Applied Soft Computing最新文献_第3页

Self-supervised Transformer for 3D point clouds completion and morphology evaluation of granular particle 三维点云补全及颗粒形态评价的自监督变压器

IF 7.2 1区计算机科学

Applied Soft Computing Pub Date : 2025-04-14 DOI: 10.1016/j.asoc.2025.113161

Haoran Zhang, Zhen-Yu Yin, Ning Zhang, Xiang Wang

{"title":"Self-supervised Transformer for 3D point clouds completion and morphology evaluation of granular particle","authors":"Haoran Zhang, Zhen-Yu Yin, Ning Zhang, Xiang Wang","doi":"10.1016/j.asoc.2025.113161","DOIUrl":"10.1016/j.asoc.2025.113161","url":null,"abstract":"<div><div>Determining the morphology characteristics of particles using 3D point cloud is promising and crucial for the quality inspection of granular materials. However, it remains challenging due to the cumbersome process and incomplete 3D point clouds obtained from laser scanning of particles. In this study, a novel intelligent method, named self-supervised transformer-based encoder and decoder model for granular materials (SSPoinTr-GM), is developed for the automatic completion of partially occluded 3D point clouds and morphology characteristics evaluation. The complete cloud points of 100 cobble and 100 gravel particles are first scanned to establish a benchmark 3D point cloud dataset. To form partial point clouds for training, the complete point cloud is divided into global seed points by the farthest point sampling (FPS) method and the local cloud points around each seed point by the k-nearest neighbor method. Then, the seed points and their local cloud points are randomly removed to generate partial cloud points as input, training the encoder and decoder in a self-supervised way with the original complete point cloud as ground truth. Experiments are conducted to validate the effectiveness of the novel method compared with four existing completion baselines based on the 3D point cloud dataset. The results indicate that the CD1 loss of the completed particles by the proposed method is, on average, 49.05 % lower than that of existing baselines. Additionally, the error rate of the calculated morphology characteristics of the completed particles is, on average, 66.06 % lower than that of the partial point clouds.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"176 ","pages":"Article 113161"},"PeriodicalIF":7.2,"publicationDate":"2025-04-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143834638","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

An intelligent feature selection-based fake news detection model for pandemic situation with optimal attention based multiscale densenet with long short-term memory layer 基于最优关注的具有长短期记忆层的多尺度密集神经网络的疫情假新闻智能检测模型

IF 7.2 1区计算机科学

Applied Soft Computing Pub Date : 2025-04-14 DOI: 10.1016/j.asoc.2025.113158

V Rathinapriya , J. Kalaivani

{"title":"An intelligent feature selection-based fake news detection model for pandemic situation with optimal attention based multiscale densenet with long short-term memory layer","authors":"V Rathinapriya , J. Kalaivani","doi":"10.1016/j.asoc.2025.113158","DOIUrl":"10.1016/j.asoc.2025.113158","url":null,"abstract":"<div><div>Fake news has recently used the strength and scope of online networking sites to efficiently propagate misinformation, eroding confidence in the press and journalism while also manipulating public perceptions and emotions. However, much information appearing on the Internet is dubious and even intended to mislead. Some fake news is so similar to the real ones that it is difficult for humans to identify them. Therefore, Fake News Detection (FND) needs to develop effectual models to overcome the existing challenges. So, in this paper, a novel deep-learning approach is developed for the recognition of fake news in pandemic situations. Initially, text data are collected from benchmark resources related to the pandemic situation and provided to the pre-processing stage. Then, the obtained pre-processed data is inputted into the feature extraction process. Here, the features are extracted using glove embedding, Bidirectional Encoder Representations from Transformers (BERT), and Term Frequency Inverse Document Frequency (TFIDF). Later, the extracted features are taken to the fused optimal weighted feature selection, and the weights are optimized using the Updated Random Variable-based Artificial Rabbits Optimization (URV-ARO), leveraging the Artificial Rabbits Optimization (ARO). The attained optimal weighted features are then given to the classification process. In the classification phase, the fake news is classified with the help of Optimal Attention-based Multiscale Densenet with Long Short-Term Memory layer (OAMDNet-LSTM). Moreover, parameters in DenseNetand LSTM are tuned by developed URV-ARO. Optimizing parameters in the DenseNetand LSTM helps fine-tune the model to achieve higher accuracy in distinguishing between genuine and fake news. The effectiveness of the proposed model is validated with conventional approaches to showcase the effectiveness of others.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"176 ","pages":"Article 113158"},"PeriodicalIF":7.2,"publicationDate":"2025-04-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143854812","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Towards energy-efficient Robotic Mobile Fulfillment System: Hybrid agent-based simulation with DEA-based surrogate machine learning 面向节能机器人移动履约系统：基于agent的混合仿真与基于dea的代理机器学习

