Advanced Engineering Informatics最新文献_第5页

Explainable wavelet-scalogram learning for quasi-stationary faults in automotive DC motors using AERIS-Wave 基于AERIS-Wave的汽车直流电机准平稳故障的可解释小波尺度学习

IF 9.9 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2026-02-03 DOI: 10.1016/j.aei.2026.104394

Willy Dharmawan , Nur Hamid , Haitham Saleh , Amalia Irma Nurwidya , Peni Laksmita Widati

Noise diagnostics in automotive brushed DC motors are crucial for ensuring reliability and compliance with strict manufacturing standards. Traditional human-based inspection methods are often subjective and inconsistent, necessitating automated, data-driven solutions. This study introduces an explainable wavelet-based classification framework for diagnosing quasi-stationary motor faults using accelerometer data. The framework employs the Continuous Wavelet Transform (CWT) to extract rich time–frequency features, including scalograms, ridge trajectories, periodograms, and statistical descriptors, serving as discriminative representations of mechanical behavior. These features are evaluated using both classical machine learning algorithms (Random Forest, XGBoost) and deep learning architectures such as ResNet18, VGG16, CNN-LSTM, CNN-GRU, and WaveNet. Building upon these baselines, the proposed AERIS-Wave (Attention-Enhanced Residual Interpretable Scalogram Network) integrates multi-level attention, LayerScale normalization, and explainable-AI components (Integrated Gradients, Grad-CAM, SHAP) to visualize spectral contributions that drive decisions. Experimental results show that AERIS-Wave achieves 97.72% accuracy and an AUC of 0.9991, surpassing all benchmark models, including WaveNet and ResNet18. The findings confirm that wavelet-based representations, combined with interpretable deep learning, enable high-precision, explainable, and scalable fault classification suitable for real-time quality control in industrial environments.

汽车有刷直流电机的噪声诊断对于确保可靠性和符合严格的制造标准至关重要。传统的人工检测方法往往是主观的和不一致的，需要自动化的、数据驱动的解决方案。本文介绍了一种基于小波的可解释分类框架，用于利用加速度计数据诊断准静止电机故障。该框架采用连续小波变换（CWT）提取丰富的时频特征，包括尺度图、脊轨迹、周期图和统计描述符，作为机械行为的判别表示。这些特征使用经典机器学习算法（Random Forest, XGBoost）和深度学习架构（如ResNet18， VGG16, CNN-LSTM， CNN-GRU和WaveNet）进行评估。在这些基线的基础上，提出的AERIS-Wave（注意力增强残余可解释尺度网络）集成了多层次注意力、LayerScale归一化和可解释的人工智能组件（集成梯度、gradcam、SHAP），以可视化驱动决策的光谱贡献。实验结果表明，AERIS-Wave的准确率为97.72%，AUC为0.9991，优于WaveNet和ResNet18等所有基准模型。研究结果证实，基于小波的表示与可解释的深度学习相结合，可以实现高精度、可解释和可扩展的故障分类，适用于工业环境中的实时质量控制。

{"title":"Explainable wavelet-scalogram learning for quasi-stationary faults in automotive DC motors using AERIS-Wave","authors":"Willy Dharmawan , Nur Hamid , Haitham Saleh , Amalia Irma Nurwidya , Peni Laksmita Widati","doi":"10.1016/j.aei.2026.104394","DOIUrl":"10.1016/j.aei.2026.104394","url":null,"abstract":"<div><div>Noise diagnostics in automotive brushed DC motors are crucial for ensuring reliability and compliance with strict manufacturing standards. Traditional human-based inspection methods are often subjective and inconsistent, necessitating automated, data-driven solutions. This study introduces an explainable wavelet-based classification framework for diagnosing quasi-stationary motor faults using accelerometer data. The framework employs the Continuous Wavelet Transform (CWT) to extract rich time–frequency features, including scalograms, ridge trajectories, periodograms, and statistical descriptors, serving as discriminative representations of mechanical behavior. These features are evaluated using both classical machine learning algorithms (Random Forest, XGBoost) and deep learning architectures such as ResNet18, VGG16, CNN-LSTM, CNN-GRU, and WaveNet. Building upon these baselines, the proposed AERIS-Wave (Attention-Enhanced Residual Interpretable Scalogram Network) integrates multi-level attention, LayerScale normalization, and explainable-AI components (Integrated Gradients, Grad-CAM, SHAP) to visualize spectral contributions that drive decisions. Experimental results show that AERIS-Wave achieves 97.72% accuracy and an AUC of 0.9991, surpassing all benchmark models, including WaveNet and ResNet18. The findings confirm that wavelet-based representations, combined with interpretable deep learning, enable high-precision, explainable, and scalable fault classification suitable for real-time quality control in industrial environments.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"71 ","pages":"Article 104394"},"PeriodicalIF":9.9,"publicationDate":"2026-02-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146188699","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}

引用次数: 0

A dual-condition diffusion-based microseismic signals denoiser for real-world engineering noise 基于双条件扩散的微震信号去噪方法

IF 9.9 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2026-02-03 DOI: 10.1016/j.aei.2026.104374

Haoran Xu, Shibin Tang, Yanjun Mao

Microseismic monitoring represents an information-rich sensing modality in rock mass engineering, which is crucial for assessing rock mass stability during construction. However, the complex non-stationary noise generated by construction and human activities limits its analytical value, resulting in degraded performance in downstream inference within automated monitoring pipelines. To address this challenge, the study innovatively proposes a diffusion-based microseismic signal denoiser (DMSD). The main contributions of this work are as follows: (1) This study encodes the noise distribution and characteristics as a condition into the diffusion model, forming a dual-condition denoising diffusion model conditioned on noise and diffusion time step, which effectively preserves the time–frequency characteristics of microseismic signals while precisely suppressing noise. (2) A ten-million-scale noise-microseismic dataset encompassing diverse engineering noise types was constructed. Experimental results on the dataset indicate that DMSD outperforms existing baseline denoising methods across multiple evaluation metrics, achieving an average signal-to-noise ratio (SNR) improvement of 41.7% and enhancing arrival picking accuracy (error < 1 ms) by 30.1% compared to the denoising baselines. (3) Furthermore, this study reveals new engineering knowledge by analyzing the adaptability of denoising methods to varying noise conditions, thereby providing actionable insights for method selection in practice. The proposed denoiser thereby enhances the reliability of downstream microseismic analysis and facilitates more robust and interpretable engineering informatics systems.

