Advanced Engineering Informatics最新文献

InceptionV3 based blockage fault diagnosis of centrifugal pump

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-09 DOI: 10.1016/j.aei.2025.103181

Deepak Kumar , Nagendra Singh Ranawat , Pavan Kumar Kankar , Ankur Miglani

Blockages in a centrifugal pump reduce the performance of the pump and can result in operational breakdown if not identified on time. Thus, it is crucial to identify blockages in the pump to ensure its reliable working. This study presents a modified InceptionV3 Deep Convolutional Neural Network (CNN) model to classify blockage faults and their severity in centrifugal pumps based on Suction pressure signal, highlighting its importance in predictive analytics. Initially, the methodology involves plotting two-dimensional Continuous Wavelet Transform (CWT) plots of pressure signals acquired from the pump test rig. The InceptionV3 model is employed to classify blockages in the pump. Validation and test accuracies of InceptionV3 are achieved as 81.71 % and 81.82 %, respectively at 2000 RPM. Afterward, the blockage fault diagnosis capabilities of InceptionV3 are improved by integrating Exponential Linear Units (ELUs) with Rectified Linear Units (ReLUs) in the model’s architecture. It is found that validation and test accuracies increased to 95.48 % and 95.36 %, respectively, in modified InceptionV3 on 35 classes (one healthy and 34 faulty conditions) at 2000RPM. The robustness of the modified InceptionV3 model is highlighted by comparing test accuracy, precision, recall, and F1 score to the original InceptionV3 model. The proposed methodology demonstrates the modified InceptionV3 model’s efficacy in classifying various blockage conditions, making it a potent tool for predictive maintenance in industrial settings.

{"title":"InceptionV3 based blockage fault diagnosis of centrifugal pump","authors":"Deepak Kumar , Nagendra Singh Ranawat , Pavan Kumar Kankar , Ankur Miglani","doi":"10.1016/j.aei.2025.103181","DOIUrl":"10.1016/j.aei.2025.103181","url":null,"abstract":"<div><div>Blockages in a centrifugal pump reduce the performance of the pump and can result in operational breakdown if not identified on time. Thus, it is crucial to identify blockages in the pump to ensure its reliable working. This study presents a modified InceptionV3 Deep Convolutional Neural Network (CNN) model to classify blockage faults and their severity in centrifugal pumps based on Suction pressure signal, highlighting its importance in predictive analytics. Initially, the methodology involves plotting two-dimensional Continuous Wavelet Transform (CWT) plots of pressure signals acquired from the pump test rig. The InceptionV3 model is employed to classify blockages in the pump. Validation and test accuracies of InceptionV3 are achieved as 81.71 % and 81.82 %, respectively at 2000 RPM. Afterward, the blockage fault diagnosis capabilities of InceptionV3 are improved by integrating Exponential Linear Units (ELUs) with Rectified Linear Units (ReLUs) in the model’s architecture. It is found that validation and test accuracies increased to 95.48 % and 95.36 %, respectively, in modified InceptionV3 on 35 classes (one healthy and 34 faulty conditions) at 2000RPM. The robustness of the modified InceptionV3 model is highlighted by comparing test accuracy, precision, recall, and F1 score to the original InceptionV3 model. The proposed methodology demonstrates the modified InceptionV3 model’s efficacy in classifying various blockage conditions, making it a potent tool for predictive maintenance in industrial settings.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103181"},"PeriodicalIF":8.0,"publicationDate":"2025-02-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143372291","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 hybrid CNN-LSTM model for involuntary fall detection using wrist-worn sensors

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-08 DOI: 10.1016/j.aei.2025.103178

Xinyao Hu, Shiling Yu, Jihan Zheng, Zhimeng Fang, Zhong Zhao, Xingda Qu

Falls have become a significant global safety and health concern, which may lead to physical injuries as well as declined mental health, reduced mobility, and deteriorated quality of life. Wearable sensor-based fall detection has emerged as a promising solution for preventing fall-related injuries. However, existing solutions are hindered by user compliance related to sensor placement locations, overall model accuracy, and dependence on simulated voluntary falls. To overcome these limitations, this study aimed to propose a novel involuntary fall detection solution by using wearable sensors and deep learning algorithms. Forty-nine participants were involved in an experimental study, in which activities of daily living and involuntary falls were simulated and kinematic data from these activities were collected using wrist-worn sensors. A novel hybrid model which integrates a convolutional neural network (CNN) and a long short-term memory (LSTM) model was proposed and its performance was compared with the CNN-alone model and LSTM-alone model. The results showed that the proposed hybrid CNN-LSTM model could effectively detect involuntary falls with 96.94% detection accuracy, 98.33% sensitivity, and 96.67% specificity, superior to the CNN-alone model and LSTM-alone model. These results highlight the effectiveness of our proposed approach in significantly improving fall detection accuracy, providing a more reliable and less intrusive solution for preventing fall-related injuries.

