Engineering Science and Technology-An International Journal-Jestech最新文献_第10页

A novel mixed Rayleigh distribution model using PID based search algorithm for wind energy applications 基于PID搜索算法的混合瑞利分布模型在风能应用中的应用

IF 5.4 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Engineering Science and Technology-An International Journal-Jestech

Pub Date : 2025-11-21 DOI: 10.1016/j.jestch.2025.102239

Hilmi Aygün , Bayram Köse

Accurate modeling of wind speed distributions is a critical prerequisite for reliable wind energy assessment, system optimization, and long-term performance prediction. Conventional probability distribution functions exhibit notable deviations between the observed and estimated wind speed frequency distributions, indicating their limited capability in capturing the actual variability of wind regimes. To address this gap, this study introduces, for the first time in the wind energy domain, the application of a Mixed Rayleigh distribution in combination with a PID-based metaheuristic optimization algorithm (PSA) for parameter estimation. The proposed approach was tested at three measurement stations: Karaburun, Mersinkoy, and Gelibolu, using extensive wind speed datasets. Comparative analyses were conducted between PSA based Rayleigh, Mixed Rayleigh, and Weibull models, alongside conventional Moment and Maximum Likelihood methods. The proposed model achieved the lowest Sum Square Error (SSE) (0.0016) and Root Mean Square Error (RMSE) (0.0091) in Karaburun, the lowest SSE (0.0014) and RMSE (0.0075) in Gelibolu, and consistently high determination coefficients (R² ≈ 0.9999) across all regions. Additionally, the model yielded the lowest Mean Absolute Percentage Error (MAPE) based on Wind Power Density (WPD) (4.11 %) in Mersinköy and relatively low MAPE values based on Average Wind Speed (3.74 % and 3.26 %) in Karaburun and Mersinköy, respectively. In particular, the Mixed Rayleigh model demonstrated superior flexibility, resulting in improved fitting accuracy and reduced estimation errors. Overall, the findings highlight the methodological novelty and practical potential of combining hybrid distribution functions with advanced optimization algorithms.

风速分布的准确建模是可靠的风能评估、系统优化和长期性能预测的关键先决条件。传统的概率分布函数在观测到的风速频率分布和估计的风速频率分布之间存在显著的偏差，表明它们在捕捉风况的实际变异性方面的能力有限。为了解决这一差距，本研究首次在风能领域引入了混合瑞利分布与基于pid的元启发式优化算法（PSA）相结合的参数估计应用。该方法在Karaburun、Mersinkoy和Gelibolu三个观测站进行了测试，使用了大量的风速数据集。在基于PSA的瑞利、混合瑞利和威布尔模型以及传统的矩和最大似然方法之间进行了比较分析。该模型在卡拉布润的和方误差（SSE）（0.0016）和均方根误差（RMSE）（0.0091）最低，在格里博卢的SSE（0.0014）和均方根误差（RMSE）（0.0075）最低，并且在所有地区都具有较高的决定系数（R2≈0.9999）。此外，该模型基于风力密度（WPD）的平均绝对百分比误差（MAPE）在Mersinköy最低（4.11%），而基于平均风速的MAPE值在卡拉布伦和Mersinköy相对较低（3.74%和3.26%）。特别是，混合瑞利模型显示出优越的灵活性，从而提高了拟合精度，减少了估计误差。总的来说，这些发现突出了将混合分布函数与先进的优化算法相结合的方法的新颖性和实际潜力。

{"title":"A novel mixed Rayleigh distribution model using PID based search algorithm for wind energy applications","authors":"Hilmi Aygün , Bayram Köse","doi":"10.1016/j.jestch.2025.102239","DOIUrl":"10.1016/j.jestch.2025.102239","url":null,"abstract":"<div><div>Accurate modeling of wind speed distributions is a critical prerequisite for reliable wind energy assessment, system optimization, and long-term performance prediction. Conventional probability distribution functions exhibit notable deviations between the observed and estimated wind speed frequency distributions, indicating their limited capability in capturing the actual variability of wind regimes. To address this gap, this study introduces, for the first time in the wind energy domain, the application of a Mixed Rayleigh distribution in combination with a PID-based metaheuristic optimization algorithm (PSA) for parameter estimation. The proposed approach was tested at three measurement stations: Karaburun, Mersinkoy, and Gelibolu, using extensive wind speed datasets. Comparative analyses were conducted between PSA based Rayleigh, Mixed Rayleigh, and Weibull models, alongside conventional Moment and Maximum Likelihood methods. The proposed model achieved the lowest Sum Square Error (SSE) (0.0016) and Root Mean Square Error (RMSE) (0.0091) in Karaburun, the lowest SSE (0.0014) and RMSE (0.0075) in Gelibolu, and consistently high determination coefficients (R<sup>2</sup> ≈ 0.9999) across all regions. Additionally, the model yielded the lowest Mean Absolute Percentage Error (MAPE) based on Wind Power Density (WPD) (4.11 %) in Mersinköy and relatively low MAPE values based on Average Wind Speed (3.74 % and 3.26 %) in Karaburun and Mersinköy, respectively. In particular, the Mixed Rayleigh model demonstrated superior flexibility, resulting in improved fitting accuracy and reduced estimation errors. Overall, the findings highlight the methodological novelty and practical potential of combining hybrid distribution functions with advanced optimization algorithms.</div></div>","PeriodicalId":48609,"journal":{"name":"Engineering Science and Technology-An International Journal-Jestech","volume":"72 ","pages":"Article 102239"},"PeriodicalIF":5.4,"publicationDate":"2025-11-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579476","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

DC-PFL: A dynamic clustering-based personalized federated learning method for human activity recognition DC-PFL：一种基于动态聚类的人类活动识别个性化联邦学习方法

IF 5.4 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Engineering Science and Technology-An International Journal-Jestech

Pub Date : 2025-11-20 DOI: 10.1016/j.jestch.2025.102230

Xiaoxu Wen , Yan Wang , Menghao Yuan , Aihui Wang , Ge Zheng , Hongnian Yu , Lin Meng

Human Activity Recognition (HAR) is essential in pervasive computing, healthcare, and human–computer interaction, where accurate interpretation of motion data underpins intelligent decision-making. Federated Learning (FL) enables privacy-preserving model training across distributed clients without sharing raw data, but suffers from degraded performance under Non-Independent and Identically Distributed (Non-IID) data, a common challenge in HAR due to user diversity and device heterogeneity. To address this, Personalized Federated Learning (PFL) introduces client-specific modeling, often via clustering. However, most existing approaches adopt static clustering strategies, lacking adaptability to dynamic changes in client data distributions. In this work, we propose DC-PFL, a Dynamic Clustering-based Personalized Federated Learning framework that performs round-wise client clustering using lightweight statistical features, like Average Peak Frequency (APF), percentiles, and Median Absolute Deviation (MAD) derived from local model parameters. This design ensures efficient and privacy-preserving similarity estimation across clients. By dynamically adjusting clusters during training, DC-PFL enables fine-grained personalization, better generalization, and improved robustness to Non-IID conditions. Experimental results on HAR benchmarks demonstrate that DC-PFL achieves superior performance in both accuracy and convergence speed compared to existing methods, including FedCHAR and standard FL baselines, validating its effectiveness in real-world federated HAR scenarios.