微震监测是岩体工程中一种信息丰富的传感方式，对工程施工过程中岩体稳定性评估具有重要意义。然而，由于施工和人类活动产生的复杂非平稳噪声限制了其分析价值，导致自动化监测管道下游推理性能下降。为了应对这一挑战，该研究创新性地提出了一种基于扩散的微震信号去噪器（DMSD）。本工作的主要贡献如下：(1)本研究将噪声的分布和特征作为条件编码到扩散模型中，形成了以噪声和扩散时间步长为条件的双条件去噪扩散模型，在精确抑制噪声的同时有效地保留了微震信号的时频特征。(2)构建了包含多种工程噪声类型的千万尺度噪声-微地震数据集。数据集上的实验结果表明，DMSD在多个评估指标上优于现有的基线去噪方法，与去噪基线相比，平均信噪比（SNR）提高41.7%，到达拾取精度（误差<； 1 ms）提高30.1%。(3)此外，通过分析降噪方法对不同噪声条件的适应性，揭示了新的工程知识，从而为实践中的方法选择提供了可操作的见解。因此，所提出的去噪方法提高了下游微震分析的可靠性，并促进了更健壮和可解释的工程信息系统。

{"title":"A dual-condition diffusion-based microseismic signals denoiser for real-world engineering noise","authors":"Haoran Xu, Shibin Tang, Yanjun Mao","doi":"10.1016/j.aei.2026.104374","DOIUrl":"10.1016/j.aei.2026.104374","url":null,"abstract":"<div><div>Microseismic monitoring represents an information-rich sensing modality in rock mass engineering, which is crucial for assessing rock mass stability during construction. However, the complex non-stationary noise generated by construction and human activities limits its analytical value, resulting in degraded performance in downstream inference within automated monitoring pipelines. To address this challenge, the study innovatively proposes a diffusion-based microseismic signal denoiser (DMSD). The main contributions of this work are as follows: (1) This study encodes the noise distribution and characteristics as a condition into the diffusion model, forming a dual-condition denoising diffusion model conditioned on noise and diffusion time step, which effectively preserves the time–frequency characteristics of microseismic signals while precisely suppressing noise. (2) A ten-million-scale noise-microseismic dataset encompassing diverse engineering noise types was constructed. Experimental results on the dataset indicate that DMSD outperforms existing baseline denoising methods across multiple evaluation metrics, achieving an average signal-to-noise ratio (SNR) improvement of 41.7% and enhancing arrival picking accuracy (error < 1 ms) by 30.1% compared to the denoising baselines. (3) Furthermore, this study reveals new engineering knowledge by analyzing the adaptability of denoising methods to varying noise conditions, thereby providing actionable insights for method selection in practice. The proposed denoiser thereby enhances the reliability of downstream microseismic analysis and facilitates more robust and interpretable engineering informatics systems.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"71 ","pages":"Article 104374"},"PeriodicalIF":9.9,"publicationDate":"2026-02-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146188612","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}

引用次数: 0

A DEM-based particle–force chain informatics framework for data-driven evaluation of pavement pre-compaction 基于dem的路面预压实数据驱动评价粒子力链信息框架

IF 9.9 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2026-02-02 DOI: 10.1016/j.aei.2026.104402

Dong Feng

Pre-compaction beneath the paver screed strongly influences the mechanical response and durability of pavements, yet it cannot be measured directly in the field. This study develops a discrete element method (DEM)-based particle–force chain informatics framework for simulating screed device–particle interaction and for deriving interpretable, data-driven indicators of pavement pre-compaction quality. A Johnson–Kendall–Roberts (JKR)-based DEM model is formulated for a vibrating screed device acting on asphalt mixtures and is used to provide particle-to-particle contact and force chain descriptors. These heterogeneous descriptors are standardized by a QuantileTransformer (QT) and embedded into a low-dimensional latent space using a denoising variational autoencoder (DVAE), from which a Gaussian mixture model (GMM) yields data-driven labels and composite pre-compaction scores. Gradient-boosted decision trees (XGBoost) are then trained to predict these scores from the original descriptors, and Shapley-additive (SHAP) explanations quantify the contributions and interactions of individual micro-mechanical features. Numerical experiments on SMA-13, AC-13 and OGFC-13 mixtures, each with three gradation variants, show that the framework consistently captures the strengthening, elongation and orientation of load-bearing force chains, the reduction of weak chains, and the concentration of high stress particles as pre-compaction progresses. The resulting indicator set, together with the trained surrogate model, constitutes a reproducible and computationally efficient informatics route to assess, rank and compare pre-compaction conditions across mixtures and operating parameters, and demonstrates how DEM simulations can be systematically transformed into predictive, interpretable metrics for data-driven evaluation of pavement pre-compaction.

