Applied Soft Computing最新文献_第2页

Towards maritime safety with industrial multimodal models: Compliance reasoning from heterogeneous maritime data 用工业多式联运模型实现海上安全：基于异构海事数据的符合性推理

IF 6.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2026-01-27 DOI: 10.1016/j.asoc.2026.114701

Shen Wang , Kejun Wu , Renjie Qiao , Runtian Wang , Chengtao Cai

The advancement of Industrial Large Models offers significant potential for enhancing intelligence and automation within specialized industrial sectors. This paper addresses the maritime industry, a critical domain where ensuring safety and regulatory compliance demands robust understanding and reasoning over complex, heterogeneous data types. Traditional methods often lack the necessary interpretability and generalization capabilities for dynamic maritime environments. To overcome these limitations, we introduce a novel framework leveraging an Industrial Large Model approach for behavioral compliance reasoning. Our methodology utilizes a transformer-based large multimodal model adept at processing diverse industrial maritime data, specifically remote sensing imagery (Synthetic Aperture Radar) and structured navigational records (Automatic Identification System). The system features two core contributions: (1) a large-scale dataset constructed with diverse maritime scenarios, integrating high-resolution visual data, structured metadata, and natural language descriptions tailored for industrial maritime contexts; and (2) a transformer-based vision-language model fine-tuned on this dataset to jointly perform vessel detection, identification, and behavioral compliance assessment—tasks crucial for industrial maritime operations. To improve the model’s reasoning depth and align with the requirements for trustworthy Artificial Intelligence in industrial settings, we incorporate Chain-of-Thought prompting, facilitating transparent, step-by-step decision-making processes. Extensive experiments validate our model’s high accuracy, interpretability, and generalization capabilities in complex operational settings, significantly enhancing anomaly detection and supporting reliable maritime safety decisions. This study underscores the capability of Industrial Large Models to interpret complex multimodal industrial data for enhanced maritime situational awareness, providing a practical illustration of adapting foundational models for safety-critical and regulatory compliance tasks, thereby advancing the understanding and application of large model technology within specialized industrial sectors.

工业大型模型的进步为提高专业工业部门的智能化和自动化提供了巨大的潜力。本文讨论了海运业，这是一个关键领域，确保安全和法规遵从性需要对复杂的异构数据类型进行强大的理解和推理。对于动态海洋环境，传统方法往往缺乏必要的可解释性和泛化能力。为了克服这些限制，我们引入了一个利用工业大模型方法进行行为遵从性推理的新框架。我们的方法利用基于变压器的大型多模态模型，擅长处理各种工业海事数据，特别是遥感图像（合成孔径雷达）和结构化导航记录（自动识别系统）。该系统有两个核心贡献：(1)由不同海事场景构建的大规模数据集，集成了高分辨率视觉数据、结构化元数据和针对工业海事环境量身定制的自然语言描述；(2)基于变压器的视觉语言模型在此数据集上进行微调，以联合执行船舶检测、识别和行为合规评估，这些任务对工业海上作业至关重要。为了提高模型的推理深度，并与工业环境中值得信赖的人工智能的要求保持一致，我们结合了思维链提示，促进透明，逐步的决策过程。大量的实验验证了我们的模型在复杂操作环境中的高精度、可解释性和泛化能力，显著增强了异常检测并支持可靠的海上安全决策。本研究强调了工业大型模型解释复杂多模态工业数据的能力，以增强海上态势感知，为安全关键和监管合规任务调整基础模型提供了一个实际的例证，从而促进了大型模型技术在专业工业部门的理解和应用。