人类活动识别（HAR）在普适计算、医疗保健和人机交互中是必不可少的，在这些领域，对运动数据的准确解释是智能决策的基础。联邦学习（FL）支持在不共享原始数据的情况下跨分布式客户端进行隐私保护模型训练，但在非独立和同分布（Non-IID）数据下性能下降，这是HAR中由于用户多样性和设备异构性而面临的常见挑战。为了解决这个问题，个性化联邦学习（PFL）通常通过集群引入了特定于客户端的建模。然而，大多数现有方法采用静态聚类策略，缺乏对客户机数据分布动态变化的适应性。在这项工作中，我们提出了DC-PFL，这是一个基于动态聚类的个性化联邦学习框架，它使用轻量级统计特征（如平均峰值频率（APF），百分位数和中位数绝对偏差（MAD））来执行round-wise客户端聚类。这种设计确保了客户端之间高效且保护隐私的相似性估计。通过在训练过程中动态调整聚类，DC-PFL可以实现细粒度个性化、更好的泛化，并提高对非iid条件的鲁棒性。HAR基准测试的实验结果表明，与现有方法（包括FedCHAR和标准FL基线）相比，DC-PFL在精度和收敛速度方面都具有优越的性能，验证了其在真实联邦HAR场景中的有效性。

{"title":"DC-PFL: A dynamic clustering-based personalized federated learning method for human activity recognition","authors":"Xiaoxu Wen , Yan Wang , Menghao Yuan , Aihui Wang , Ge Zheng , Hongnian Yu , Lin Meng","doi":"10.1016/j.jestch.2025.102230","DOIUrl":"10.1016/j.jestch.2025.102230","url":null,"abstract":"<div><div>Human Activity Recognition (HAR) is essential in pervasive computing, healthcare, and human–computer interaction, where accurate interpretation of motion data underpins intelligent decision-making. Federated Learning (FL) enables privacy-preserving model training across distributed clients without sharing raw data, but suffers from degraded performance under Non-Independent and Identically Distributed (Non-IID) data, a common challenge in HAR due to user diversity and device heterogeneity. To address this, Personalized Federated Learning (PFL) introduces client-specific modeling, often via clustering. However, most existing approaches adopt static clustering strategies, lacking adaptability to dynamic changes in client data distributions. In this work, we propose DC-PFL, a Dynamic Clustering-based Personalized Federated Learning framework that performs round-wise client clustering using lightweight statistical features, like Average Peak Frequency (APF), percentiles, and Median Absolute Deviation (MAD) derived from local model parameters. This design ensures efficient and privacy-preserving similarity estimation across clients. By dynamically adjusting clusters during training, DC-PFL enables fine-grained personalization, better generalization, and improved robustness to Non-IID conditions. Experimental results on HAR benchmarks demonstrate that DC-PFL achieves superior performance in both accuracy and convergence speed compared to existing methods, including FedCHAR and standard FL baselines, validating its effectiveness in real-world federated HAR scenarios.</div></div>","PeriodicalId":48609,"journal":{"name":"Engineering Science and Technology-An International Journal-Jestech","volume":"72 ","pages":"Article 102230"},"PeriodicalIF":5.4,"publicationDate":"2025-11-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579364","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Electromagnetic Torque Prediction and Modeling of a Doubly Fed Induction Generator for Wind Energy Conversion Systems Using Machine Learning and Deep Learning Algorithms 基于机器学习和深度学习算法的风能转换系统双馈感应发电机电磁转矩预测与建模

IF 5.4 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Engineering Science and Technology-An International Journal-Jestech

Pub Date : 2025-11-19 DOI: 10.1016/j.jestch.2025.102227

M. Murat Tezcan , Ebru Efeoğlu

According to the 2023 Wind Energy Report published by the Global Energy Council, the total installed power of wind energy conversion systems worldwide is around 1 TW. In addition, in 2024 and the following years, an average annual increase of around 15% on this installed capacity is envisaged. This situation reveals the importance and rapid development of wind energy conversion systems (WECS) in renewable energy systems. Accordingly, during the design, modeling and production of AC generators at different power levels used in wind turbines, new generation design and modeling techniques are used in addition to classical modeling methods, and wind turbine generator R&D is developing rapidly. New design and optimization methods have begun to be used in the modeling and performance analysis of Double Fed Asynchronous Generators (DFIG), which are frequently used in the field for different output powers. Modeling DFIG with classical numerical modeling and FEA-based magnetic simulation programs is a time-consuming operation, especially in transient or dynamic analysis. Depending on the performance of the computer, obtaining a transient field distribution solution may take hours or even days to obtain iteration-based field distribution solutions that use the finite difference method as a reference. Therefore, machine learning and deep learning-based iterative optimization and prediction methods stand out as a powerful alternative.

In this study, electromagnetic torque values obtained through FEA-based simulations for three different DFIGs numerically modeled at medium power levels (250 kVA) with different winding materials (copper and aluminum) were used as reference. These torque curves were estimated using deep neural network algorithms based on K Nearest Neighbors (KNN), Support Vector Regression (SVR), Extra Tree (ET), Random Forest (RF), and Long Short-Term Memory (LSTM). Thus, the FEA results were compared with the predictions obtained from these algorithms, and the predictive performance of the algorithms was evaluated. The performances of the aforementioned algorithms in trainings and cross-validations were compared using R², MAE, and RMSE metrics. The LSTM-based deep neural network outperformed the other algorithms for electromagnetic torque estimation. Using this approach, R² values of 0.990, 0.976 and 0.994 were obtained for DFIG-1, DFIG-2 and DFIG-3 in cross-validation, respectively.