铺装层下的预压实对路面的力学响应和耐久性有很大影响，但无法在现场直接测量。本研究开发了一个基于离散元方法（DEM）的颗粒-力链信息框架，用于模拟找平设备-颗粒的相互作用，并用于推导可解释的、数据驱动的路面预压实质量指标。基于Johnson-Kendall-Roberts （JKR）的DEM模型是为作用于沥青混合料的振动筛设备制定的，用于提供颗粒间的接触和力链描述符。这些异构描述符由QuantileTransformer （QT）标准化，并使用去噪变分自编码器（DVAE）嵌入到低维潜在空间中，高斯混合模型（GMM）从中产生数据驱动的标签和复合预压缩分数。然后训练梯度增强决策树（XGBoost）来从原始描述符中预测这些分数，Shapley-additive （SHAP）解释量化单个微观力学特征的贡献和相互作用。对SMA-13、AC-13和OGFC-13混合料的数值实验表明，随着预压实的进行，框架一致地捕捉了承重力链的强化、延伸和取向，弱链的减少以及高应力颗粒的集中。由此产生的指标集与经过训练的代理模型一起，构成了可重复且计算效率高的信息学路线，用于评估、排序和比较混合料和操作参数的预压实条件，并展示了如何将DEM模拟系统地转化为可预测的、可解释的指标，用于数据驱动的路面预压实评估。

{"title":"A DEM-based particle–force chain informatics framework for data-driven evaluation of pavement pre-compaction","authors":"Dong Feng","doi":"10.1016/j.aei.2026.104402","DOIUrl":"10.1016/j.aei.2026.104402","url":null,"abstract":"<div><div>Pre-compaction beneath the paver screed strongly influences the mechanical response and durability of pavements, yet it cannot be measured directly in the field. This study develops a discrete element method (DEM)-based particle–force chain informatics framework for simulating screed device–particle interaction and for deriving interpretable, data-driven indicators of pavement pre-compaction quality. A Johnson–Kendall–Roberts (JKR)-based DEM model is formulated for a vibrating screed device acting on asphalt mixtures and is used to provide particle-to-particle contact and force chain descriptors. These heterogeneous descriptors are standardized by a QuantileTransformer (QT) and embedded into a low-dimensional latent space using a denoising variational autoencoder (DVAE), from which a Gaussian mixture model (GMM) yields data-driven labels and composite pre-compaction scores. Gradient-boosted decision trees (XGBoost) are then trained to predict these scores from the original descriptors, and Shapley-additive (SHAP) explanations quantify the contributions and interactions of individual micro-mechanical features. Numerical experiments on SMA-13, AC-13 and OGFC-13 mixtures, each with three gradation variants, show that the framework consistently captures the strengthening, elongation and orientation of load-bearing force chains, the reduction of weak chains, and the concentration of high stress particles as pre-compaction progresses. The resulting indicator set, together with the trained surrogate model, constitutes a reproducible and computationally efficient informatics route to assess, rank and compare pre-compaction conditions across mixtures and operating parameters, and demonstrates how DEM simulations can be systematically transformed into predictive, interpretable metrics for data-driven evaluation of pavement pre-compaction.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"71 ","pages":"Article 104402"},"PeriodicalIF":9.9,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146188610","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}

引用次数: 0

Dynamic distance trajectory correlation for associating unsafe behavior with worker identity: A multimodal CV–RFID integration approach 将不安全行为与工人身份相关联的动态距离轨迹关联：一种多模式CV-RFID集成方法

IF 9.9 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2026-02-02 DOI: 10.1016/j.aei.2026.104380

Ruying Cai , Ying Liang , Xiangsheng Chen , Jingru Li , Jingyuan Tang , Yi Tan

Convolutional neural networks (CNNs) have been widely adopted in construction safety for detecting unsafe behaviors. However, accurately associating unsafe behaviors with the identities of specific workers remains a significant challenge. This paper proposes a trajectory-consistent multimodal identity association approach integrating computer vision (CV) and radio frequency identification (RFID) technologies. To address the instability and unreliability of RSSI signals caused by environmental interference, signal loss, and multipath effects, this paper introduces a novel dynamic trajectory correlation-based matching algorithm, which shifts the focus from instantaneous distance values to temporal similarity of distance trends. This matching is performed only when both CV- and RFID-derived distances co-exist and exhibit consistent dynamic patterns, ensuring robustness against signal fluctuation and partial occlusions. The proposed method comprises: (1) a novel unsafe behavior identification method based on the PPE-based overlap rate between bounding boxes of workers and their personal protective equipment (PPE), where object detection is performed using YOLO-based models (YOLOv5, YOLOv8–YOLOv12); (2) visual tracking of individuals and estimation of left-wrist tag positions via human pose estimation; (3) dual-modality distance estimation using spatial coordinate transformation with spatial adjustment for CV and a path-loss model for RFID; (4) the trajectory-based identity alignment algorithm that leverages cross-modal dynamic consistency. Experimental evaluations demonstrate that YOLOv8 achieves the best object detection accuracy (mAP 0.969, F1-score 0.950) of the test dataset, YOLOv9 yields the highest unsafe behavior detection accuracy (0.8663) of the actual video data. Notably, the proposed trajectory-based identity association method reaches 1.0000 accuracy in stable tracking segments. These results confirm the method’s potential to accurately determine “who is unsafe” in real-world construction environments.