{"title":"A hybrid CNN-LSTM model for involuntary fall detection using wrist-worn sensors","authors":"Xinyao Hu, Shiling Yu, Jihan Zheng, Zhimeng Fang, Zhong Zhao, Xingda Qu","doi":"10.1016/j.aei.2025.103178","DOIUrl":"10.1016/j.aei.2025.103178","url":null,"abstract":"<div><div>Falls have become a significant global safety and health concern, which may lead to physical injuries as well as declined mental health, reduced mobility, and deteriorated quality of life. Wearable sensor-based fall detection has emerged as a promising solution for preventing fall-related injuries. However, existing solutions are hindered by user compliance related to sensor placement locations, overall model accuracy, and dependence on simulated voluntary falls. To overcome these limitations, this study aimed to propose a novel involuntary fall detection solution by using wearable sensors and deep learning algorithms. Forty-nine participants were involved in an experimental study, in which activities of daily living and involuntary falls were simulated and kinematic data from these activities were collected using wrist-worn sensors. A novel hybrid model which integrates a convolutional neural network (CNN) and a long short-term memory (LSTM) model was proposed and its performance was compared with the CNN-alone model and LSTM-alone model. The results showed that the proposed hybrid CNN-LSTM model could effectively detect involuntary falls with 96.94% detection accuracy, 98.33% sensitivity, and 96.67% specificity, superior to the CNN-alone model and LSTM-alone model. These results highlight the effectiveness of our proposed approach in significantly improving fall detection accuracy, providing a more reliable and less intrusive solution for preventing fall-related injuries.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103178"},"PeriodicalIF":8.0,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143350700","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

Investigating the effects of data and image enhancement techniques on crack detection accuracy in FMPI

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-08 DOI: 10.1016/j.aei.2025.103169

Qiang Wu , Xunpeng Qin , Xiaochen Xiong

Fluorescent magnetic particle inspection (FMPI) is a vital non-destructive testing technique for detecting surface defects in ferromagnetic materials. However, existing research on FMPI crack detection using deep learning models has been hindered by the limited availability of high-quality and diverse training data. This study addresses this challenge by proposing an approach to synthesize and enhance FMPI crack images, enabling comprehensive exploration of data augmentation strategies and their impact on model performance. A large-scale dataset of high-quality FMPI crack images is generated through a stepwise image synthesis method combining a diffusion model and Poisson image blending. Leveraging the synthesized dataset, the effects of various spatial and pixel-level transformations on crack detection accuracy are systematically investigated, leading to the identification of optimal data augmentation strategies tailored to the unique characteristics of FMPI crack images. A ToneCurve mapping method is developed for image enhancement, enhancing the contrast between crack indications and backgrounds, further improving model performance. The proposed image synthesis and enhancement methods significantly boost crack detection precision on a small-sample FMPI dataset, achieving a 35.2% and 17.6% improvement in mean Average Precision ([email protected], YOLOv5s), and a 27.6% and 8.3% improvement ([email protected], YOLOv8s), compared to non-enhancement and conventional enhancement methods, respectively, demonstrating their practical applicability. The findings underscore the importance of data augmentation strategies and the effectiveness of the proposed methods in enhancing FMPI crack detection accuracy, particularly in scenarios with limited training data. The synthesized dataset is open-sourced (https://drive.google.com/drive/folders/1ES47PcW1y6CobrOVr29jGmU6kMdeECJl?usp=sharing) to facilitate further research in this field.

