Transformer (ViT) network with embedded Coordinate Attention (CA) and multiple auxiliary branches for real time angle estimation. A custom magnetic excitation system enables controllable arc blow simulation and consistent data acquisition. The method introduces a dual angle representation, namely the maximum curvature angle (<span><math><msub><mrow><mi>θ</mi></mrow><mrow><mi>curv</mi></mrow></msub></math></span>) and the equivalent deviation angle (<span><math><msub><mrow><mi>θ</mi></mrow><mrow><mi>eq</mi></mrow></msub></math></span>), to comprehensively describe arc geometry. The Artificial Intelligence (AI) framework integrates segmentation, keypoint localization, and regression tasks to improve accuracy and robustness. Trained on a self constructed HDR dataset containing 3,191 annotated images, model achieves a mean absolute error (MAE) of <span><math><mrow><mn>1</mn><mo>.</mo><mn>12</mn><mo>°</mo></mrow></math></span>, a root mean square error (RMSE) of <span><math><mrow><mn>2</mn><mo>.</mo><mn>84</mn><mo>°</mo></mrow></math></span>, a determination coefficient (<span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span>) of 0.96, and a per frame inference latency of 12.96 ms (ms) on an NVIDIA RTX 2080Ti graphics processing unit (GPU). These results demonstrate that AI based methods combined with HDR imaging cannot only achieve accurate monitoring of welding arc states, but also provide potential support for closed loop control in all position welding applications.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112432"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159775","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Interpretable chest X-ray localization using principal component-based feature selection in deep learning","authors":"Diwakar Diwakar ,&nbsp;Deepa Raj","doi":"10.1016/j.engappai.2025.112358","DOIUrl":"10.1016/j.engappai.2025.112358","url":null,"abstract":"<div><div>Accurate identification and localization of diseases in chest X-ray (CXR) images are crucial for early diagnosis and timely medical intervention. Traditional localization techniques like Class Activation Mapping (CAM), depend on Global Average Pooling (GAP) layers, restricting their flexibility, while gradient-based methods like Grad-CAM involve computational overhead and limited interpretability. To address these limitations, this study introduces a novel Principal Component Analysis (PCA)-based localization method that eliminates reliance on GAP layers and gradient computations. Utilizing publicly available Kaggle datasets, namely the COVID-19 Radiography Dataset and Tuberculosis (TB) Chest X-ray Database. The proposed approach employs PCA to compress high-dimensional convolutional feature maps extracted from the pretrained VGG16 model into a lower-dimensional, spatially meaningful representation. This enables rapid, interpretable heatmap generation highlighting precise abnormal regions. Experimental results demonstrate that the proposed method achieved an average training loss of <span><math><mrow><mn>0</mn><mo>.</mo><mn>0835</mn><mo>±</mo><mn>0</mn><mo>.</mo><mn>1830</mn></mrow></math></span> and validation loss of <span><math><mrow><mn>0</mn><mo>.</mo><mn>1385</mn><mo>±</mo><mn>0</mn><mo>.</mo><mn>0741</mn></mrow></math></span> across 5-fold cross-validation. In addition, it achieved an impressive accuracy of 97.5%, sensitivity of 98.2%, specificity of 99.4%, a Dice Similarity Coefficient (DSC) of 97.5%, and an Intersection-over-Union (IoU) of 95.1%. Compared to CAM, and Grad-CAM, PCA-based localization significantly reduces inference time, enhances interpretability, and provides robust multi-class localization performance suitable for clinical deployment.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112358"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222591","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Carbon trading price prediction with spikes: A novel hybrid model framework using heuristic multi-head attention convolutional bidirectional recurrent neural network","authors":"Rongquan Zhang ,&nbsp;Siqi Bu ,&nbsp;Gangqiang Li ,&nbsp;Min Zhou","doi":"10.1016/j.engappai.2025.112438","DOIUrl":"10.1016/j.engappai.2025.112438","url":null,"abstract":"<div><div>Accurate forecasting of carbon trading prices (CTPs) with spikes is crucial for developing carbon emission reduction policies and planning corporate investments. However, most existing CTP approaches usually focus on designing a cutting-edge model without considering spike prediction. Therefore, this paper presents a novel heuristic optimization-based hybrid model framework for CTP prediction with spikes. First, random forest is exploited to identify the relevant features of spikes and non-spikes for CTPs, and categorical boosting is employed to predict the spike occurrences of CTPs. Then, a novel hybrid model based on multiple linear regression, categorical boosting, and two dimensions convolutional neural network and bidirectional gated recurrent unit with multi-head regularized attention mechanism (2DCNN-BiGRU-MRA) is proposed to predict spikes and non-spikes for CTPs. In this model, multiple linear regression and categorical boosting are respectively applied to capture the linear and complex nonlinear features of the CTPs, in which their prediction results and deviations are integrated into the 2DCNN-BiGRU-MRA model as relevant features. The proposed 2DCNN-BiGRU-MRA can learn the spatiotemporal features and enhance representation capabilities by introducing 2DCNN, BiGRU, and MRA, thereby improving the accuracy of CTP prediction. In addition, to construct appropriate model hyperparameters of 2DCNN-BiGRU-MRA, the strength honey badger algorithm based on the adaptive momentum estimation is proposed to optimize the hyperparameters of 2DCNN-BiGRU-MRA. Finally, the proposed framework is tested on the actual data of European Union emissions trading and the carbon market in Hubei, China, and case studies have confirmed the superiority and achievable local interpretability of the proposed hybrid model framework.