IF 7.2 1区计算机科学

Applied Soft Computing Pub Date : 2025-04-12 DOI: 10.1016/j.asoc.2025.113141

Zakka Ugih Rizqi , Shuo-Yan Chou , Adi Dharma Oscar

{"title":"Towards energy-efficient Robotic Mobile Fulfillment System: Hybrid agent-based simulation with DEA-based surrogate machine learning","authors":"Zakka Ugih Rizqi , Shuo-Yan Chou , Adi Dharma Oscar","doi":"10.1016/j.asoc.2025.113141","DOIUrl":"10.1016/j.asoc.2025.113141","url":null,"abstract":"<div><div>The rapid growth of retail e-commerce has increased demand for warehouses to handle large volumes and diverse SKUs. To meet these demands, Robotic Mobile Fulfillment System (RMFS) is widely adopted. However, the automation in RMFS significantly raises energy consumption. The challenge is that the dynamic complexity of RMFS operations poses a major challenge in improving energy efficiency. This research proposes a hybrid optimization model to optimize traffic policy, routing strategy, number of robots, and robot’s max speed for reducing energy consumption while maintaining throughput rate. We first formulated a realistic RMFS energy consumption. A new priority rule for traffic policy was then proposed to reduce unnecessary stoppages. Two routing strategies namely Aisles Only and Underneath Pod were evaluated. Agent-based model was finally developed. Simulation experiment shows that the proposed priority rule reduces energy consumption by 3.41 % and increases the throughput by 26.07 % compared to FCFS. Further, global optimization was performed by first unifying conflicting objectives into a single-efficiency objective using Data Envelopment Analysis. Surrogate-based machine learning was then fitted and optimized via metaheuristic algorithm. The near-optimal configuration for RMFS was achieved by implementing the Priority Rule as traffic policy, Underneath Pod as routing strategy, 26 as number of robots, and 1.372 m/s as max speed. ANOVA reveals that the number of robots is the most influential factors to overall RMFS performance.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"176 ","pages":"Article 113141"},"PeriodicalIF":7.2,"publicationDate":"2025-04-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143829448","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A local and global multi-head relation self-attention network for fault diagnosis of rotating machinery under noisy environments 基于局部和全局多头关系自关注网络的噪声环境下旋转机械故障诊断

IF 7.2 1区计算机科学

Applied Soft Computing Pub Date : 2025-04-11 DOI: 10.1016/j.asoc.2025.113138

Yiwei Cheng , Xinnuo Lin , Wenwei Liu , Ming Zeng , Pengfei Liang

{"title":"A local and global multi-head relation self-attention network for fault diagnosis of rotating machinery under noisy environments","authors":"Yiwei Cheng , Xinnuo Lin , Wenwei Liu , Ming Zeng , Pengfei Liang","doi":"10.1016/j.asoc.2025.113138","DOIUrl":"10.1016/j.asoc.2025.113138","url":null,"abstract":"<div><div>Fault diagnosis under noisy environments (FDUNE) for rotating machinery is a highly challenging task. In recent years, deep learning models have become research hotspots in the field of FDUNE. However, the existing FDUNE approaches suffer from a limitation that insufficient consideration of both local and global features in the feature extraction process leads to unsatisfactory diagnostic performance. In this paper, a local and global multi-head relation self-attention network (LGMHRSANet) is proposed to improve the diagnostic accuracy of rotating machinery under noisy environments, which integrates convolution and self-attention into the transformer form, enabling it to capture local features and global long-range temporal features from vibration signals. Two experimental cases on rolling bearings and gearboxes are implemented to verify the effectiveness of LGMHRSANet under noisy environments. Experimental results demonstrate that LGMHRSANet has superior diagnostic performance compared to other deep learning models, regardless of whether it is in a non-noise environment, or a strong noise environment. In addition, the adaptive performance analysis in the variable noise domain indicates that LGMHRSANet has good robustness in noisy environments.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"176 ","pages":"Article 113138"},"PeriodicalIF":7.2,"publicationDate":"2025-04-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143834637","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Parallel graph neural architecture search optimization with incomplete features 具有不完全特征的并行图神经结构搜索优化