{"title":"Investigating the effects of data and image enhancement techniques on crack detection accuracy in FMPI","authors":"Qiang Wu , Xunpeng Qin , Xiaochen Xiong","doi":"10.1016/j.aei.2025.103169","DOIUrl":"10.1016/j.aei.2025.103169","url":null,"abstract":"<div><div>Fluorescent magnetic particle inspection (FMPI) is a vital non-destructive testing technique for detecting surface defects in ferromagnetic materials. However, existing research on FMPI crack detection using deep learning models has been hindered by the limited availability of high-quality and diverse training data. This study addresses this challenge by proposing an approach to synthesize and enhance FMPI crack images, enabling comprehensive exploration of data augmentation strategies and their impact on model performance. A large-scale dataset of high-quality FMPI crack images is generated through a stepwise image synthesis method combining a diffusion model and Poisson image blending. Leveraging the synthesized dataset, the effects of various spatial and pixel-level transformations on crack detection accuracy are systematically investigated, leading to the identification of optimal data augmentation strategies tailored to the unique characteristics of FMPI crack images. A ToneCurve mapping method is developed for image enhancement, enhancing the contrast between crack indications and backgrounds, further improving model performance. The proposed image synthesis and enhancement methods significantly boost crack detection precision on a small-sample FMPI dataset, achieving a 35.2% and 17.6% improvement in mean Average Precision ([email protected], YOLOv5s), and a 27.6% and 8.3% improvement ([email protected], YOLOv8s), compared to non-enhancement and conventional enhancement methods, respectively, demonstrating their practical applicability. The findings underscore the importance of data augmentation strategies and the effectiveness of the proposed methods in enhancing FMPI crack detection accuracy, particularly in scenarios with limited training data. The synthesized dataset is open-sourced (<span><span>https://drive.google.com/drive/folders/1ES47PcW1y6CobrOVr29jGmU6kMdeECJl?usp=sharing</span><svg><path></path></svg></span>) to facilitate further research in this field.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103169"},"PeriodicalIF":8.0,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143372290","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

Robust correlation measures for informative frequency band selection in heavy-tailed signals

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-08 DOI: 10.1016/j.aei.2025.103174

Justyna Hebda-Sobkowicz , Radosław Zimroz , Anil Kumar , Agnieszka Wyłomańska

Vibration signals are commonly used to detect local damage in rotating machinery. However, raw signals are often noisy, particularly in crusher machines, where the technological process (falling pieces of rock) generates random impulses that complicate detection. To address this, signal pre-filtering (extracting the informative frequency band from noise-affected signals) is necessary. This paper proposes an algorithm for processing vibration signals from a bearing used in an ore crusher. Selecting informative frequency bands (IFBs) in the presence of impulsive noise is notably challenging. The approach employs correlation maps to detect cyclic behavior within specific frequency bands in the time–frequency domain (spectrogram), enabling the identification of IFBs. Robust correlation measures and median filtering are applied to enhance the correlation maps and improve the final IFB selection. Signal segmentation and the use of averaged results for IFB selection are also highlighted. The proposed trimmed and quadrant correlations are compared with the Pearson and Kendall correlations using simulated signal, real vibration signal from crusher in mining industry and acoustic signal measured on the test rig. Furthermore, the results of real vibration analyses are compared with established IFB selectors, including the spectral kurtosis, the alpha selector and the conditional variance-based selector.

{"title":"Robust correlation measures for informative frequency band selection in heavy-tailed signals","authors":"Justyna Hebda-Sobkowicz , Radosław Zimroz , Anil Kumar , Agnieszka Wyłomańska","doi":"10.1016/j.aei.2025.103174","DOIUrl":"10.1016/j.aei.2025.103174","url":null,"abstract":"<div><div>Vibration signals are commonly used to detect local damage in rotating machinery. However, raw signals are often noisy, particularly in crusher machines, where the technological process (falling pieces of rock) generates random impulses that complicate detection. To address this, signal pre-filtering (extracting the informative frequency band from noise-affected signals) is necessary. This paper proposes an algorithm for processing vibration signals from a bearing used in an ore crusher. Selecting informative frequency bands (IFBs) in the presence of impulsive noise is notably challenging. The approach employs correlation maps to detect cyclic behavior within specific frequency bands in the time–frequency domain (spectrogram), enabling the identification of IFBs. Robust correlation measures and median filtering are applied to enhance the correlation maps and improve the final IFB selection. Signal segmentation and the use of averaged results for IFB selection are also highlighted. The proposed trimmed and quadrant correlations are compared with the Pearson and Kendall correlations using simulated signal, real vibration signal from crusher in mining industry and acoustic signal measured on the test rig. Furthermore, the results of real vibration analyses are compared with established IFB selectors, including the spectral kurtosis, the alpha selector and the conditional variance-based selector.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103174"},"PeriodicalIF":8.0,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143350699","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

Uncertainty quantification of aerosol jet 3D printing process using non-intrusive polynomial chaos and stochastic collocation

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-08 DOI: 10.1016/j.aei.2025.103175