根据全球能源理事会发布的《2023年风能报告》，全球风能转换系统的总装机容量约为1太瓦。此外，在2024年和接下来的几年里，预计这一装机容量的平均年增长率约为15%。这种情况揭示了风能转换系统在可再生能源系统中的重要性和快速发展。因此，在风力发电机组所使用的不同功率级交流发电机的设计、建模和生产过程中，除了经典的建模方法外，还采用了新一代的设计和建模技术，风力发电机组的研发发展迅速。双馈异步发电机（DFIG）是电力领域中常用的一种具有不同输出功率的发电机，其建模和性能分析开始采用新的设计和优化方法。用经典的数值模拟和基于有限元的磁仿真程序对DFIG进行建模是一项耗时的工作，特别是在瞬态或动态分析中。根据计算机性能的不同，获得瞬态场分布解可能需要数小时甚至数天的时间才能获得以有限差分法为参考的基于迭代的场分布解。因此，机器学习和基于深度学习的迭代优化和预测方法作为一种强大的替代方案脱颖而出。本研究以三种不同绕组材料（铜和铝）的dfig在中等功率（250 kVA）下的电磁转矩数值模拟结果为参考。使用基于K近邻（KNN）、支持向量回归（SVR）、额外树（ET）、随机森林（RF）和长短期记忆（LSTM）的深度神经网络算法估计这些扭矩曲线。将有限元结果与算法的预测结果进行了比较，并对算法的预测性能进行了评价。使用R2、MAE和RMSE指标比较上述算法在训练和交叉验证中的性能。基于lstm的深度神经网络在电磁转矩估计方面优于其他算法。采用该方法交叉验证DFIG-1、DFIG-2和DFIG-3的R2值分别为0.990、0.976和0.994。

{"title":"Electromagnetic Torque Prediction and Modeling of a Doubly Fed Induction Generator for Wind Energy Conversion Systems Using Machine Learning and Deep Learning Algorithms","authors":"M. Murat Tezcan , Ebru Efeoğlu","doi":"10.1016/j.jestch.2025.102227","DOIUrl":"10.1016/j.jestch.2025.102227","url":null,"abstract":"<div><div>According to the 2023 Wind Energy Report published by the Global Energy Council, the total installed power of wind energy conversion systems worldwide is around 1 TW. In addition, in 2024 and the following years, an average annual increase of around 15% on this installed capacity is envisaged. This situation reveals the importance and rapid development of wind energy conversion systems (WECS) in renewable energy systems. Accordingly, during the design, modeling and production of AC generators at different power levels used in wind turbines, new generation design and modeling techniques are used in addition to classical modeling methods, and wind turbine generator R&D is developing rapidly. New design and optimization methods have begun to be used in the modeling and performance analysis of Double Fed Asynchronous Generators (DFIG), which are frequently used in the field for different output powers. Modeling DFIG with classical numerical modeling and FEA-based magnetic simulation programs is a time-consuming operation, especially in transient or dynamic analysis. Depending on the performance of the computer, obtaining a transient field distribution solution may take hours or even days to obtain iteration-based field distribution solutions that use the finite difference method as a reference. Therefore, machine learning and deep learning-based iterative optimization and prediction methods stand out as a powerful alternative.</div><div>In this study, electromagnetic torque values obtained through FEA-based simulations for three different DFIGs numerically modeled at medium power levels (250 kVA) with different winding materials (copper and aluminum) were used as reference. These torque curves were estimated using deep neural network algorithms based on K Nearest Neighbors (KNN), Support Vector Regression (SVR), Extra Tree (ET), Random Forest (RF), and Long Short-Term Memory (LSTM). Thus, the FEA results were compared with the predictions obtained from these algorithms, and the predictive performance of the algorithms was evaluated. The performances of the aforementioned algorithms in trainings and cross-validations were compared using R<sup>2</sup>, MAE, and RMSE metrics. The LSTM-based deep neural network outperformed the other algorithms for electromagnetic torque estimation. Using this approach, R<sup>2</sup> values of 0.990, 0.976 and 0.994 were obtained for DFIG-1, DFIG-2 and DFIG-3 in cross-validation, respectively.</div></div>","PeriodicalId":48609,"journal":{"name":"Engineering Science and Technology-An International Journal-Jestech","volume":"72 ","pages":"Article 102227"},"PeriodicalIF":5.4,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579477","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Opposition learning & PID-based grey wolf optimizer with swarm intelligence for improved load forecasting 基于群智能的基于对立学习和pid的灰狼优化算法改进负荷预测

IF 5.4 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Engineering Science and Technology-An International Journal-Jestech

Pub Date : 2025-11-17 DOI: 10.1016/j.jestch.2025.102237

Murat Akil , Ugur Yuzgec , Emrah Dokur

Electricity load forecasting helps grid operators to make informed decisions in terms of planning and managing demand response. Electric power companies utilize load forecasting to make optimal power management. Therefore, accurate forecasting of total electrical load in a region is of great importance. To overcome this problem, this paper proposes a multi-layer perceptron (MLP) hybrid model that contain Swarm Decomposition (SWD) aided Opposition Learning and proportional–integral–derivative based Grey Wolf Optimizer (OLPIDGWO) using historical electricity demand data in non-consecutive years. The dataset used for load forecasting includes loads with different characteristics. Empirical mode decomposition method and swarm decomposition are applied to the original data to decompose the data features. Then, MLP hybrid model is applied for each decomposed signal of the data as the load forecasting model. The advantages of the proposed hybrid model include a significant improvement in forecast accuracy and capture of local maxima. The advantage of the proposed hybrid model over other hybrid models and existing single forecasting models is also verified by error performance metrics. The result of the hybrid forecast model shows that the error performance metrics of MSE, RMSE, MAE and MAPE for the year 2020 are 35 MW, 0.591MW, 0.452MW and 1.47%, respectively, and the error performance metrics of MSE, RMSE, MAE and MAPE for the year 2022 are 22.6MW, 0.475MW, 0.367MW and 1.21%, respectively. The results reveal the SWD decomposition and GWO optimizer module of MLP improve the load prediction, and the proposed model outperforms other load prediction models.