IF 7.2 1区计算机科学

Applied Soft Computing Pub Date : 2025-04-11 DOI: 10.1016/j.asoc.2025.113068

Haitao Yang , Zhaowei Liu , Dong Yang , Lihong Wang

{"title":"Parallel graph neural architecture search optimization with incomplete features","authors":"Haitao Yang , Zhaowei Liu , Dong Yang , Lihong Wang","doi":"10.1016/j.asoc.2025.113068","DOIUrl":"10.1016/j.asoc.2025.113068","url":null,"abstract":"<div><div>Graph neural networks (GNNs) have shown remarkable success in many fields. However, the results of different model architectures for different scenarios can be very different. Designing effective neural architectures requires a great deal of specialized knowledge, which limits the application of GNNs models. In recent years, graph neural architecture search (GNAS) has attracted widespread attention. GNAS selects the GNNs structure in predefined search space using a suitable search algorithm. The search direction is constrained based on the evaluation made by the estimation strategy. Traditional GNAS methods suffer from long search times, difficulty in parameter selection, and high sensitivity to data quality. When feature information is missing, the candidate architectures explored during the search process cannot obtain complete feature information, which significantly reduces the accuracy of GNAS. To tackle these challenges, we propose a novel optimization framework for parallel graph neural architecture search, named AutoPGO. In AutoPGO, we complement the features based on a feature propagation algorithm generated by minimizing the Dirichlet energy function, improve the search algorithm using the mutation decay strategy and complete the optimization of the parameters using the Bayesian optimization method. Experimental results show that AutoPGO has good performance and some degree of robustness.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"176 ","pages":"Article 113068"},"PeriodicalIF":7.2,"publicationDate":"2025-04-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143845253","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Oscillating activation functions can improve the performance of convolutional neural networks 振荡激活函数可以提高卷积神经网络的性能

IF 7.2 1区计算机科学

Applied Soft Computing Pub Date : 2025-04-11 DOI: 10.1016/j.asoc.2025.113077

Mathew Mithra Noel , Arunkumar L. , Advait Trivedi , Praneet Dutta

{"title":"Oscillating activation functions can improve the performance of convolutional neural networks","authors":"Mathew Mithra Noel , Arunkumar L. , Advait Trivedi , Praneet Dutta","doi":"10.1016/j.asoc.2025.113077","DOIUrl":"10.1016/j.asoc.2025.113077","url":null,"abstract":"<div><div>Convolutional neural networks have been successful in solving many socially important and economically significant problems. Their ability to learn complex high-dimensional functions hierarchically can be attributed to the use of nonlinear activation functions. A key discovery that made training deep networks feasible was the adoption of the Rectified Linear Unit (ReLU) activation function to alleviate the vanishing gradient problem caused by using saturating activation functions. Since then, many improved variants of the ReLU activation have been proposed. However, a majority of activation functions used today are non-oscillatory and monotonically increasing due to their biological plausibility. This paper demonstrates that oscillatory activation functions can improve gradient flow and reduce network size. Two theorems on limits of non-oscillatory activation functions are presented. A new oscillatory activation function called Growing Cosine Unit(GCU) defined as <span><math><mrow><mi>C</mi><mrow><mo>(</mo><mi>z</mi><mo>)</mo></mrow><mo>=</mo><mi>z</mi><mi>⋅</mi><mo>cos</mo><mi>z</mi></mrow></math></span> that outperforms Sigmoids, Swish, Mish and ReLU on a variety of architectures and benchmarks is presented. The GCU activation has multiple zeros enabling single GCU neurons to have multiple hyperplanes in the decision boundary. This allows single GCU neurons to learn the XOR function without feature engineering. Extensive experimental comparison with 16 popular activation functions indicate that the GCU activation function significantly improves performance on CIFAR-10, CIFAR-100, Imagenette and the 1000 class ImageNet benchmarks.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113077"},"PeriodicalIF":7.2,"publicationDate":"2025-04-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143823666","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Selecting a health emergency strategy through large-scale multi-criteria decision-making based on intuitionistic fuzzy self-confidence data 基于直觉模糊自信数据的大规模多准则决策卫生应急策略选择