Haining Zhang , Jingyuan Huang , Xiaoge Zhang , Chak-Nam Wong

Aerosol jet printing (AJP) emerges as an innovative three-dimensional (3D) printing technology by offering a versatile approach for fabricating customized and conformal microelectronic devices directly onto various flexible substrates. There is, though, an inherent process uncertainty in AJP that often leads to variations in critical geometrical properties, particularly printing overspray, which diminishes the reproducibility and uniformity of the produced components. While notable advancements have been made in recent years in modeling and elucidating the critical geometrical properties of AJP, a significant research gap persists in systematically quantifying the uncertainties inherent in the developed physics-based models, which may undermine process reliability and hamper informed decision-making during printing. In this study, an uncertainty quantification (UQ) analysis is conducted through non-intrusive generalized polynomial chaos expansion (gPCE) and stochastic collocation within a developed computational fluid dynamics (CFD) model applied in AJP. This analysis quantifies the variability in model responses due to uncertainties in the input parameters. Specifically, uncertainties in the main process parameters are effectively captured by modeling them as Gaussian random variables. Subsequently, the modeled input uncertainties are mapped into the stochastic space via a stochastic collocation technique. This is followed by computational simulations of the Navier–Stokes equations conducted using the designated collocation points within a developed CFD model. Finally, a non-intrusive gPCE approach is employed to quantify the uncertainties in velocity and pressure fields, as well as in particle trajectories, based on fluctuations in input process parameters. To the best of the authors’ knowledge, there is no prior investigations made to conduct formal UQ analysis on physics-based models for AJP process. The primary contribution of this study is to address the research gap concerning the lack of systematic studies on UQ analysis for CFD models used in AJP.

{"title":"Uncertainty quantification of aerosol jet 3D printing process using non-intrusive polynomial chaos and stochastic collocation","authors":"Haining Zhang , Jingyuan Huang , Xiaoge Zhang , Chak-Nam Wong","doi":"10.1016/j.aei.2025.103175","DOIUrl":"10.1016/j.aei.2025.103175","url":null,"abstract":"<div><div>Aerosol jet printing (AJP) emerges as an innovative three-dimensional (3D) printing technology by offering a versatile approach for fabricating customized and conformal microelectronic devices directly onto various flexible substrates. There is, though, an inherent process uncertainty in AJP that often leads to variations in critical geometrical properties, particularly printing overspray, which diminishes the reproducibility and uniformity of the produced components. While notable advancements have been made in recent years in modeling and elucidating the critical geometrical properties of AJP, a significant research gap persists in systematically quantifying the uncertainties inherent in the developed physics-based models, which may undermine process reliability and hamper informed decision-making during printing. In this study, an uncertainty quantification (UQ) analysis is conducted through non-intrusive generalized polynomial chaos expansion (gPCE) and stochastic collocation within a developed computational fluid dynamics (CFD) model applied in AJP. This analysis quantifies the variability in model responses due to uncertainties in the input parameters. Specifically, uncertainties in the main process parameters are effectively captured by modeling them as Gaussian random variables. Subsequently, the modeled input uncertainties are mapped into the stochastic space via a stochastic collocation technique. This is followed by computational simulations of the Navier–Stokes equations conducted using the designated collocation points within a developed CFD model. Finally, a non-intrusive gPCE approach is employed to quantify the uncertainties in velocity and pressure fields, as well as in particle trajectories, based on fluctuations in input process parameters. To the best of the authors’ knowledge, there is no prior investigations made to conduct formal UQ analysis on physics-based models for AJP process. The primary contribution of this study is to address the research gap concerning the lack of systematic studies on UQ analysis for CFD models used in AJP.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103175"},"PeriodicalIF":8.0,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143372289","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

Create or revise? A comparative study on CAD rework after team-based engineering design review in virtual reality and desktop interface

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-06 DOI: 10.1016/j.aei.2025.103177

Nikola Horvat , Jelena Šklebar , Mario Štorga , Stanko Škec

While numerous studies have delved into the effect of virtual reality (VR) on design reviews (DR), this research study explores how the use of VR in team-based DRs influences the design work after the review. Thus, it explores the effects of VR on broader design contexts. The study employed an experimental case study involving 14 design teams working in CAD over several weeks, engaging in a design review with external reviewers, and subsequently revising the design based on the feedback. The experimental aspect involved randomly allocating teams to one of two conditions: low-immersion (desktop interface) or high-immersion (VR). Furthermore, the results indicate that teams that had DR in VR executed slightly more CAD actions compared to those that underwent DR in low-immersion. Furthermore, the VR group exhibited a significantly higher proportion of creation actions and assembly actions compared to the low-immersion group. These findings suggest that incorporating VR into DRs has the potential to change the course of the design process, making it a valuable tool for early design phases or agile methodologies, primarily due to an increased focus on creation during the rework phases. The findings also highlight the distinct focus of designers before and after the DR in terms of creation and revision, emphasizing the need for CAD tools to be more adaptable and responsive to the evolving needs of designers, considering both the phase of design and the broader ecosystem of design support tools. In summary, this study serves as an initial step for implementing VR in the industry, demonstrating that its use can indeed change the course of the design.