{"title":"Towards maritime safety with industrial multimodal models: Compliance reasoning from heterogeneous maritime data","authors":"Shen Wang , Kejun Wu , Renjie Qiao , Runtian Wang , Chengtao Cai","doi":"10.1016/j.asoc.2026.114701","DOIUrl":"10.1016/j.asoc.2026.114701","url":null,"abstract":"<div><div>The advancement of Industrial Large Models offers significant potential for enhancing intelligence and automation within specialized industrial sectors. This paper addresses the maritime industry, a critical domain where ensuring safety and regulatory compliance demands robust understanding and reasoning over complex, heterogeneous data types. Traditional methods often lack the necessary interpretability and generalization capabilities for dynamic maritime environments. To overcome these limitations, we introduce a novel framework leveraging an Industrial Large Model approach for behavioral compliance reasoning. Our methodology utilizes a transformer-based large multimodal model adept at processing diverse industrial maritime data, specifically remote sensing imagery (Synthetic Aperture Radar) and structured navigational records (Automatic Identification System). The system features two core contributions: (1) a large-scale dataset constructed with diverse maritime scenarios, integrating high-resolution visual data, structured metadata, and natural language descriptions tailored for industrial maritime contexts; and (2) a transformer-based vision-language model fine-tuned on this dataset to jointly perform vessel detection, identification, and behavioral compliance assessment—tasks crucial for industrial maritime operations. To improve the model’s reasoning depth and align with the requirements for trustworthy Artificial Intelligence in industrial settings, we incorporate Chain-of-Thought prompting, facilitating transparent, step-by-step decision-making processes. Extensive experiments validate our model’s high accuracy, interpretability, and generalization capabilities in complex operational settings, significantly enhancing anomaly detection and supporting reliable maritime safety decisions. This study underscores the capability of Industrial Large Models to interpret complex multimodal industrial data for enhanced maritime situational awareness, providing a practical illustration of adapting foundational models for safety-critical and regulatory compliance tasks, thereby advancing the understanding and application of large model technology within specialized industrial sectors.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"191 ","pages":"Article 114701"},"PeriodicalIF":6.6,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080301","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

Interpretable fake review detection based on multimodal information and human–computer collaboration 基于多模态信息和人机协作的可解释性虚假评论检测

IF 6.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2026-01-27 DOI: 10.1016/j.asoc.2026.114722

Tingting Xu, Yuhong Huo

The growing prevalence of fake reviews on online platforms poses a significant threat to consumer decision-making and corporate reputations. Although existing fake review detection (FRD) methods have, to some extent, curbed the spread of fake reviews, their detection accuracy is challenged by the continuing evolution of techniques for generating fake reviews. In this paper, we propose a novel FRD framework based on multimodal information and human–computer collaboration, aiming to enhance the accuracy and interpretability of detection. We first employ natural language processing techniques to extract and categorize item and context attribute variables in reviews, thereby constructing comprehensive user, item, and review features. Then, we introduce an interpretable FRD model, called IFML, based on multimodal variational autoencoders and large language models (LLMs). This model achieves an improved detection accuracy and generates detailed interpretations of the results by integrating multimodal data (textual, image, and structured features) and exploiting the semantic understanding capabilities of LLMs. Additionally, we design a Bayesian result aggregation framework that combines machine detection results (from the IFML model) with human feedback (trust features). This framework aims to determine the authenticity of reviews with greater accuracy by integrating the strengths of both machine and human judgments through Bayesian modeling. Experiments on three real-world Yelp datasets demonstrate that the proposed framework achieves an average performance improvement of 2.55% over the optimal baseline model. This study provides a novel perspective on FRD and demonstrates the value of human–machine collaboration in enhancing detection accuracy and interpretability.

在线平台上越来越普遍的虚假评论对消费者的决策和企业声誉构成了重大威胁。尽管现有的虚假评论检测（FRD）方法在一定程度上遏制了虚假评论的传播，但其检测准确性受到生成虚假评论技术不断发展的挑战。本文提出了一种基于多模态信息和人机协作的FRD框架，旨在提高检测的准确性和可解释性。我们首先采用自然语言处理技术对评论中的条目和上下文属性变量进行提取和分类，从而构建综合的用户、条目和评论特征。然后，我们引入了一个可解释的FRD模型，称为IFML，基于多模态变分自编码器和大型语言模型（llm）。该模型通过集成多模态数据（文本、图像和结构化特征）和利用llm的语义理解能力，提高了检测精度，并生成了对结果的详细解释。此外，我们设计了一个贝叶斯结果聚合框架，将机器检测结果（来自IFML模型）与人类反馈（信任特征）相结合。该框架旨在通过贝叶斯建模整合机器和人类判断的优势，从而更准确地确定评论的真实性。在三个真实Yelp数据集上的实验表明，该框架比最优基线模型的平均性能提高了2.55%。该研究为FRD研究提供了一个新的视角，并证明了人机协作在提高检测准确性和可解释性方面的价值。