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112438"},"PeriodicalIF":8.0,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145134846","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-dimensional feature fusion network design and performance optimisation for small target detection","authors":"Xiaoyao Yang ,&nbsp;Wenyang Zhao ,&nbsp;Pengchao Sun ,&nbsp;Wenda Zhao ,&nbsp;Wenlong Yang","doi":"10.1016/j.engappai.2025.112425","DOIUrl":"10.1016/j.engappai.2025.112425","url":null,"abstract":"<div><div>Due to the long distance of image acquisition, high imaging resolution, complex feature background, shooting angle, etc. The result is that there are few features available for small targets and they are easily interfered by background noise, which poses a challenge to the detection of small targets. To address the above problems, this paper proposes a target detection network (Convolution-based Small Target Detection Network, CSTDNet) with enhanced feature information, which integrates a multi-dimensional information fusion strategy for small target features. An all-round efficient feature fusion mudule (AeFusion) is introduced, which emphasises the fusion of multi-dimensional feature information, enhances the model's ability to focus on key information and suppress redundant information, and strengthens the ability to characterise local features and details, improving the effectiveness of the information and computational efficiency. In order to further enhance the location-awareness capability in cross-layer interaction, this paper introduces a novel decoupling head (Self-aware task decomposition for fine-grained feature sharing, STFS), which improves the accuracy of the small-target classification and localisation tasks through efficient detail sharing and task auto-alignment functions. And localisation tasks through efficient detail sharing and task auto-alignment. This study evaluates the effectiveness of the algorithm on five different scenarios containing small target datasets. Experimental results show that CSTDNet achieved improvements of 6.6 %, 5.8 %, 5.8 %, 5.5 %, and 5.6 % over the baseline model in terms of the mean average precision ([email protected]) metric on the Visdrone 2019, BDD100K, WiderPerson, SODA10M, and AppleDatas datasets, respectively, demonstrating stronger detection performance.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112425"},"PeriodicalIF":8.0,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159654","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A novel federated deep learning for intrusion detection in smart grid cyber-physical systems","authors":"Rong Xie ,&nbsp;Bin Wang ,&nbsp;Xin Xu","doi":"10.1016/j.engappai.2025.112404","DOIUrl":"10.1016/j.engappai.2025.112404","url":null,"abstract":"<div><div>The fusion of sophisticated computational, communicative, and physical elements in Smart Grid Cyber-Physical Systems (SGCPS) has greatly improved the efficiency and reliability of power grids. However, this complexity introduces enhanced cybersecurity risks, evidenced by significant cyberattacks on the Ukrainian power grid during 2015 and 2016. Despite progress in Artificial Intelligence (AI)-driven security solutions for SGCPS, practical deployment of these technologies is often limited due to a lack of high-quality attack data and owners’ hesitance to distribute sensitive details. This paper introduces an innovative strategy to fortify SGCPS against diverse network threats via a comprehensive intrusion detection system. We present a deep learning model leveraging a temporal convolutional network with multi-feature integration, aimed at robust threat identification. We also propose a federated learning framework enabling various SGCPS to jointly develop an extensive intrusion detection model, ensuring data privacy. Moreover, we incorporate a gradient compression technique utilizing the Long Short Term Memory-<span><math><mi>β</mi></math></span>-Total Correlation Variational Autoencoder (LSTM-<span><math><mi>β</mi></math></span>-TCVAE) model to enhance and secure model parameters throughout the training phase. Thorough experimental validations confirm the efficacy of our method in recognizing multiple cyber threat types to SGCPS and its advantages over current methods.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112404"},"PeriodicalIF":8.0,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145134851","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Machine learning methods comparison by using statistical tests in solar energy forecasting based on weather features","authors":"Mohammadreza pourmir ,&nbsp;Seyedeh Mohadeseh Miri","doi":"10.1016/j.engappai.2025.112239","DOIUrl":"10.1016/j.engappai.2025.112239","url":null,"abstract":"<div><div>Climate change necessitates precise solar forecasting due to its weather-dependent intermittency. Key parameters - temperature, visibility, altitude, pressure, and wind speed - were analyzed using non-parametric tests. We prioritized short-term weather patterns over random data splitting for enhanced accuracy.Non-parametric tests, such as the Kolmogorov-Smirnov test, were used to assess data normality and select highly correlated features. Principal Component Analysis (PCA) reduces dataset dimensionality while preserving critical trends. Various machine learning approaches were evaluated, including: weighted linear regression (both with and without dimensionality reduction), boosted regression trees, and deep learning architectures-comprising both fundamental models (Convolutional Neural Networks [CNNs] and Recurrent Neural Networks [RNNs]) and advanced hybrid architectures (Temporal Convolutional Networks (TCN) Convolutional Neural Network-Long Short-Term Memory network (CNN-LSTM). All models were optimized through systematic hyperparameter tuning to enhance predictive performance, reduce computational complexity, and improve learning convergence rates. Special attention was given to addressing vanishing gradient problems in deep neural network implementations. Results show TCN outperform other deep learning models, achieving lower training and testing errors with fewer parameters and reduced time complexity. CNN-LSTM models, designed for spatial-sequence prediction, perform well but require more parameters and computational time. The lowest test and training errors belong to CNN-LSTM and TCN, with approximately 9 % and 2 % lower than the maximum amount, respectively. A trade-off between model complexity, error rates, and computational efficiency must be considered when selecting the optimal approach. Since relevant weather features vary by location, the proposed methodology serves as an adaptable algorithm for solar energy prediction in diverse geographical regions.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112239"},"PeriodicalIF":8.0,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159656","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A multi-scale feature extraction and fusion framework based on wavelet Kolmogorov–Arnold networks and parallel Bi-directional gated recurrent units for electric load forecasting","authors":"Chunliang Mai ,&nbsp;Lixin Zhang ,&nbsp;Xuewei Chao ,&nbsp;Xue Hu ,&nbsp;Omar Behar","doi":"10.1016/j.engappai.2025.112517","DOIUrl":"10.1016/j.engappai.2025.112517","url":null,"abstract":"<div><div>Short-term electric load forecasting remains challenged by the dual requirements of accuracy and robustness due to the combined effects of strong seasonality, multi-scale spikes, and stochastic disturbances. To address this, we propose a novel multi-scale forecasting framework, NP-WavKAN-Fusion, which integrates Neural Prophet for data decomposition and a Wavelet-based Kolmogorov–Arnold Network (WavKAN) with learnable wavelet kernels for multi-scale encoding. This fusion model utilizes a Bi-directional Gated Recurrent Unit (BiGRU) to capture long-term temporal dependencies and an adaptive feature fusion gate (AFF) to dynamically re-weight static and dynamic features for final load predictions. Extensive experiments on two public datasets from Australia and Morocco show that NP-WavKAN-Fusion consistently outperforms traditional models, reducing the mean absolute error by at least 30 %. For multi-step forecasting tasks, NP-WavKAN-Fusion maintains error inflation within 15 %, demonstrating superior performance compared to state-of-the-art long-sequence models such as Informer and PatchTST. The Diebold–Mariano test confirms that NP-WavKAN-Fusion yields statistically significant improvements, with 19 out of 20 comparisons showing lower errors. Ablation studies show that removing either the Neural Prophet component or the AFF significantly increases the forecasting error, validating the necessity of our layered denoising and fusion strategies. The proposed NP-WavKAN-Fusion framework demonstrates strong potential for real-world applications in electric load forecasting, offering robust performance under various temporal and non-stationary conditions.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112517"},"PeriodicalIF":8.0,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145134845","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A reliable degradation prediction method for proton exchange membrane fuel cells based on uncertainty Bayesian self-attention","authors":"Mengyu Liu ,&nbsp;Zhe Cheng ,&nbsp;Yu Yang ,&nbsp;Niaoqing Hu ,&nbsp;Guoji Shen ,&nbsp;Yi Yang","doi":"10.1016/j.engappai.2025.112444","DOIUrl":"10.1016/j.engappai.2025.112444","url":null,"abstract":"<div><div>Health state prediction of Proton Exchange Membrane Fuel Cells (PEMFCs) is a critical technology to ensure their long-term reliable operation. Prediction accuracy directly influences the effectiveness of maintenance strategies and risk management. However, existing PEMFC degradation prediction methods based on Recurrent Neural Networks (RNNs) or Transformer architectures mostly focus on point estimation while neglecting uncertainty quantification. This limitation makes it difficult to assess the confidence level of predictions in practical engineering applications, reducing the models' reliability in decision support. To address this issue, this paper proposes a novel Bayesian Patch Time Series Transformer (B-PatchTST) method. By deeply integrating Bayesian variational inference with time series patch modeling, the method enables probabilistic prediction of PEMFC degradation trajectories and disentangled analysis of uncertainty sources. Unlike traditional Bayesian Neural Networks (BNNs) that primarily apply Bayesian modeling to fully connected layers, B-PatchTST introduces a Bayesian Self-Attention Mechanism, which models epistemic uncertainty in three stages: patch embedding, uncertainty-aware self-attention computation, and adaptive regularization. This design significantly enhances the credibility of the model. Extensive experiments on the fuel cell datasets demonstrate the proposed method's outstanding performance. It achieves an average reduction of 36.31 % in root mean square error and an average compression of 83.39 % in the 95 % confidence interval, significantly outperforming existing methods. This approach offers a trustworthy basis for predictive maintenance in PEMFC systems, promoting a shift from “experience-based maintenance” to “reliable prognostics” in hydrogen energy applications.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112444"},"PeriodicalIF":8.0,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145134847","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}