尽管许多研究已经深入探讨了虚拟现实（VR）对设计评审（DR）的影响，但本研究探讨了在基于团队的设计评审中使用 VR 如何影响评审后的设计工作。因此，本研究探讨了虚拟现实技术对更广泛的设计环境的影响。本研究采用了一项实验性案例研究，涉及 14 个设计团队，他们在数周内使用 CAD 进行工作，与外部评审人员一起参与设计评审，随后根据反馈意见修改设计。实验方面包括随机分配团队到两种条件之一：低沉浸（桌面界面）或高沉浸（VR）。此外，实验结果表明，与在低浸入式环境中进行 DR 的团队相比，在 VR 环境中进行 DR 的团队执行的 CAD 操作略多。此外，与低浸入式组相比，VR 组的创建操作和装配操作比例明显更高。这些研究结果表明，将 VR 纳入 DR 有可能改变设计流程的进程，使其成为早期设计阶段或敏捷方法的重要工具，这主要是因为在返工阶段会更加关注创造。研究结果还强调了设计人员在 DR 之前和之后在创建和修改方面的不同侧重点，强调了 CAD 工具在考虑设计阶段和更广泛的设计支持工具生态系统的同时，需要具有更强的适应性，以满足设计人员不断变化的需求。总之，这项研究为在行业中实施虚拟现实技术迈出了第一步，证明了虚拟现实技术的使用确实可以改变设计的进程。

{"title":"Create or revise? A comparative study on CAD rework after team-based engineering design review in virtual reality and desktop interface","authors":"Nikola Horvat , Jelena Šklebar , Mario Štorga , Stanko Škec","doi":"10.1016/j.aei.2025.103177","DOIUrl":"10.1016/j.aei.2025.103177","url":null,"abstract":"<div><div>While numerous studies have delved into the effect of virtual reality (VR) on design reviews (DR), this research study explores how the use of VR in team-based DRs influences the design work after the review. Thus, it explores the effects of VR on broader design contexts. The study employed an experimental case study involving 14 design teams working in CAD over several weeks, engaging in a design review with external reviewers, and subsequently revising the design based on the feedback. The experimental aspect involved randomly allocating teams to one of two conditions: low-immersion (desktop interface) or high-immersion (VR). Furthermore, the results indicate that teams that had DR in VR executed slightly more CAD actions compared to those that underwent DR in low-immersion. Furthermore, the VR group exhibited a significantly higher proportion of creation actions and assembly actions compared to the low-immersion group. These findings suggest that incorporating VR into DRs has the potential to change the course of the design process, making it a valuable tool for early design phases or agile methodologies, primarily due to an increased focus on creation during the rework phases. The findings also highlight the distinct focus of designers before and after the DR in terms of creation and revision, emphasizing the need for CAD tools to be more adaptable and responsive to the evolving needs of designers, considering both the phase of design and the broader ecosystem of design support tools. In summary, this study serves as an initial step for implementing VR in the industry, demonstrating that its use can indeed change the course of the design.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103177"},"PeriodicalIF":8.0,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143368498","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

Retrieval augmented generation-driven information retrieval and question answering in construction management

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-06 DOI: 10.1016/j.aei.2025.103158

Chengke Wu , Wenjun Ding , Qisen Jin , Junjie Jiang , Rui Jiang , Qinge Xiao , Longhui Liao , Xiao Li

Construction management is a communication-intensive field, requiring prompt responses to queries from various stakeholders to ensure project continuity. However, retrieving accurate information from project documents is hampered by the mismatch in granularity between queries and vast contents and by inherent ambiguities in information. Large language models (LLMs) and retrieval-augmented generation (RAG) offer new opportunities to address the challenges. However, their effectiveness is limited by the segmentation of documents and insufficient consideration of engineers’ preferences. Therefore, we propose a novel paradigm: RAG for Construction Management (RAG4CM). It includes three components: 1) a pipeline that parses project documents into hierarchical structures to establish a knowledge pool; 2) novel RAG search algorithms; and 3) a user preference learning mechanism. The first two components enhance granularity alignment and RAG results by integrating document-level hierarchical features with raw contents. The preference learning realizes continuously improved responses along with user-system interactions. We developed a prototype system and conducted extensive experiments, demonstrating that the knowledge pool efficiently accommodates texts, tables, and images. RAG4CM realized a 0.924 Top-3 and 0.898 answer accuracy, surpassing both open-source frameworks and commercial products. In addition, preference learning further increases answer accuracy by 1.3 % to 9.5 %. Consequently, RAG4CM enables multi-source information retrieval in a user-friendly manner, improving communication efficiency and facilitating construction management activities.