{"title":"Interpretable fake review detection based on multimodal information and human–computer collaboration","authors":"Tingting Xu, Yuhong Huo","doi":"10.1016/j.asoc.2026.114722","DOIUrl":"10.1016/j.asoc.2026.114722","url":null,"abstract":"<div><div>The growing prevalence of fake reviews on online platforms poses a significant threat to consumer decision-making and corporate reputations. Although existing fake review detection (FRD) methods have, to some extent, curbed the spread of fake reviews, their detection accuracy is challenged by the continuing evolution of techniques for generating fake reviews. In this paper, we propose a novel FRD framework based on multimodal information and human–computer collaboration, aiming to enhance the accuracy and interpretability of detection. We first employ natural language processing techniques to extract and categorize item and context attribute variables in reviews, thereby constructing comprehensive user, item, and review features. Then, we introduce an interpretable FRD model, called IFML, based on multimodal variational autoencoders and large language models (LLMs). This model achieves an improved detection accuracy and generates detailed interpretations of the results by integrating multimodal data (textual, image, and structured features) and exploiting the semantic understanding capabilities of LLMs. Additionally, we design a Bayesian result aggregation framework that combines machine detection results (from the IFML model) with human feedback (trust features). This framework aims to determine the authenticity of reviews with greater accuracy by integrating the strengths of both machine and human judgments through Bayesian modeling. Experiments on three real-world Yelp datasets demonstrate that the proposed framework achieves an average performance improvement of 2.55% over the optimal baseline model. This study provides a novel perspective on FRD and demonstrates the value of human–machine collaboration in enhancing detection accuracy and interpretability.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"191 ","pages":"Article 114722"},"PeriodicalIF":6.6,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080300","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

Model predictive control for 3D multi-UAV conflict resolution based on velocity obstacle and dynamic window triggering 基于速度障碍和动态窗口触发的三维多无人机冲突模型预测控制

IF 6.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2026-01-27 DOI: 10.1016/j.asoc.2026.114726

Yingxue Yu , Qingwei Zhong , Yongxiang Zhang , Weijun Pan , Yi Ai

The urban low-altitude airspace is characterized by high density, narrow spacing, and strong dynamics, posing significant challenges to existing conflict management approaches. Traditional methods based on rule-based buffers, geometric feasible domains, and rolling optimization struggle to achieve a balance among real-time performance, system stability, and scalability. To address these limitations, this study proposes a trigger-based fusion framework integrating Model Predictive Control (MPC) for conflict resolution in multi-UAV operations. The framework introduces a dynamic time window mechanism to initiate single-step optimization only when predicted conflicts enter the effective response horizon, thereby reducing unnecessary computational overhead. A local feasible velocity domain is constructed based on an improved Velocity Obstacle (VO) method to define the searchable decision space. At the decision layer, a bi-objective principle minimizes velocity deviations and suppresses trajectory disturbances, while explicitly penalizing potential secondary conflicts to enhance separation stability. At the execution layer, optimal velocity decisions are mapped to real-time state updates. At the system level, periodic rolling scheduling is employed to reconstruct Unmanned Aerial Vehicle (UAV) fleet states and evolving conflict scenarios, enabling long-term continuous operation. Simulation results demonstrate an average response delay of 139 seconds, significantly reducing false alarms. The dynamic time window constrains disengagement initiation to within 200 s prior to conflict onset, ensuring timely response and maneuver feasibility. On average, 37.83 % of trajectories need to be adjusted to avoid all secondary conflicts, confirming the method’s comprehensive advantages in safety, efficiency, and stability.