卷积神经网络（Convolutional neural networks, cnn）被广泛应用于建筑安全中，用于检测不安全行为。然而，准确地将不安全行为与特定工人的身份联系起来仍然是一个重大挑战。本文提出了一种结合计算机视觉（CV）和射频识别（RFID）技术的轨迹一致的多模态身份关联方法。针对RSSI信号受环境干扰、信号丢失、多径效应等因素影响而产生的不稳定性和不可靠性问题，提出了一种基于动态轨迹相关的匹配算法，该算法将焦点从瞬时距离值转移到距离趋势的时间相似性上。只有当CV和rfid衍生的距离共存并表现出一致的动态模式时，才会进行匹配，以确保对信号波动和部分遮挡的鲁棒性。提出的方法包括：(1)一种基于PPE的工人与个人防护装备（PPE）边界盒重叠率的不安全行为识别新方法，其中使用基于YOLOv5、YOLOv8-YOLOv12的模型进行目标检测；(2)对个体进行视觉跟踪，利用人体姿态估计左手腕标签位置；(3)基于CV空间坐标变换和RFID路径损耗模型的双模距离估计；(4)利用跨模态动态一致性的轨迹同一性对齐算法。实验评价表明，YOLOv8达到了测试数据集的最佳目标检测精度（mAP 0.969, F1-score 0.950）， YOLOv9达到了实际视频数据的最高不安全行为检测精度（0.8663）。值得注意的是，所提出的基于轨迹的身份关联方法在稳定跟踪段中达到了1.0000的精度。这些结果证实了该方法在真实建筑环境中准确判断“谁是不安全的”的潜力。

{"title":"Dynamic distance trajectory correlation for associating unsafe behavior with worker identity: A multimodal CV–RFID integration approach","authors":"Ruying Cai , Ying Liang , Xiangsheng Chen , Jingru Li , Jingyuan Tang , Yi Tan","doi":"10.1016/j.aei.2026.104380","DOIUrl":"10.1016/j.aei.2026.104380","url":null,"abstract":"<div><div>Convolutional neural networks (CNNs) have been widely adopted in construction safety for detecting unsafe behaviors. However, accurately associating unsafe behaviors with the identities of specific workers remains a significant challenge. This paper proposes a trajectory-consistent multimodal identity association approach integrating computer vision (CV) and radio frequency identification (RFID) technologies. To address the instability and unreliability of RSSI signals caused by environmental interference, signal loss, and multipath effects, this paper introduces a novel dynamic trajectory correlation-based matching algorithm, which shifts the focus from instantaneous distance values to temporal similarity of distance trends. This matching is performed only when both CV- and RFID-derived distances co-exist and exhibit consistent dynamic patterns, ensuring robustness against signal fluctuation and partial occlusions. The proposed method comprises: (1) a novel unsafe behavior identification method based on the PPE-based overlap rate between bounding boxes of workers and their personal protective equipment (PPE), where object detection is performed using YOLO-based models (YOLOv5, YOLOv8–YOLOv12); (2) visual tracking of individuals and estimation of left-wrist tag positions via human pose estimation; (3) dual-modality distance estimation using spatial coordinate transformation with spatial adjustment for CV and a path-loss model for RFID; (4) the trajectory-based identity alignment algorithm that leverages cross-modal dynamic consistency. Experimental evaluations demonstrate that YOLOv8 achieves the best object detection accuracy (mAP 0.969, F1-score 0.950) of the test dataset, YOLOv9 yields the highest unsafe behavior detection accuracy (0.8663) of the actual video data. Notably, the proposed trajectory-based identity association method reaches 1.0000 accuracy in stable tracking segments. These results confirm the method’s potential to accurately determine “who is unsafe” in real-world construction environments.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"71 ","pages":"Article 104380"},"PeriodicalIF":9.9,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146188613","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}

引用次数: 0

Mapping high-resolution real estate value distribution: a multi-attention deep generative model inspired by image inpainting 高分辨率房地产价值分布映射：基于图像绘画的多注意力深度生成模型

IF 9.9 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2026-02-01 DOI: 10.1016/j.aei.2026.104386

Feifeng Jiang , Jun Ma

Accurate real estate valuation is essential for urban planning, investment, and policy development. Traditional point-based methods treat properties as isolated units, oversimplifying complex urban environments and failing to capture spatial interactions and continuous value variation for informed decision-making. This study introduces REIN (Real Estate Inpainting Network), a novel deep generative model that leverages both local and surrounding context to predict high-resolution, spatially continuous property value distributions. By reframing valuation as a spatial inpainting task, REIN transforms multi-source urban data into image-like inputs and employs a hybrid multi-attention architecture—integrating channel–spatial interactions and dense–sparse contextual dependencies—to learn urban spatial structure and infer center values from their surroundings. A relative value estimation strategy further enhances adaptability across diverse regions. Applied to New York City, REIN outperforms existing models in both accuracy and visual coherence, demonstrating the effectiveness of its attention mechanisms and context-to-center inference strategy in property valuation. The model also exhibits strong generalizability under missing spatial context, incomplete features, and cross-region transfer, making it suitable for data-scarce planning scenarios. Feature importance analysis through the Squeeze-and-Excitation block reveals globally consistent and regionally adaptive value drivers across heterogeneous settings. By combining predictive precision, adaptivity, and interpretability, REIN provides an engineering informatics framework that supports planning simulations and data-driven urban policy decisions.