电力负荷预测有助于电网运营商在规划和管理需求响应方面做出明智的决策。电力公司利用负荷预测进行电力优化管理。因此，准确预测某一地区的总电力负荷是十分重要的。为了克服这一问题，本文提出了一种多层感知器（MLP）混合模型，该模型包含群体分解（SWD）辅助的对立学习和基于比例-积分-导数的灰狼优化器（OLPIDGWO），使用非连续年的历史电力需求数据。用于负荷预测的数据集包括具有不同特征的负荷。对原始数据采用经验模态分解方法和群分解方法对数据特征进行分解。然后，对数据的每个分解信号采用MLP混合模型作为负荷预测模型。该混合模型的优点包括预测精度的显著提高和局部极大值的捕获。通过误差性能指标验证了混合预测模型相对于其他混合预测模型和现有单一预测模型的优越性。混合预测模型结果表明，2020年MSE、RMSE、MAE和MAPE的误差性能指标分别为35 MW、0.591MW、0.452MW和1.47%，2022年MSE、RMSE、MAE和MAPE的误差性能指标分别为22.6MW、0.475MW、0.367MW和1.21%。结果表明，MLP的SWD分解和GWO优化器模块改善了负荷预测，该模型优于其他负荷预测模型。

{"title":"Opposition learning & PID-based grey wolf optimizer with swarm intelligence for improved load forecasting","authors":"Murat Akil , Ugur Yuzgec , Emrah Dokur","doi":"10.1016/j.jestch.2025.102237","DOIUrl":"10.1016/j.jestch.2025.102237","url":null,"abstract":"<div><div>Electricity load forecasting helps grid operators to make informed decisions in terms of planning and managing demand response. Electric power companies utilize load forecasting to make optimal power management. Therefore, accurate forecasting of total electrical load in a region is of great importance. To overcome this problem, this paper proposes a multi-layer perceptron (MLP) hybrid model that contain Swarm Decomposition (SWD) aided Opposition Learning and proportional–integral–derivative based Grey Wolf Optimizer (OLPIDGWO) using historical electricity demand data in non-consecutive years. The dataset used for load forecasting includes loads with different characteristics. Empirical mode decomposition method and swarm decomposition are applied to the original data to decompose the data features. Then, MLP hybrid model is applied for each decomposed signal of the data as the load forecasting model. The advantages of the proposed hybrid model include a significant improvement in forecast accuracy and capture of local maxima. The advantage of the proposed hybrid model over other hybrid models and existing single forecasting models is also verified by error performance metrics. The result of the hybrid forecast model shows that the error performance metrics of MSE, RMSE, MAE and MAPE for the year 2020 are 35 MW, 0.591MW, 0.452MW and 1.47%, respectively, and the error performance metrics of MSE, RMSE, MAE and MAPE for the year 2022 are 22.6MW, 0.475MW, 0.367MW and 1.21%, respectively. The results reveal the SWD decomposition and GWO optimizer module of MLP improve the load prediction, and the proposed model outperforms other load prediction models.</div></div>","PeriodicalId":48609,"journal":{"name":"Engineering Science and Technology-An International Journal-Jestech","volume":"72 ","pages":"Article 102237"},"PeriodicalIF":5.4,"publicationDate":"2025-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579400","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 PatchNet-Attention based deep learning architecture for multi-type fabric defect classification in textile manufacturing and quality control 基于PatchNet-Attention的混合深度学习体系结构在纺织制造和质量控制中的多类型织物缺陷分类

IF 5.4 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Engineering Science and Technology-An International Journal-Jestech

Pub Date : 2025-11-17 DOI: 10.1016/j.jestch.2025.102231

Isil Karabey Aksakalli , Kubra Demir , Ozlem Sokmen

Accurate and timely detection of defects that may occur on fabric surfaces is a critical requirement for ensuring sustainable production quality in the textile industry. Due to human resource, time, and cost limitations, there is a growing interest in advanced image processing and deep learning-based automatic defect detection systems to improve the accuracy and efficiency of quality control in fabric manufacturing processes. In this study, we propose a novel hybrid PatchNet–Attention architecture that integrates patch-based feature extraction with an attention mechanism to improve defect localization and recognition. To evaluate the generalizability of the proposed architecture, its performance was tested on three public datasets using different class structures. Specifically, four classification scenarios were conducted: (i) classification with baseline models, (ii) patch-based classification, (iii) classification with a Convolutional Block Attention Module (CBAM)-enhanced model, and (iv) the proposed hybrid PatchNet–Attention architecture. Initially, 15 pre-trained Convolutional Neural Network (CNN) architectures were evaluated using transfer learning on the ZD001 dataset. The best-performing models, ResNet101V2 and Xception, were then selected as the foundation for constructing the hybrid PatchNet–Attention model. The experimental results demonstrate that configurations incorporating the attention mechanism consistently achieve the highest performance across all evaluated datasets. Specifically, the hybrid PatchNet–Attention model attained superior outcomes on the ZD001 dataset, with an F1-score of 99.15% and a Receiver Operating Characteristic–Area Under the Curve (ROC–AUC) of 99.5% in the three-class setting, and an F1-score of 97.28% with a ROC–AUC of 99.74% in the nine-subclass configuration. In the TILDA data set, the proposed model produced an F1 score of 87.74% and an ROC-AUC of 98.09%, while in the FDD data set it achieved an F1 score of 98.95% and a ROC-AUC of 99.50%. The source code of the proposed method can be accessed from the Data Availability section.

准确和及时地检测织物表面可能出现的缺陷是确保纺织工业可持续生产质量的关键要求。由于人力资源、时间和成本的限制，人们对先进的图像处理和基于深度学习的自动缺陷检测系统越来越感兴趣，以提高织物制造过程中质量控制的准确性和效率。在这项研究中，我们提出了一种新的混合补丁-注意力架构，该架构将基于补丁的特征提取与注意力机制相结合，以提高缺陷的定位和识别。为了评估所提出的体系结构的泛化性，在使用不同类结构的三个公共数据集上测试了其性能。具体来说，进行了四种分类场景：(i)使用基线模型进行分类，（ii）基于补丁的分类，（iii）使用卷积块注意力模块（CBAM）增强模型进行分类，以及（iv）提出的混合PatchNet-Attention架构。首先，在ZD001数据集上使用迁移学习对15个预训练的卷积神经网络（CNN）架构进行评估。然后选择表现最好的模型ResNet101V2和Xception作为构建混合PatchNet-Attention模型的基础。实验结果表明，包含注意机制的配置在所有评估的数据集上都一致地获得了最高的性能。具体而言，混合PatchNet-Attention模型在ZD001数据集上获得了更好的结果，在3类设置下的f1得分为99.15%，受试者工作特征曲线下面积（ROC-AUC）为99.5%；在9类配置下的f1得分为97.28%，ROC-AUC为99.74%。在TILDA数据集中，该模型的F1得分为87.74%，ROC-AUC为98.09%，而在FDD数据集中，该模型的F1得分为98.95%，ROC-AUC为99.50%。建议的方法的源代码可以从Data Availability部分访问。