城市低空空域具有高密度、窄间距、强动态性等特点，对现有的冲突管理方法提出了重大挑战。基于规则缓冲、几何可行域和滚动优化的传统方法难以在实时性、系统稳定性和可扩展性之间取得平衡。为了解决这些限制，本研究提出了一种基于触发器的融合框架，将模型预测控制（MPC）集成到多无人机作战中的冲突解决中。该框架引入了动态时间窗口机制，仅在预测冲突进入有效响应范围时才启动单步优化，从而减少了不必要的计算开销。基于改进的速度障碍（velocity Obstacle， VO）方法，构造了一个局部可行速度域来定义可搜索的决策空间。在决策层，双目标原则最小化速度偏差并抑制轨迹干扰，同时明确惩罚潜在的次要冲突以增强分离稳定性。在执行层，最佳速度决策被映射到实时状态更新。在系统层面，采用周期滚动调度方法重构无人机机群状态和不断演变的冲突场景，实现无人机的长期连续运行。仿真结果表明，平均响应延迟为139秒，大大减少了误报。动态时间窗将脱离接触起始时间限制在冲突开始前200 秒内，确保了及时响应和机动的可行性。平均有37.83 %的轨迹需要调整以避免所有次要冲突，证实了该方法在安全性、效率和稳定性方面的综合优势。

{"title":"Model predictive control for 3D multi-UAV conflict resolution based on velocity obstacle and dynamic window triggering","authors":"Yingxue Yu , Qingwei Zhong , Yongxiang Zhang , Weijun Pan , Yi Ai","doi":"10.1016/j.asoc.2026.114726","DOIUrl":"10.1016/j.asoc.2026.114726","url":null,"abstract":"<div><div>The urban low-altitude airspace is characterized by high density, narrow spacing, and strong dynamics, posing significant challenges to existing conflict management approaches. Traditional methods based on rule-based buffers, geometric feasible domains, and rolling optimization struggle to achieve a balance among real-time performance, system stability, and scalability. To address these limitations, this study proposes a trigger-based fusion framework integrating Model Predictive Control (MPC) for conflict resolution in multi-UAV operations. The framework introduces a dynamic time window mechanism to initiate single-step optimization only when predicted conflicts enter the effective response horizon, thereby reducing unnecessary computational overhead. A local feasible velocity domain is constructed based on an improved Velocity Obstacle (VO) method to define the searchable decision space. At the decision layer, a bi-objective principle minimizes velocity deviations and suppresses trajectory disturbances, while explicitly penalizing potential secondary conflicts to enhance separation stability. At the execution layer, optimal velocity decisions are mapped to real-time state updates. At the system level, periodic rolling scheduling is employed to reconstruct Unmanned Aerial Vehicle (UAV) fleet states and evolving conflict scenarios, enabling long-term continuous operation. Simulation results demonstrate an average response delay of 139 seconds, significantly reducing false alarms. The dynamic time window constrains disengagement initiation to within 200 s prior to conflict onset, ensuring timely response and maneuver feasibility. On average, 37.83 % of trajectories need to be adjusted to avoid all secondary conflicts, confirming the method’s comprehensive advantages in safety, efficiency, and stability.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"191 ","pages":"Article 114726"},"PeriodicalIF":6.6,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080198","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 learning-driven artificial bee colony algorithm for mobile robot multi-objective path planning 移动机器人多目标路径规划的学习驱动人工蜂群算法

IF 6.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2026-01-26 DOI: 10.1016/j.asoc.2026.114692