准确的房地产估值对城市规划、投资和政策制定至关重要。传统的基于点的方法将属性视为孤立的单元，过度简化了复杂的城市环境，无法捕捉空间相互作用和连续的价值变化，从而无法做出明智的决策。本研究引入了REIN (Real Estate Inpainting Network)，这是一种新颖的深度生成模型，它利用本地和周围环境来预测高分辨率、空间连续的财产价值分布。通过将评估重新定义为空间绘制任务，REIN将多源城市数据转换为类似图像的输入，并采用混合多关注架构（集成通道-空间交互和密集-稀疏上下文依赖）来学习城市空间结构并从周围环境中推断中心值。相对价值估计策略进一步增强了跨不同区域的适应性。在纽约市的应用中，REIN在准确性和视觉一致性方面都优于现有模型，证明了其注意机制和上下文到中心推理策略在房地产估值中的有效性。该模型在空间背景缺失、特征不完整、跨区域迁移等情况下具有较强的泛化能力，适用于数据稀缺的规划场景。通过挤压和激励块进行特征重要性分析，揭示了跨异构设置的全球一致和区域自适应的价值驱动因素。通过结合预测精度、适应性和可解释性，REIN提供了一个工程信息学框架，支持规划模拟和数据驱动的城市政策决策。

{"title":"Mapping high-resolution real estate value distribution: a multi-attention deep generative model inspired by image inpainting","authors":"Feifeng Jiang , Jun Ma","doi":"10.1016/j.aei.2026.104386","DOIUrl":"10.1016/j.aei.2026.104386","url":null,"abstract":"<div><div>Accurate real estate valuation is essential for urban planning, investment, and policy development. Traditional point-based methods treat properties as isolated units, oversimplifying complex urban environments and failing to capture spatial interactions and continuous value variation for informed decision-making. This study introduces REIN (Real Estate Inpainting Network), a novel deep generative model that leverages both local and surrounding context to predict high-resolution, spatially continuous property value distributions. By reframing valuation as a spatial inpainting task, REIN transforms multi-source urban data into image-like inputs and employs a hybrid multi-attention architecture—integrating channel–spatial interactions and dense–sparse contextual dependencies—to learn urban spatial structure and infer center values from their surroundings. A relative value estimation strategy further enhances adaptability across diverse regions. Applied to New York City, REIN outperforms existing models in both accuracy and visual coherence, demonstrating the effectiveness of its attention mechanisms and context-to-center inference strategy in property valuation. The model also exhibits strong generalizability under missing spatial context, incomplete features, and cross-region transfer, making it suitable for data-scarce planning scenarios. Feature importance analysis through the Squeeze-and-Excitation block reveals globally consistent and regionally adaptive value drivers across heterogeneous settings. By combining predictive precision, adaptivity, and interpretability, REIN provides an engineering informatics framework that supports planning simulations and data-driven urban policy decisions.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"71 ","pages":"Article 104386"},"PeriodicalIF":9.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146188922","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}

引用次数: 0

A VAE-GAT-based approach for energy consumption analysis and prediction in manufacturing workshops 制造车间能耗分析与预测的一种基于vae - gat的方法

IF 9.9 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2026-02-01 DOI: 10.1016/j.aei.2026.104390

Wei Chen, Liping Wang, Changchun Liu, Dunbing Tang, Zequn Zhang

In the global pursuit of carbon neutrality, the manufacturing industry is under increasing pressure to reduce energy waste. Excess consumption not only depletes resources but also hinders sustainable development. Accurate energy consumption prediction is therefore essential for scientific production scheduling and resource allocation, enabling loss reduction, efficiency improvement, and environmental performance enhancement. However, the complexity of modern manufacturing environments results in energy consumption data that is high-dimensional, noisy, and strongly spatiotemporal, which poses challenges to traditional prediction methods. To address these issues, this paper constructs an energy consumption behavior model considering key factors such as equipment status, processing techniques, and environmental conditions. A comprehensive feature analysis and data preprocessing are carried out to identify the key factors influencing consumption. Based on this, an optimization model is proposed that integrates an improved Variational Autoencoder (VAE) with an enhanced Graph Attention Network (GAT). VAE extracts compact latent representations from high-dimensional noisy inputs, suppressing redundancy while preserving essential patterns. GAT then captures complex spatiotemporal dependencies among energy-related features, thereby revealing intrinsic consumption dynamics. Experimental evaluations on both public and real-world datasets demonstrate that the proposed VAE-GAT model achieves superior prediction accuracy and generalization compared with other deep learning baselines. This approach provides a reliable foundation for energy management and contributes to advancing green intelligent manufacturing.

在全球追求碳中和的过程中，制造业面临着越来越大的减少能源浪费的压力。过度消费不仅消耗资源，而且阻碍可持续发展。因此，准确的能源消耗预测对于科学的生产调度和资源分配至关重要，从而实现减少损失、提高效率和增强环境绩效。然而，现代制造业环境的复杂性导致能耗数据具有高维、噪声和强烈的时空特征，这对传统的预测方法提出了挑战。为了解决这些问题，本文构建了考虑设备状态、加工工艺和环境条件等关键因素的能耗行为模型。通过综合特征分析和数据预处理，找出影响消费的关键因素。在此基础上，提出了一种将改进的变分自编码器（VAE）与增强的图注意网络（GAT）相结合的优化模型。VAE从高维噪声输入中提取紧凑的潜在表示，在保留基本模式的同时抑制冗余。然后，GAT捕获能源相关特征之间复杂的时空依赖关系，从而揭示内在的消费动态。在公共和现实数据集上的实验评估表明，与其他深度学习基线相比，所提出的VAE-GAT模型具有更高的预测精度和泛化能力。该方法为能源管理提供了可靠的基础，有助于推进绿色智能制造。