{"title":"A hybrid PatchNet-Attention based deep learning architecture for multi-type fabric defect classification in textile manufacturing and quality control","authors":"Isil Karabey Aksakalli , Kubra Demir , Ozlem Sokmen","doi":"10.1016/j.jestch.2025.102231","DOIUrl":"10.1016/j.jestch.2025.102231","url":null,"abstract":"<div><div>Accurate and timely detection of defects that may occur on fabric surfaces is a critical requirement for ensuring sustainable production quality in the textile industry. Due to human resource, time, and cost limitations, there is a growing interest in advanced image processing and deep learning-based automatic defect detection systems to improve the accuracy and efficiency of quality control in fabric manufacturing processes. In this study, we propose a novel hybrid PatchNet–Attention architecture that integrates patch-based feature extraction with an attention mechanism to improve defect localization and recognition. To evaluate the generalizability of the proposed architecture, its performance was tested on three public datasets using different class structures. Specifically, four classification scenarios were conducted: (i) classification with baseline models, (ii) patch-based classification, (iii) classification with a Convolutional Block Attention Module (CBAM)-enhanced model, and (iv) the proposed hybrid PatchNet–Attention architecture. Initially, 15 pre-trained Convolutional Neural Network (CNN) architectures were evaluated using transfer learning on the ZD001 dataset. The best-performing models, ResNet101V2 and Xception, were then selected as the foundation for constructing the hybrid PatchNet–Attention model. The experimental results demonstrate that configurations incorporating the attention mechanism consistently achieve the highest performance across all evaluated datasets. Specifically, the hybrid PatchNet–Attention model attained superior outcomes on the ZD001 dataset, with an F1-score of 99.15% and a Receiver Operating Characteristic–Area Under the Curve (ROC–AUC) of 99.5% in the three-class setting, and an F1-score of 97.28% with a ROC–AUC of 99.74% in the nine-subclass configuration. In the TILDA data set, the proposed model produced an F1 score of 87.74% and an ROC-AUC of 98.09%, while in the FDD data set it achieved an F1 score of 98.95% and a ROC-AUC of 99.50%. The source code of the proposed method can be accessed from the Data Availability section.</div></div>","PeriodicalId":48609,"journal":{"name":"Engineering Science and Technology-An International Journal-Jestech","volume":"72 ","pages":"Article 102231"},"PeriodicalIF":5.4,"publicationDate":"2025-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579478","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Experimental and modeled response of hydraulic motors under pulsating flow 脉动流量下液压马达响应的实验与建模

IF 5.4 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Engineering Science and Technology-An International Journal-Jestech

Pub Date : 2025-11-17 DOI: 10.1016/j.jestch.2025.102234

Osama A. Gaheen , M.A. Aziz , Ernesto Benini , Mostafa E.A. Elsayed , Mostafa R. Rashed , Haitham Elshimy

This study investigates the dynamic response and performance of a hydraulic motor operating under controlled pulsating flow conditions. An experimental setup was developed incorporating a variable frequency pulse generator within an electro-hydraulic control circuit. Tests were conducted at inlet pressures of 20, 40, and 60 bar and pulsation frequencies of 2, 4, and 6 Hz. The results revealed that increasing flow pulsation frequency from 0 to 6 Hz significantly enhanced motor performance. At 60 bar, the motor speed increased from 71 RPM at 2 Hz to 114 RPM at 6 Hz, while torque rose from 6.11 kNm to 7.07 kNm. Similarly, increasing inlet pressure from 20 to 60 bar at 6 Hz improved speed from 67 to 114 RPM and torque from 3.65 to 7.07 kNm. At lower operating conditions (20 bar and 2 Hz), speed and pressure decreased by 60.74 % and 15 %, respectively, confirming the high sensitivity of motor output to pulsation parameters. Simulation results using Automation Studio closely matched the experimental findings, particularly at moderate frequencies and pressures with less than 4 % error. The developed empirical correlations accurately predicted motor speed and torque, with maximum deviations of ±10.49 %. The results demonstrate that controlling pulsation frequency provides an effective means of optimizing hydraulic motor performance, enhancing energy efficiency, and enabling dynamic regulation of speed and torque.

研究了液压马达在可控脉动工况下的动态响应和性能。在电液控制回路中加入变频脉冲发生器，建立了实验装置。试验在进口压力为20、40和60 bar，脉动频率为2、4和6 Hz的情况下进行。结果表明，将流量脉动频率从0增加到6 Hz可显著提高电机性能。在60 bar时，电机转速从2hz时的71 RPM增加到6hz时的114 RPM，扭矩从6.11 kNm增加到7.07 kNm。同样，在6赫兹下，将进口压力从20 bar增加到60 bar，速度从67 RPM提高到114 RPM，扭矩从3.65 kn提高到7.07 kn。在较低的工作条件下（20 bar和2 Hz），速度和压力分别下降了60.74%和15%，证实了电机输出对脉动参数的高灵敏度。使用Automation Studio的仿真结果与实验结果非常吻合，特别是在中等频率和压力下，误差小于4%。所建立的经验相关性能准确预测电机转速和转矩，最大偏差为±10.49%。结果表明，控制脉动频率是优化液压马达性能、提高能效、实现转速和转矩动态调节的有效手段。