{"title":"Retrieval augmented generation-driven information retrieval and question answering in construction management","authors":"Chengke Wu , Wenjun Ding , Qisen Jin , Junjie Jiang , Rui Jiang , Qinge Xiao , Longhui Liao , Xiao Li","doi":"10.1016/j.aei.2025.103158","DOIUrl":"10.1016/j.aei.2025.103158","url":null,"abstract":"<div><div>Construction management is a communication-intensive field, requiring prompt responses to queries from various stakeholders to ensure project continuity. However, retrieving accurate information from project documents is hampered by the mismatch in granularity between queries and vast contents and by inherent ambiguities in information. Large language models (LLMs) and retrieval-augmented generation (RAG) offer new opportunities to address the challenges. However, their effectiveness is limited by the segmentation of documents and insufficient consideration of engineers’ preferences. Therefore, we propose a novel paradigm: RAG for Construction Management (RAG4CM). It includes three components: 1) a pipeline that parses project documents into hierarchical structures to establish a knowledge pool; 2) novel RAG search algorithms; and 3) a user preference learning mechanism. The first two components enhance granularity alignment and RAG results by integrating document-level hierarchical features with raw contents. The preference learning realizes continuously improved responses along with user-system interactions. We developed a prototype system and conducted extensive experiments, demonstrating that the knowledge pool efficiently accommodates texts, tables, and images. RAG4CM realized a 0.924 Top-3 and 0.898 answer accuracy, surpassing both open-source frameworks and commercial products. In addition, preference learning further increases answer accuracy by 1.3 % to 9.5 %. Consequently, RAG4CM enables multi-source information retrieval in a user-friendly manner, improving communication efficiency and facilitating construction management activities.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103158"},"PeriodicalIF":8.0,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143368297","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 novel user scenario and behavior sequence recognition approach based on vision-context fusion architecture

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-06 DOI: 10.1016/j.aei.2025.103161

Wenyu Yuan, Danni Chang, Chenlu Mao, Luyao Wang, Ke Ren, Ting Han

Understanding user scenario and behavior is essential for the development of human-centered intelligent service systems. However, the presence of cluttered objects, uncertain human behaviors, and overlapping timelines in daily life scenarios complicates the problem of scenario understanding. This paper aims to address the challenges of identifying and predicting user scenario and behavior sequences through a multimodal data fusion approach, focusing on the integration of visual and environmental data to capture subtle scenario and behavioral features.

For the purpose, a novel Vision-Context Fusion Scenario Recognition (VCFSR) approach was proposed, encompassing three stages. First, four categories of context data related to home scenarios were acquired: physical context, time context, user context, and inferred context. Second, scenarios were represented as multidimensional data relationships through modeling technologies. Third, a scenario recognition model was developed, comprising context feature processing, visual feature handling, and multimodal feature fusion. For illustration, a smart home environment was built, and twenty-six participants were recruited to perform various home activities. Integral sensors were used to collect environmental context data, and video data was captured simultaneously, both of which jointly form a multimodal dataset. Results demonstrated that the VCFSR model achieved an average accuracy of 98.1 %, outperforming traditional machine learning models such as decision trees and support vector machines. This method was then employed for fine-grained human behavior sequence prediction tasks, showing good performance in predicting behavior sequences across all scenarios constructed in this study. Furthermore, the results of ablation experiments revealed that the multimodal feature fusion method increased the average accuracy by at least 1.8 % compared to single-modality data-driven methods.

This novel approach to user behavior modeling simultaneously handles the relationship threads across scenarios and the rich details provided by visual data, paving the way for advanced intelligent services in complex interactive environments such as smart homes and hospitals.