{"title":"A VAE-GAT-based approach for energy consumption analysis and prediction in manufacturing workshops","authors":"Wei Chen, Liping Wang, Changchun Liu, Dunbing Tang, Zequn Zhang","doi":"10.1016/j.aei.2026.104390","DOIUrl":"10.1016/j.aei.2026.104390","url":null,"abstract":"<div><div>In the global pursuit of carbon neutrality, the manufacturing industry is under increasing pressure to reduce energy waste. Excess consumption not only depletes resources but also hinders sustainable development. Accurate energy consumption prediction is therefore essential for scientific production scheduling and resource allocation, enabling loss reduction, efficiency improvement, and environmental performance enhancement. However, the complexity of modern manufacturing environments results in energy consumption data that is high-dimensional, noisy, and strongly spatiotemporal, which poses challenges to traditional prediction methods. To address these issues, this paper constructs an energy consumption behavior model considering key factors such as equipment status, processing techniques, and environmental conditions. A comprehensive feature analysis and data preprocessing are carried out to identify the key factors influencing consumption. Based on this, an optimization model is proposed that integrates an improved Variational Autoencoder (VAE) with an enhanced Graph Attention Network (GAT). VAE extracts compact latent representations from high-dimensional noisy inputs, suppressing redundancy while preserving essential patterns. GAT then captures complex spatiotemporal dependencies among energy-related features, thereby revealing intrinsic consumption dynamics. Experimental evaluations on both public and real-world datasets demonstrate that the proposed VAE-GAT model achieves superior prediction accuracy and generalization compared with other deep learning baselines. This approach provides a reliable foundation for energy management and contributes to advancing green intelligent manufacturing.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"71 ","pages":"Article 104390"},"PeriodicalIF":9.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146188614","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}

引用次数: 0

DSC-DETR: Real-time detection of transmission line defects using dynamic shuffle context DSC-DETR：利用动态洗牌上下文实时检测传输线缺陷

IF 9.9 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2026-01-31 DOI: 10.1016/j.aei.2026.104403

Yuanxue Xin , Jiabin Huang , Mengyao Xu , Jiawei Chen , Pengfei Shi

Power transmission line inspection is critical for ensuring grid reliability, yet accurately detecting diverse defects in real-time remains a persistent challenge due to the complexity of defect patterns and environmental variations. To address this problem, we propose the Dynamic Shuffle Context Detection Transformer (DSC-DETR), an end-to-end framework designed for efficient and high-precision defect detection in aerial inspection scenarios. Dynamic Shuffle Context Network (DSCNet) enhances the model’s detection capability for small-scale targets by dynamically integrating local details and global contextual information through cross-scale feature fusion. We designed an Efficient Coupled Encoder, which incorporates a Multilayer Coupled Focusing (MCF) Module. Within the Efficient Coupled Encoder, we deployed large-scale convolution operations using a parallel strategy. By progressively constructing a multi-scale perceptual field, we enhanced semantic information fusion across different feature levels. We incorporated Haar Wavelet Pooling Sampling (HWPS), implicitly encoding spatial relationships into channel dimensions to provide location cues for defect detection and improve accuracy. Our DSC-DETR achieves 52.8% AP (Average Precision) and 138.8 FPS (Frames Per Second) on the Transmission Line Multi-Category Defect (TLMD) dataset and 73% AP and 158.7 FPS on the Inspection of Power Line Assets Dataset (InsPLAD), respectively. This performance surpasses other State-of-the-Art (SOTA) methods in accuracy and real-time, and is beneficial to practical application in real-time transmission line fault scenarios.

输电线路检测是确保电网可靠性的关键，但由于缺陷模式和环境变化的复杂性，实时准确检测各种缺陷仍然是一个持续的挑战。为了解决这个问题，我们提出了动态Shuffle上下文检测转换器（DSC-DETR），这是一个端到端的框架，旨在在航空检测场景中高效、高精度地检测缺陷。动态随机上下文网络（Dynamic Shuffle Context Network, DSCNet）通过跨尺度特征融合动态整合局部细节和全局上下文信息，增强了模型对小尺度目标的检测能力。我们设计了一个高效的耦合编码器，它包含了一个多层耦合聚焦（MCF）模块。在高效耦合编码器中，我们使用并行策略部署了大规模卷积操作。通过逐步构建多尺度感知场，增强了不同特征层次的语义信息融合。采用Haar小波池采样（HWPS）方法，将空间关系隐式编码到通道维度中，为缺陷检测提供位置线索，提高检测精度。我们的DSC-DETR在传输线多类别缺陷（TLMD）数据集上实现了52.8%的AP（平均精度）和138.8 FPS（帧每秒），在电力线资产检查数据集（InsPLAD）上分别实现了73%的AP和158.7 FPS。这一性能在准确性和实时性上均优于其他SOTA方法，有利于实时输电线路故障场景的实际应用。

{"title":"DSC-DETR: Real-time detection of transmission line defects using dynamic shuffle context","authors":"Yuanxue Xin , Jiabin Huang , Mengyao Xu , Jiawei Chen , Pengfei Shi","doi":"10.1016/j.aei.2026.104403","DOIUrl":"10.1016/j.aei.2026.104403","url":null,"abstract":"<div><div>Power transmission line inspection is critical for ensuring grid reliability, yet accurately detecting diverse defects in real-time remains a persistent challenge due to the complexity of defect patterns and environmental variations. To address this problem, we propose the Dynamic Shuffle Context Detection Transformer (DSC-DETR), an end-to-end framework designed for efficient and high-precision defect detection in aerial inspection scenarios. Dynamic Shuffle Context Network (DSCNet) enhances the model’s detection capability for small-scale targets by dynamically integrating local details and global contextual information through cross-scale feature fusion. We designed an Efficient Coupled Encoder, which incorporates a Multilayer Coupled Focusing (MCF) Module. Within the Efficient Coupled Encoder, we deployed large-scale convolution operations using a parallel strategy. By progressively constructing a multi-scale perceptual field, we enhanced semantic information fusion across different feature levels. We incorporated Haar Wavelet Pooling Sampling (HWPS), implicitly encoding spatial relationships into channel dimensions to provide location cues for defect detection and improve accuracy. Our DSC-DETR achieves 52.8% AP (Average Precision) and 138.8 FPS (Frames Per Second) on the Transmission Line Multi-Category Defect (TLMD) dataset and 73% AP and 158.7 FPS on the Inspection of Power Line Assets Dataset (InsPLAD), respectively. This performance surpasses other State-of-the-Art (SOTA) methods in accuracy and real-time, and is beneficial to practical application in real-time transmission line fault scenarios.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"71 ","pages":"Article 104403"},"PeriodicalIF":9.9,"publicationDate":"2026-01-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146188923","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}