{"title":"Experimental and modeled response of hydraulic motors under pulsating flow","authors":"Osama A. Gaheen , M.A. Aziz , Ernesto Benini , Mostafa E.A. Elsayed , Mostafa R. Rashed , Haitham Elshimy","doi":"10.1016/j.jestch.2025.102234","DOIUrl":"10.1016/j.jestch.2025.102234","url":null,"abstract":"<div><div>This study investigates the dynamic response and performance of a hydraulic motor operating under controlled pulsating flow conditions. An experimental setup was developed incorporating a variable frequency pulse generator within an electro-hydraulic control circuit. Tests were conducted at inlet pressures of 20, 40, and 60 bar and pulsation frequencies of 2, 4, and 6 Hz. The results revealed that increasing flow pulsation frequency from 0 to 6 Hz significantly enhanced motor performance. At 60 bar, the motor speed increased from 71 RPM at 2 Hz to 114 RPM at 6 Hz, while torque rose from 6.11 kNm to 7.07 kNm. Similarly, increasing inlet pressure from 20 to 60 bar at 6 Hz improved speed from 67 to 114 RPM and torque from 3.65 to 7.07 kNm. At lower operating conditions (20 bar and 2 Hz), speed and pressure decreased by 60.74 % and 15 %, respectively, confirming the high sensitivity of motor output to pulsation parameters. Simulation results using <em>Automation Studio</em> closely matched the experimental findings, particularly at moderate frequencies and pressures with less than 4 % error. The developed empirical correlations accurately predicted motor speed and torque, with maximum deviations of ±10.49 %. The results demonstrate that controlling pulsation frequency provides an effective means of optimizing hydraulic motor performance, enhancing energy efficiency, and enabling dynamic regulation of speed and torque.</div></div>","PeriodicalId":48609,"journal":{"name":"Engineering Science and Technology-An International Journal-Jestech","volume":"72 ","pages":"Article 102234"},"PeriodicalIF":5.4,"publicationDate":"2025-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

New ACCWOA algorithm based on the acceleration mechanism for solving engineering design problems and global optimization 基于加速机制的ACCWOA算法求解工程设计问题及全局优化

IF 5.4 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Engineering Science and Technology-An International Journal-Jestech

Pub Date : 2025-11-17 DOI: 10.1016/j.jestch.2025.102202

Ahmed Noori Algburi, Isa Avci

The Accelerated Whale Algorithm (ACCWOA) is a new version of the Whale Optimization Algorithm (WOA) that incorporates a velocity factor into the individuals. The WOA is a nature-inspired metaheuristic algorithm widely applied to solve engineering problems, which uses shrinking and spiral insertion movements to exploit the solution, and random search to discover new solutions. This algorithm suffers from slow convergence in early iterations, a lack of fine-tuned local search, stagnation in local optima, a lack of mechanisms to maintain diversity, and a lack of historical learning. To address these limitations, this work proposes an acceleration technique that mimics the rapid movement of whales as they pursue their prey. Acceleration technology utilizes the velocity equation to achieve accelerated convergence, enhanced exploitation, improved diversity retention, dynamic behavior, increased scalability and stability, and emergent memory capabilities. The algorithm was evaluated on standard benchmarks, IEEE CEC-2014 and CEC-2017 suites, and five engineering design problems: spring, three-bar truss, pressure vessel, welded beam, and cantilever beam. Results show that ACCWOA achieves rapid convergence, accurate solutions, and competitive efficiency compared to state-of-the-art methods.

加速鲸鱼算法（ACCWOA）是鲸鱼优化算法（WOA）的新版本，它将速度因素纳入个体。WOA是一种受自然启发的元启发式算法，广泛应用于解决工程问题，它使用收缩和螺旋插入运动来挖掘解，并使用随机搜索来发现新的解。该算法在早期迭代中收敛缓慢，缺乏微调的局部搜索，局部最优停滞，缺乏保持多样性的机制，以及缺乏历史学习。为了解决这些限制，这项工作提出了一种加速技术，模仿鲸鱼在追捕猎物时的快速运动。加速技术利用速度方程来实现加速收敛、增强开发、改进多样性保留、动态行为、增强可扩展性和稳定性以及应急存储能力。该算法在标准基准、IEEE CEC-2014和CEC-2017套件以及弹簧、三杆桁架、压力容器、焊接梁和悬臂梁五个工程设计问题上进行了评估。结果表明，与最先进的方法相比，ACCWOA实现了快速收敛，准确的解决方案和具有竞争力的效率。

{"title":"New ACCWOA algorithm based on the acceleration mechanism for solving engineering design problems and global optimization","authors":"Ahmed Noori Algburi, Isa Avci","doi":"10.1016/j.jestch.2025.102202","DOIUrl":"10.1016/j.jestch.2025.102202","url":null,"abstract":"<div><div>The Accelerated Whale Algorithm (ACCWOA) is a new version of the Whale Optimization Algorithm (WOA) that incorporates a velocity factor into the individuals. The WOA is a nature-inspired metaheuristic algorithm widely applied to solve engineering problems, which uses shrinking and spiral insertion movements to exploit the solution, and random search to discover new solutions. This algorithm suffers from slow convergence in early iterations, a lack of fine-tuned local search, stagnation in local optima, a lack of mechanisms to maintain diversity, and a lack of historical learning. To address these limitations, this work proposes an acceleration technique that mimics the rapid movement of whales as they pursue their prey. Acceleration technology utilizes the velocity equation to achieve accelerated convergence, enhanced exploitation, improved diversity retention, dynamic behavior, increased scalability and stability, and emergent memory capabilities. The algorithm was evaluated on standard benchmarks, IEEE CEC-2014 and CEC-2017 suites, and five engineering design problems: spring, three-bar truss, pressure vessel, welded beam, and cantilever beam. Results show that ACCWOA achieves rapid convergence, accurate solutions, and competitive efficiency compared to state-of-the-art methods.</div></div>","PeriodicalId":48609,"journal":{"name":"Engineering Science and Technology-An International Journal-Jestech","volume":"72 ","pages":"Article 102202"},"PeriodicalIF":5.4,"publicationDate":"2025-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579475","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Numerical and experimental investigation of the seismic behavior of rigid steel beam-to-hollow circular column connections using a steel box 钢箱梁-空心圆柱连接抗震性能的数值与试验研究

IF 5.4 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Engineering Science and Technology-An International Journal-Jestech

Pub Date : 2025-11-15 DOI: 10.1016/j.jestch.2025.102233

Abdulrahman Kamil Aakool AL-Hureshat , Mehrzad TahamouliRoudsari , AllahReza Moradi Garoosi , Mohsen Oghabi

Rigid beam-to-circular steel column connections have always attracted the attention of researchers and design codes due to their geometric complexity and construction challenges. In this study, the use of a steel box at the connection region is proposed for joining I-shaped sections to hollow circular columns. This method offers advantages such as simplified detailing and eliminates the need for continuity plates or doubler plates. The initial design of the experimental specimens was performed numerically, and then three specimens exhibiting superior performance were selected for experimental investigation. The experimental studies examined the effect of the configuration of column-to-box and beam-to-box stiffeners on their seismic performance. Subsequently, further numerical studies were conducted on the experimentally superior specimen. After validating the numerical model, the effect of the steel box thickness on the connection with various beam and column sections was evaluated. The aim of this part was to determine the minimum required thickness for the steel box, beyond which further increases would not affect the results. Additionally, the rigidity of the connection was also investigated. The experimental results indicated that the model with internal box stiffeners aligned with the beam flanges provided the highest strength, ductility, and elastic stiffness. Parametric analyses indicated that rigid behavior and full beam capacity can be ensured when the steel box thickness is at least 1.4 times the beam flange thickness.