{"title":"A novel user scenario and behavior sequence recognition approach based on vision-context fusion architecture","authors":"Wenyu Yuan, Danni Chang, Chenlu Mao, Luyao Wang, Ke Ren, Ting Han","doi":"10.1016/j.aei.2025.103161","DOIUrl":"10.1016/j.aei.2025.103161","url":null,"abstract":"<div><div>Understanding user scenario and behavior is essential for the development of human-centered intelligent service systems. However, the presence of cluttered objects, uncertain human behaviors, and overlapping timelines in daily life scenarios complicates the problem of scenario understanding. This paper aims to address the challenges of identifying and predicting user scenario and behavior sequences through a multimodal data fusion approach, focusing on the integration of visual and environmental data to capture subtle scenario and behavioral features.</div><div>For the purpose, a novel Vision-Context Fusion Scenario Recognition (VCFSR) approach was proposed, encompassing three stages. First, four categories of context data related to home scenarios were acquired: physical context, time context, user context, and inferred context. Second, scenarios were represented as multidimensional data relationships through modeling technologies. Third, a scenario recognition model was developed, comprising context feature processing, visual feature handling, and multimodal feature fusion. For illustration, a smart home environment was built, and twenty-six participants were recruited to perform various home activities. Integral sensors were used to collect environmental context data, and video data was captured simultaneously, both of which jointly form a multimodal dataset. Results demonstrated that the VCFSR model achieved an average accuracy of 98.1 %, outperforming traditional machine learning models such as decision trees and support vector machines. This method was then employed for fine-grained human behavior sequence prediction tasks, showing good performance in predicting behavior sequences across all scenarios constructed in this study. Furthermore, the results of ablation experiments revealed that the multimodal feature fusion method increased the average accuracy by at least 1.8 % compared to single-modality data-driven methods.</div><div>This novel approach to user behavior modeling simultaneously handles the relationship threads across scenarios and the rich details provided by visual data, paving the way for advanced intelligent services in complex interactive environments such as smart homes and hospitals.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103161"},"PeriodicalIF":8.0,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143368501","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

The human-centric framework integrating knowledge distillation architecture with fine-tuning mechanism for equipment health monitoring

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-06 DOI: 10.1016/j.aei.2025.103167

Jr-Fong Dang , Tzu-Li Chen , Hung-Yi Huang

Human-centricity serves as the cornerstone of the evolution of manufacturing into Industry 5.0. Accordingly, modern manufacturing prioritizes both the well-being of human workers and their collaboration with production systems. Successful systems would be user-friendly and market-appropriate, effectively identifying and analyzing user needs. This study aims to integrate user requirements into the framework for equipment health monitoring (EHM). The proposed framework addresses issues related to insufficient training samples and variable-length data by combining an encoder-decoder architecture with an attention mechanism and a conditional generative adversarial network (EDA-CGAN). Furthermore, the authors utilize a teacher-student network to reduce model complexity through knowledge distillation (KD). To prevent negative knowledge distillation, this study incorporates user requirements using Kullback-Leibler divergence (KLD) to determine whether the teacher model would be fine-tuned. Consequently, we employ the explainable AI (XAI) to provide a clear and understandable explanation for the prediction results. Thus, the proposed human-centric EHM consisting of four modules: (i) the data augmentation (ii) the fine-tuning mechanism (ii) the equipment health prediction model (iv) the explainable AI (XAI). The authors employ these methods to uncover new research insights that are vital for advancing the methodological innovation within the proposed framework. To evaluate model performance, this study conducts an empirical investigation to illustrate the capability and practicality of the proposed framework. The results indicate that our algorithm outperforms existing machine learning models, enabling the implementation of the proposed framework in the real-world manufacturing environment to maintain equipment health.

以人为本是制造业向工业 5.0 演进的基石。因此，现代制造业既要优先考虑人类工人的福祉，也要考虑他们与生产系统的协作。成功的系统应具有用户友好性和市场适应性，并能有效识别和分析用户需求。本研究旨在将用户需求纳入设备健康监测（EHM）框架。建议的框架通过将编码器-解码器架构与注意力机制和条件生成对抗网络（EDA-CGAN）相结合，解决了与训练样本不足和数据长度可变相关的问题。此外，作者还利用师生网络，通过知识提炼（KD）来降低模型的复杂性。为防止出现消极的知识蒸馏，本研究利用库尔贝-莱布勒发散（KLD）将用户需求纳入其中，以确定是否对教师模型进行微调。因此，我们采用了可解释人工智能（XAI），为预测结果提供清晰易懂的解释。因此，所提出的以人为本的 EHM 由四个模块组成：(i) 数据增强；(ii) 微调机制；(ii) 设备健康预测模型；(iv) 可解释的人工智能（XAI）。作者采用这些方法来揭示新的研究见解，这些见解对于推进拟议框架内的方法创新至关重要。为评估模型性能，本研究开展了一项实证调查，以说明拟议框架的能力和实用性。结果表明，我们的算法优于现有的机器学习模型，从而能够在现实世界的制造环境中实施所提出的框架，以维护设备健康。