引用次数: 0

State prediction of adjacent operational tunnels under zoned foundation-pit excavation: a temporal decomposition network 分区基坑开挖下邻近运行隧道状态预测：一个时间分解网络

IF 9.9 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2026-01-31 DOI: 10.1016/j.aei.2026.104382

Yi Rui , Zeyu Dai , Xiaojun Li , Hehua Zhu , Yang Liu , Xi Wang , Mengqi Zhu , Chao Chen , Zhao Yuan

As urban underground space becomes increasingly congested, adjacent foundation-pit excavation poses a growing threat to the structural safety of operational tunnels. To address the complex disturbances induced by zonal excavation of large foundation-pits, we propose a spatiotemporally decoupled framework with dual inputs—retaining-structure displacements and construction constraints. The method integrates graph convolutional networks (GCNs), a Transformer architecture, and an interpretable module, leveraging multi-layer decomposition blocks to sequentially disentangle short-term fluctuations from long-term trends. Validated through a case study in Shanghai, the model achieves an average MAE of 0.1 mm, RMSE of 0.131 mm, and R² of 94.3% across three tasks. Ablation studies underscore the necessity of the temporal decomposition network and construction constraints for capturing subtle disturbances. The interpretable module further elucidates key influencing factors and disturbance propagation mechanisms. Compared to existing methods, the proposed approach demonstrates superior accuracy and robustness, offering critical insights for safeguarding operational tunnel safety.

随着城市地下空间的日益拥挤，相邻基坑开挖对运营隧道的结构安全构成越来越大的威胁。为了解决大基坑分区开挖引起的复杂扰动，我们提出了一个具有双输入-挡土墙位移和施工约束的时空解耦框架。该方法集成了图形卷积网络（GCNs）、Transformer架构和可解释模块，利用多层分解块依次从长期趋势中分离出短期波动。通过上海地区的案例研究验证，该模型在三个任务中的平均MAE为0.1 mm， RMSE为0.131 mm， R2为94.3%。消融研究强调了时间分解网络和构造约束对捕获细微扰动的必要性。可解释模块进一步阐明了关键影响因素和扰动传播机制。与现有方法相比，该方法具有更高的准确性和鲁棒性，为保障隧道运营安全提供了重要见解。

{"title":"State prediction of adjacent operational tunnels under zoned foundation-pit excavation: a temporal decomposition network","authors":"Yi Rui , Zeyu Dai , Xiaojun Li , Hehua Zhu , Yang Liu , Xi Wang , Mengqi Zhu , Chao Chen , Zhao Yuan","doi":"10.1016/j.aei.2026.104382","DOIUrl":"10.1016/j.aei.2026.104382","url":null,"abstract":"<div><div>As urban underground space becomes increasingly congested, adjacent foundation-pit excavation poses a growing threat to the structural safety of operational tunnels. To address the complex disturbances induced by zonal excavation of large foundation-pits, we propose a spatiotemporally decoupled framework with dual inputs—retaining-structure displacements and construction constraints. The method integrates graph convolutional networks (GCNs), a Transformer architecture, and an interpretable module, leveraging multi-layer decomposition blocks to sequentially disentangle short-term fluctuations from long-term trends. Validated through a case study in Shanghai, the model achieves an average MAE of 0.1 mm, RMSE of 0.131 mm, and R<sup>2</sup> of 94.3% across three tasks. Ablation studies underscore the necessity of the temporal decomposition network and construction constraints for capturing subtle disturbances. The interpretable module further elucidates key influencing factors and disturbance propagation mechanisms. Compared to existing methods, the proposed approach demonstrates superior accuracy and robustness, offering critical insights for safeguarding operational tunnel safety.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"71 ","pages":"Article 104382"},"PeriodicalIF":9.9,"publicationDate":"2026-01-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146188921","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}

引用次数: 0

Encoder-decoder based active learning approach for corrosion segmentation in industrial and lab environments 基于编码器-解码器的主动学习方法在工业和实验室环境腐蚀分割

IF 9.9 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2026-01-31 DOI: 10.1016/j.aei.2026.104366

Zhen Qi Chee , Cheng Siong Chin , Hao Chen , Zi Jie Choong , Jun Jie Chong , Carla Canturri , Tom Portafaix , ShiLiang Johnathan Tan

Despite significant progress in corrosion monitoring, accurately targeting critical areas remains a persistent challenge due to irregular textures and environmental variability, which limit the effectiveness of traditional transfer learning approaches. To address this, this study explores the potential of optimised pool-based active learning to enhance corrosion detection. Pool-based active learning prioritises high-value samples, improving segmentation performance while reducing annotation costs by focusing on refining sample selection for corrosion-specific features rather than generic image uncertainty. Two distinct datasets were used to validate the segmentation model rigorously. The first is a laboratory-controlled dataset featuring standardised corrosion samples with precise ground-truth annotations, and the second is a site-realistic dataset captured under real-world environmental conditions. The laboratory experiments were conducted first to validate the methodology under controlled conditions, ensuring accurate segmentation against well-defined corrosion samples, before progressing to the site dataset. Experimental results demonstrate that the DeepLabv3 + model with an EfficientNet backbone, train with batch size of 16 and 50 epochs with an 80% train, 10% validation and 10% test dataset split using the Bayesian Active Learning by Disagreement (BALD) method, achieves 98% ± 0.16% pixel accuracy in controlled laboratory conditions and 87.8% ± 0.98% pixel accuracy on real-world on-site images. Furthermore, the on-site model demonstrated robust segmentation capabilities with a mean Intersection over Union (IoU) of 86.7% ± 0.28%, under challenging conditions. The findings underscore the strengths and trade-offs of active learning in corrosion detection. Future work would explore further optimisation methods to balance accuracy, efficiency, and scalability across diverse operating conditions.