IF 7.2 1区计算机科学

Applied Soft Computing Pub Date : 2025-04-11 DOI: 10.1016/j.asoc.2025.113085

Priya Sharma , Mukesh Kumar Mehlawat , Pankaj Gupta , Shilpi Verma

{"title":"Selecting a health emergency strategy through large-scale multi-criteria decision-making based on intuitionistic fuzzy self-confidence data","authors":"Priya Sharma , Mukesh Kumar Mehlawat , Pankaj Gupta , Shilpi Verma","doi":"10.1016/j.asoc.2025.113085","DOIUrl":"10.1016/j.asoc.2025.113085","url":null,"abstract":"<div><div>In complex decision-making scenarios involving multiple stakeholders, the uncertainty and individual confidence of decision-makers (DMs) are crucial in determining the outcomes. A novel approach is proposed in this paper to improve decision-making processes within a large group of DMs operating under an “Intuitionistic Fuzzy Self-Confidence (IFN-SC)” setting. The research presents a hybrid clustering algorithm to categorize DMs based on their numerical similarities and psychological factors. A multi-objective nonlinear optimization problem is employed to determine the criteria weights in the IFN-SC environment when the weight vector is either partially or fully unknown. Using the max operator, we derive a single-objective nonlinear optimization problem, which is solved by the “Particle Swarm Optimization (PSO)” algorithm. Furthermore, extending the “Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS)” for the IFN-SC environment significantly enhances the model’s effectiveness in ranking alternatives. The study exemplified its capability in managing a large-scale decision-making problem based on health emergency strategy selection and presented various analyses highlighting its utility, adaptability, and robustness in practical situations.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"176 ","pages":"Article 113085"},"PeriodicalIF":7.2,"publicationDate":"2025-04-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143834639","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Photovoltaic parameter extraction through an adaptive differential evolution algorithm with multiple linear regression 光伏参数提取采用多元线性回归自适应差分进化算法

IF 7.2 1区计算机科学

Applied Soft Computing Pub Date : 2025-04-11 DOI: 10.1016/j.asoc.2025.113117

Bozhen Chen , Haibin Ouyang , Steven Li , Liqun Gao , Weiping Ding

{"title":"Photovoltaic parameter extraction through an adaptive differential evolution algorithm with multiple linear regression","authors":"Bozhen Chen , Haibin Ouyang , Steven Li , Liqun Gao , Weiping Ding","doi":"10.1016/j.asoc.2025.113117","DOIUrl":"10.1016/j.asoc.2025.113117","url":null,"abstract":"<div><div>Solar cells play a crucial role in generating clean, renewable energy. Accurate modeling of photovoltaic (PV) systems is essential for their development, and simulating their behaviors requires precise estimation of their parameters. However, many optimization methods exhibit high or unstable root mean square error (RMSE) due to local optima entrapment and parameter interdependence. To address these challenges, we propose MLR-DE, a novel hybrid approach that integrates adaptive differential evolution (DE) with multiple linear regression (MLR). The main innovation is to decompose the PV model into linear coefficients and non-linear functions, the latter being iteratively estimated using DE. By treating nonlinear function outputs as independent variables and known measured currents as dependent variables, linear coefficients are analytically solved through MLR. Additionally, we introduce a data-fusion-based parameter generation scheme to improve DE’s reliability by integrating historical crossover rates with estimated crossover rates. We validate MLR-DE through experiments across 11 PV configurations: 3 standard diode models and 8 environmental variants. The results demonstrate MLR-DE’s superiority in all tests. It achieves the lowest average RMSE compared to other algorithms, with standard deviations at or below 2e−16. In the Friedman test, MLR-DE ranked first with a score of 1.94, outperforming the second-place (3.72) and last-place (7.58) competitors. The convergence curve shows that MLR-DE achieves convergence in less than 3,000 function evaluations over standard models, with an average convergence time of less than 0.6 s.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"176 ","pages":"Article 113117"},"PeriodicalIF":7.2,"publicationDate":"2025-04-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143834640","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Application of multi-objective evolutionary algorithm based on transfer learning in sliding bearing 基于迁移学习的多目标进化算法在滑动轴承中的应用