{"title":"The human-centric framework integrating knowledge distillation architecture with fine-tuning mechanism for equipment health monitoring","authors":"Jr-Fong Dang , Tzu-Li Chen , Hung-Yi Huang","doi":"10.1016/j.aei.2025.103167","DOIUrl":"10.1016/j.aei.2025.103167","url":null,"abstract":"<div><div>Human-centricity serves as the cornerstone of the evolution of manufacturing into Industry 5.0. Accordingly, modern manufacturing prioritizes both the well-being of human workers and their collaboration with production systems. Successful systems would be user-friendly and market-appropriate, effectively identifying and analyzing user needs. This study aims to integrate user requirements into the framework for equipment health monitoring (EHM). The proposed framework addresses issues related to insufficient training samples and variable-length data by combining an encoder-decoder architecture with an attention mechanism and a conditional generative adversarial network (EDA-CGAN). Furthermore, the authors utilize a teacher-student network to reduce model complexity through knowledge distillation (KD). To prevent negative knowledge distillation, this study incorporates user requirements using Kullback-Leibler divergence (KLD) to determine whether the teacher model would be fine-tuned. Consequently, we employ the explainable AI (XAI) to provide a clear and understandable explanation for the prediction results. Thus, the proposed human-centric EHM consisting of four modules: (i) the data augmentation (ii) the fine-tuning mechanism (ii) the equipment health prediction model (iv) the explainable AI (XAI). The authors employ these methods to uncover new research insights that are vital for advancing the methodological innovation within the proposed framework. To evaluate model performance, this study conducts an empirical investigation to illustrate the capability and practicality of the proposed framework. The results indicate that our algorithm outperforms existing machine learning models, enabling the implementation of the proposed framework in the real-world manufacturing environment to maintain equipment health.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103167"},"PeriodicalIF":8.0,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143368497","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

Vehicle spatiotemporal distribution identification in low-light environment based on image enhancement and object detection

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-06 DOI: 10.1016/j.aei.2025.103165

Jie Zhang , Jiaqiang Peng , Xuan Kong , Shuo Wang , Jiexuan Hu

The spatiotemporal distribution of vehicles on roads and bridges is important for the operation and maintenance of transportation systems. The accuracy of vehicle identification is affected by the lighting conditions, especially low-light environments. This study proposes a vehicle spatiotemporal distribution identification method using image enhancement and object detection. First, the FP-ZeroDCE algorithm is used to enhance low-light images, which improves the brightness and contrast of images. Next, the enhanced images are input into the AFF-YOLO model to identify the spatiotemporal distribution of vehicles. Finally, the proposed method is validated using public datasets and tested in the field. The results indicate that the proposed method can enhance the quality of low-light images, with an increase in the Peak Signal-to-Noise Ratio by 8.257 dB, and improve the accuracy of vehicle detection, with an accuracy of 92.7 %. The proposed method is an effective means for identifying vehicle spatiotemporal distribution under low-light conditions.

{"title":"Vehicle spatiotemporal distribution identification in low-light environment based on image enhancement and object detection","authors":"Jie Zhang , Jiaqiang Peng , Xuan Kong , Shuo Wang , Jiexuan Hu","doi":"10.1016/j.aei.2025.103165","DOIUrl":"10.1016/j.aei.2025.103165","url":null,"abstract":"<div><div>The spatiotemporal distribution of vehicles on roads and bridges is important for the operation and maintenance of transportation systems. The accuracy of vehicle identification is affected by the lighting conditions, especially low-light environments. This study proposes a vehicle spatiotemporal distribution identification method using image enhancement and object detection. First, the FP-ZeroDCE algorithm is used to enhance low-light images, which improves the brightness and contrast of images. Next, the enhanced images are input into the AFF-YOLO model to identify the spatiotemporal distribution of vehicles. Finally, the proposed method is validated using public datasets and tested in the field. The results indicate that the proposed method can enhance the quality of low-light images, with an increase in the Peak Signal-to-Noise Ratio by 8.257 dB, and improve the accuracy of vehicle detection, with an accuracy of 92.7 %. The proposed method is an effective means for identifying vehicle spatiotemporal distribution under low-light conditions.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103165"},"PeriodicalIF":8.0,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143368502","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