尽管在腐蚀监测方面取得了重大进展，但由于结构不规则和环境可变性，准确定位关键区域仍然是一个持续的挑战，这限制了传统迁移学习方法的有效性。为了解决这个问题，本研究探索了优化的基于池的主动学习的潜力，以增强腐蚀检测。基于池的主动学习优先考虑高价值样本，提高分割性能，同时通过专注于细化腐蚀特定特征的样本选择而不是通用图像不确定性来降低注释成本。使用两个不同的数据集严格验证分割模型。第一个是实验室控制的数据集，具有标准化腐蚀样品和精确的地面真相注释，第二个是在真实环境条件下捕获的现场真实数据集。在进入现场数据集之前，首先进行实验室实验，在受控条件下验证该方法，确保对定义明确的腐蚀样品进行准确分割。实验结果表明，采用高效网（EfficientNet）骨架、16次和50次批处理训练、80%训练、10%验证和10%测试数据集分割的DeepLabv3 +模型，在实验室控制条件下达到98%±0.16%的像素精度，在真实现场图像上达到87.8%±0.98%的像素精度。此外，现场模型显示出强大的分割能力，在具有挑战性的条件下，平均交汇比（IoU）为86.7%±0.28%。研究结果强调了主动学习在腐蚀检测中的优势和权衡。未来的工作将探索进一步的优化方法，以平衡不同操作条件下的准确性、效率和可扩展性。

{"title":"Encoder-decoder based active learning approach for corrosion segmentation in industrial and lab environments","authors":"Zhen Qi Chee , Cheng Siong Chin , Hao Chen , Zi Jie Choong , Jun Jie Chong , Carla Canturri , Tom Portafaix , ShiLiang Johnathan Tan","doi":"10.1016/j.aei.2026.104366","DOIUrl":"10.1016/j.aei.2026.104366","url":null,"abstract":"<div><div>Despite significant progress in corrosion monitoring, accurately targeting critical areas remains a persistent challenge due to irregular textures and environmental variability, which limit the effectiveness of traditional transfer learning approaches. To address this, this study explores the potential of optimised pool-based active learning to enhance corrosion detection. Pool-based active learning prioritises high-value samples, improving segmentation performance while reducing annotation costs by focusing on refining sample selection for corrosion-specific features rather than generic image uncertainty. Two distinct datasets were used to validate the segmentation model rigorously. The first is a laboratory-controlled dataset featuring standardised corrosion samples with precise ground-truth annotations, and the second is a site-realistic dataset captured under real-world environmental conditions. The laboratory experiments were conducted first to validate the methodology under controlled conditions, ensuring accurate segmentation against well-defined corrosion samples, before progressing to the site dataset. Experimental results demonstrate that the DeepLabv3 + model with an EfficientNet backbone, train with batch size of 16 and 50 epochs with an 80% train, 10% validation and 10% test dataset split using the Bayesian Active Learning by Disagreement (BALD) method, achieves 98% ± 0.16% pixel accuracy in controlled laboratory conditions and 87.8% ± 0.98% pixel accuracy on real-world on-site images. Furthermore, the on-site model demonstrated robust segmentation capabilities with a mean Intersection over Union (IoU) of 86.7% ± 0.28%, under challenging conditions. The findings underscore the strengths and trade-offs of active learning in corrosion detection. Future work would explore further optimisation methods to balance accuracy, efficiency, and scalability across diverse operating conditions.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"71 ","pages":"Article 104366"},"PeriodicalIF":9.9,"publicationDate":"2026-01-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077884","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}

引用次数: 0

AI-supported seismic performance evaluation of structures: challenges, gaps, and future directions at early design stages 人工智能支持的结构抗震性能评估：早期设计阶段的挑战、差距和未来方向

IF 9.9 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2026-01-30 DOI: 10.1016/j.aei.2025.104301

Fatma Ak, Berk Ekici, Ugur Demir

This study reviews 91 journal articles that intersect with earthquake-resistant building design and artificial intelligence (AI)- based modeling, utilizing machine learning, deep learning, and metaheuristic optimization algorithms. Previous reviews on AI applications have examined engineering problems without considering the impact of architectural design parameters and structural irregularities on seismic performance. This review discusses the role of AI in integrating architectural design variables and seismic performance objectives, highlighting challenges, gaps, and future directions in the early design phase. The reviewed articles demonstrate that AI is successful in addressing seismic performance objectives; however, a holistic framework for assessing architectural and structural variables has not been presented. The review highlights key findings, gaps, and future directions for those involved in earthquake-resistant building design utilizing AI.

本研究回顾了91篇与抗震建筑设计和基于人工智能（AI）的建模相关的期刊文章，这些文章利用了机器学习、深度学习和元启发式优化算法。以前对人工智能应用的评论检查了工程问题，而没有考虑建筑设计参数和结构不规则对抗震性能的影响。本文讨论了人工智能在整合建筑设计变量和抗震性能目标方面的作用，强调了早期设计阶段的挑战、差距和未来方向。回顾的文章表明，人工智能在解决地震性能目标方面是成功的；然而，评估建筑和结构变量的整体框架尚未提出。该综述强调了利用人工智能进行抗震建筑设计的主要发现、差距和未来方向。

引用次数: 0