钢梁与圆钢柱的刚性连接由于其几何复杂性和施工难度，一直受到研究人员和设计规范的关注。在本研究中，建议在连接区域使用钢箱将工字截面连接到空心圆形柱。这种方法的优点是简化了细节，并且不需要连续版或加倍版。通过数值方法对试验试件进行初步设计，选取性能较好的3个试件进行试验研究。试验研究考察了柱对箱和梁对箱加强筋结构对其抗震性能的影响。随后，对实验上较优的试样进行了进一步的数值研究。在验证了数值模型的基础上，评估了钢箱厚度对不同梁柱截面连接的影响。这部分的目的是确定钢盒所需的最小厚度，超过该厚度的进一步增加不会影响结果。此外，还对连接的刚度进行了研究。试验结果表明，内箱型加劲筋与梁缘对齐的模型具有最高的强度、延性和弹性刚度。参数分析表明，当钢箱厚度至少为梁翼缘厚度的1.4倍时，可以保证梁的刚性性能和全梁承载力。

{"title":"Numerical and experimental investigation of the seismic behavior of rigid steel beam-to-hollow circular column connections using a steel box","authors":"Abdulrahman Kamil Aakool AL-Hureshat , Mehrzad TahamouliRoudsari , AllahReza Moradi Garoosi , Mohsen Oghabi","doi":"10.1016/j.jestch.2025.102233","DOIUrl":"10.1016/j.jestch.2025.102233","url":null,"abstract":"<div><div>Rigid beam-to-circular steel column connections have always attracted the attention of researchers and design codes due to their geometric complexity and construction challenges. In this study, the use of a steel box at the connection region is proposed for joining I-shaped sections to hollow circular columns. This method offers advantages such as simplified detailing and eliminates the need for continuity plates or doubler plates. The initial design of the experimental specimens was performed numerically, and then three specimens exhibiting superior performance were selected for experimental investigation. The experimental studies examined the effect of the configuration of column-to-box and beam-to-box stiffeners on their seismic performance. Subsequently, further numerical studies were conducted on the experimentally superior specimen. After validating the numerical model, the effect of the steel box thickness on the connection with various beam and column sections was evaluated. The aim of this part was to determine the minimum required thickness for the steel box, beyond which further increases would not affect the results. Additionally, the rigidity of the connection was also investigated. The experimental results indicated that the model with internal box stiffeners aligned with the beam flanges provided the highest strength, ductility, and elastic stiffness. Parametric analyses indicated that rigid behavior and full beam capacity can be ensured when the steel box thickness is at least 1.4 times the beam flange thickness.</div></div>","PeriodicalId":48609,"journal":{"name":"Engineering Science and Technology-An International Journal-Jestech","volume":"72 ","pages":"Article 102233"},"PeriodicalIF":5.4,"publicationDate":"2025-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145529275","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 secure multi-hop routing algorithm based-on fuzzy logic for IoT communication 物联网通信中一种基于模糊逻辑的安全多跳路由算法

IF 5.4 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Engineering Science and Technology-An International Journal-Jestech

Pub Date : 2025-11-14 DOI: 10.1016/j.jestch.2025.102208

Tao Fu, Guoxin Han, Xuming Qin, Jinfang Li, Weiting Lin

The fast growth of the Internet of Things (IoT) into mission-critical applications requires secure and efficient routing protocols. Nevertheless, the resource limitations of IoT devices and their susceptibility to attacks require smart, dynamic solutions. To overcome these challenges, this paper introduces a new, safe, multi-hop routing algorithm that combines the use of fuzzy logic and reinforcement learning. We initially build a high-performance communication backbone over a Connected Dominating Set (CDS) to reduce network overhead. A fuzzy inference system then intelligently considers the possible paths using path energy, distance, and node credibility to choose the best path to transmit the data. A Q-learning model is used to dynamically evaluate the reliability of each node to provide security, and to identify and isolate malicious actors. Our algorithm is shown to be better in experimental results, with the ability to increase the ratio of packet delivery by up to 2.4 percent and at the same time lower the average energy consumption by about 6.53 percent of the current state-of-the-art protocols. These results demonstrate that our hybrid solution has a great potential to improve the reliability and safety of data routing in contemporary IoT networks.

物联网（IoT）向关键任务应用的快速发展需要安全高效的路由协议。然而，物联网设备的资源限制及其对攻击的易感性需要智能、动态的解决方案。为了克服这些挑战，本文引入了一种新的、安全的、多跳路由算法，该算法结合了模糊逻辑和强化学习的使用。我们首先在连接支配集（CDS）上构建高性能通信骨干，以减少网络开销。然后，模糊推理系统利用路径能量、距离和节点可信度来智能地考虑可能的路径，选择最佳路径来传输数据。使用q学习模型动态评估每个节点的可靠性以提供安全性，并识别和隔离恶意行为者。实验结果表明，我们的算法具有更好的性能，能够将分组传送率提高2.4%，同时将平均能耗降低约6.53%。这些结果表明，我们的混合解决方案在提高当代物联网网络中数据路由的可靠性和安全性方面具有巨大的潜力。