IF 7.2 1区计算机科学

Applied Soft Computing Pub Date : 2025-04-11 DOI: 10.1016/j.asoc.2025.113111

Xuepeng Ren, Maocai Wang, Guangming Dai, Lei Peng

{"title":"Application of multi-objective evolutionary algorithm based on transfer learning in sliding bearing","authors":"Xuepeng Ren, Maocai Wang, Guangming Dai, Lei Peng","doi":"10.1016/j.asoc.2025.113111","DOIUrl":"10.1016/j.asoc.2025.113111","url":null,"abstract":"<div><div>In recent years, decomposition-based multi-objective evolutionary algorithms have gained increasing attention for solving complex optimization problems. However, existing weight vector adaptation methods often struggle to balance diversity and convergence. To address this issue, we propose a multi-objective evolutionary algorithm based on transfer learning (MOEA/D-TL), which integrates joint distribution adaptation (JDA) to coordinate the populations generated by genetic and differential operators. The key innovations of MOEA/D-TL include: (1) a dual-operator framework that leverages JDA to integrate the strengths of both operators; (2) auxiliary population labeling using Pareto dominance, leveraging JDA’s characteristics; and (3) sparsity-driven adaptive weight vector adjustment to refine population distribution. Extensive experiments on 44 benchmark problems demonstrate that MOEA/D-TL outperforms nine state-of-the-art algorithms, achieving a 42%–60% improvement across three performance metrics. When applied to the optimization of sliding bearings with conflicting objectives (load capacity, heat generation, and friction coefficient), MOEA/D-TL yields solutions with broader distribution and improved uniformity compared to seven other algorithms. These results validate the algorithm’s capability to balance diversity and convergence effectively.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"176 ","pages":"Article 113111"},"PeriodicalIF":7.2,"publicationDate":"2025-04-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143829442","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Defect recognition network for optical fiber cables based on feature information compensation 基于特征信息补偿的光纤电缆缺陷识别网络

IF 7.2 1区计算机科学

Applied Soft Computing Pub Date : 2025-04-11 DOI: 10.1016/j.asoc.2025.113139

Shao-Kai Zheng, Sheng-Su Ni, Peng Yan, Hao Wang, Dao-Lei Wang

{"title":"Defect recognition network for optical fiber cables based on feature information compensation","authors":"Shao-Kai Zheng, Sheng-Su Ni, Peng Yan, Hao Wang, Dao-Lei Wang","doi":"10.1016/j.asoc.2025.113139","DOIUrl":"10.1016/j.asoc.2025.113139","url":null,"abstract":"<div><div>The occurrence of electrical corrosion defects in ADSS optical fiber cables presents a significant challenge to the reliable operation of communication lines. Despite the importance of this issue, there has been limited research on accurately detecting electrical corrosion defects in recent years. Moreover, existing defect detection algorithms for industrial issues, such as electrical corrosion in ADSS optical fiber cables, are prone to feature information loss. To address this, we propose an improved Feature Compensation You Only Look Once (FC-YOLO) algorithm for effective detection of electrical corrosion defects in optical cables. First, we proposed the Feature Information Compensated Fusion Network (FICFN), which compensates for fusion features, mitigates the loss of defect information during cross-layer fusion, and enhances feature fusion. Second, an auxiliary training head is integrated into the head network, improving the information expression capability of the FICFN. Finally, an Efficient Local Attention (ELA) mechanism is incorporated into the neck network to boost the localization capabilities of the FICFN. To evaluate the efficacy of the proposed FC-YOLO, we conducted comparison experiments using different mainstream algorithms on both the ADSS electrical corrosion defects dataset and the NEU-DET dataset. Results from the ADSS dataset show that, compared to the YOLOv10s algorithm, the proposed algorithm achieves a 4.7 % increase in mean average precision (mAP@50), reaching 90.2 %, and a 4.1 % improvement in mAP@50–95. These enhancements meet the specifications required for power inspection. On the NEU-DET dataset, the algorithm improved mAP@50 and mAP@50–95 by 8.0 % and 6.1 %, respectively, demonstrating its adaptability for industrial defect detection tasks.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"176 ","pages":"Article 113139"},"PeriodicalIF":7.2,"publicationDate":"2025-04-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143829446","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0