{"title":"A secure multi-hop routing algorithm based-on fuzzy logic for IoT communication","authors":"Tao Fu, Guoxin Han, Xuming Qin, Jinfang Li, Weiting Lin","doi":"10.1016/j.jestch.2025.102208","DOIUrl":"10.1016/j.jestch.2025.102208","url":null,"abstract":"<div><div>The fast growth of the Internet of Things (IoT) into mission-critical applications requires secure and efficient routing protocols. Nevertheless, the resource limitations of IoT devices and their susceptibility to attacks require smart, dynamic solutions. To overcome these challenges, this paper introduces a new, safe, multi-hop routing algorithm that combines the use of fuzzy logic and reinforcement learning. We initially build a high-performance communication backbone over a Connected Dominating Set (CDS) to reduce network overhead. A fuzzy inference system then intelligently considers the possible paths using path energy, distance, and node credibility to choose the best path to transmit the data. A Q-learning model is used to dynamically evaluate the reliability of each node to provide security, and to identify and isolate malicious actors. Our algorithm is shown to be better in experimental results, with the ability to increase the ratio of packet delivery by up to 2.4 percent and at the same time lower the average energy consumption by about 6.53 percent of the current state-of-the-art protocols. These results demonstrate that our hybrid solution has a great potential to improve the reliability and safety of data routing in contemporary IoT networks.</div></div>","PeriodicalId":48609,"journal":{"name":"Engineering Science and Technology-An International Journal-Jestech","volume":"72 ","pages":"Article 102208"},"PeriodicalIF":5.4,"publicationDate":"2025-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145529276","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Comparative analysis of autoencoder architectures for breast cancer detection using dynamic infrared thermography 动态红外热像仪检测乳腺癌的自编码器结构比较分析

IF 5.4 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Engineering Science and Technology-An International Journal-Jestech

Pub Date : 2025-11-10 DOI: 10.1016/j.jestch.2025.102225

Burcu Acar Demirci , Mehmet Engin , Erkan Zeki Engin

Breast cancer is the most diagnosed cancer among women worldwide. Early detection substantially improves treatment outcomes, especially when lesions are small and localized. Although conventional imaging modalities such as mammography, CT, MRI, and ultrasonography play a vital role in diagnosis, they often entail radiation exposure, high cost, and the use of contrast agents. These drawbacks have motivated increasing interest in non-invasive and cost-effective alternatives such as Infrared Thermal Imaging (ITI), which captures surface temperature variations that may indicate malignancy. This study proposes a novel ITI-based diagnostic framework integrating deep learning-driven feature extraction with conventional machine learning classifiers. Three autoencoder architectures—Vanilla Autoencoder (VanAE), Convolutional Autoencoder (CAE), and Variational Autoencoder (VAE)—were utilized to extract discriminative latent features from dynamic breast thermograms. The extracted features were subsequently classified using Support Vector Machine (SVM) and Random Forest (RF) algorithms. Experimental evaluation on a balanced DMR-IR dynamic dataset comprising 3,600 thermograms demonstrated that the CAE-SVM combination achieved the highest performance, reaching 92.28% accuracy, 89.11% sensitivity, 95.94% specificity, and a 92.26% F1-score. In addition to its superior classification performance, the CAE model exhibited the shortest training time, underscoring its potential for practical clinical implementation. Overall, the findings confirm the effectiveness of autoencoder-based architectures in learning meaningful representations directly from raw thermograms without relying on handcrafted or pre-trained features.

乳腺癌是全世界女性中诊断最多的癌症。早期发现可以显著改善治疗效果，特别是当病变很小且局部时。虽然传统的成像方式，如乳房x光检查、CT、MRI和超声检查在诊断中起着至关重要的作用，但它们通常需要辐射暴露、高成本和使用造影剂。这些缺点激发了人们对非侵入性和成本效益替代方案的兴趣，例如红外热成像（ITI），它可以捕获可能指示恶性肿瘤的表面温度变化。本研究提出了一种新的基于it的诊断框架，将深度学习驱动的特征提取与传统的机器学习分类器相结合。三种自编码器架构——香草自编码器（VanAE）、卷积自编码器（CAE）和变分自编码器（VAE）——被用于从动态乳房热像图中提取判别潜在特征。随后使用支持向量机（SVM）和随机森林（RF）算法对提取的特征进行分类。在包含3600张热图的平衡DMR-IR动态数据集上的实验评估表明，CAE-SVM组合达到了最高的性能，准确率为92.28%，灵敏度为89.11%，特异性为95.94%，f1评分为92.26%。除了其优越的分类性能外，CAE模型还具有最短的训练时间，强调了其在实际临床应用中的潜力。总的来说，研究结果证实了基于自动编码器的架构在直接从原始热图中学习有意义的表示方面的有效性，而不依赖于手工制作或预训练的特征。

{"title":"Comparative analysis of autoencoder architectures for breast cancer detection using dynamic infrared thermography","authors":"Burcu Acar Demirci , Mehmet Engin , Erkan Zeki Engin","doi":"10.1016/j.jestch.2025.102225","DOIUrl":"10.1016/j.jestch.2025.102225","url":null,"abstract":"<div><div>Breast cancer is the most diagnosed cancer among women worldwide. Early detection substantially improves treatment outcomes, especially when lesions are small and localized. Although conventional imaging modalities such as mammography, CT, MRI, and ultrasonography play a vital role in diagnosis, they often entail radiation exposure, high cost, and the use of contrast agents. These drawbacks have motivated increasing interest in non-invasive and cost-effective alternatives such as Infrared Thermal Imaging (ITI), which captures surface temperature variations that may indicate malignancy. This study proposes a novel ITI-based diagnostic framework integrating deep learning-driven feature extraction with conventional machine learning classifiers. Three autoencoder architectures—Vanilla Autoencoder (VanAE), Convolutional Autoencoder (CAE), and Variational Autoencoder (VAE)—were utilized to extract discriminative latent features from dynamic breast thermograms. The extracted features were subsequently classified using Support Vector Machine (SVM) and Random Forest (RF) algorithms. Experimental evaluation on a balanced DMR-IR dynamic dataset comprising 3,600 thermograms demonstrated that the CAE-SVM combination achieved the highest performance, reaching 92.28% accuracy, 89.11% sensitivity, 95.94% specificity, and a 92.26% F1-score. In addition to its superior classification performance, the CAE model exhibited the shortest training time, underscoring its potential for practical clinical implementation. Overall, the findings confirm the effectiveness of autoencoder-based architectures in learning meaningful representations directly from raw thermograms without relying on handcrafted or pre-trained features.</div></div>","PeriodicalId":48609,"journal":{"name":"Engineering Science and Technology-An International Journal-Jestech","volume":"72 ","pages":"Article 102225"},"PeriodicalIF":5.4,"publicationDate":"2025-11-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145529360","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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