Fan Ye , Peng Duan , Leilei Meng , Hongyan Sang , Maodi Ru , Kaizhou Gao

Path planning is a critical component enabling mobile robots to perform diverse tasks, with complexity increasing significantly when simultaneously optimizing multiple objectives. In this study, a learning-driven artificial bee colony algorithm (LBABC) is proposed to address multi-objective path planning problems, where three objectives are optimized, i.e., path length, path safety, and path smoothness. In the initialization phase, a competition-based initialization strategy is investigated to generate a diverse set of initial solutions. In the employed bee phase, a differential evolution strategy is developed to enhance the algorithm’s global exploration ability, with differential mutation and crossing operators redesigned to suit the specific characteristics of the path planning problems. In the onlooker bee phase, a Q-learning-based evolutionary operator selection strategy is designed to improve the algorithm’s local search ability, utilizing six problem-specific evolutionary operators. In the scout bee phase, an adaptive evolutionary restart strategy is presented to replace those stagnant solutions with promising ones to enhance population activity. Finally, the proposed LBABC was compared with seven state-of-the-art algorithms across 16 instances from four respective environments. Additionally, all components proposed by the LBABC were examined through ablation testing. Experimental results showed that the proposed LBABC obtained average improvements of 1.37% and 40.74% on the hypervolume and inverted generational distance metrics, respectively, compared with the results obtained by the corresponding algorithm with the second-best performance. These results demonstrated the effectiveness of the designed components and the superior performance of the proposed LBABC in solving multi-objective path planning problems.

路径规划是移动机器人实现多样化任务的关键组成部分，当同时优化多个目标时，其复杂性会显著增加。针对多目标路径规划问题，提出了一种学习驱动的人工蜂群算法（LBABC），对路径长度、路径安全、路径平滑三个目标进行优化。在初始化阶段，研究了一种基于竞争的初始化策略，以生成不同的初始解集。在蜜蜂阶段，采用差分进化策略增强算法的全局搜索能力，并根据路径规划问题的具体特点重新设计差分变异算子和杂交算子。在围观蜂阶段，设计了一种基于q学习的进化算子选择策略，利用6个针对特定问题的进化算子来提高算法的局部搜索能力。在侦察蜂阶段，提出了一种适应性进化重启策略，以有希望的方案取代停滞不前的方案，以增强种群活动。最后，将提出的LBABC与来自四个不同环境的16个实例中的七种最先进算法进行比较。此外，通过烧蚀测试对LBABC提出的所有组件进行了检查。实验结果表明，与性能次优的相应算法相比，LBABC算法在超大容量和倒代距离指标上的平均性能分别提高了1.37%和40.74%。这些结果证明了所设计组件的有效性以及所提出的LBABC在解决多目标路径规划问题方面的优越性能。

{"title":"A learning-driven artificial bee colony algorithm for mobile robot multi-objective path planning","authors":"Fan Ye , Peng Duan , Leilei Meng , Hongyan Sang , Maodi Ru , Kaizhou Gao","doi":"10.1016/j.asoc.2026.114692","DOIUrl":"10.1016/j.asoc.2026.114692","url":null,"abstract":"<div><div>Path planning is a critical component enabling mobile robots to perform diverse tasks, with complexity increasing significantly when simultaneously optimizing multiple objectives. In this study, a learning-driven artificial bee colony algorithm (LBABC) is proposed to address multi-objective path planning problems, where three objectives are optimized, i.e., path length, path safety, and path smoothness. In the initialization phase, a competition-based initialization strategy is investigated to generate a diverse set of initial solutions. In the employed bee phase, a differential evolution strategy is developed to enhance the algorithm’s global exploration ability, with differential mutation and crossing operators redesigned to suit the specific characteristics of the path planning problems. In the onlooker bee phase, a Q-learning-based evolutionary operator selection strategy is designed to improve the algorithm’s local search ability, utilizing six problem-specific evolutionary operators. In the scout bee phase, an adaptive evolutionary restart strategy is presented to replace those stagnant solutions with promising ones to enhance population activity. Finally, the proposed LBABC was compared with seven state-of-the-art algorithms across 16 instances from four respective environments. Additionally, all components proposed by the LBABC were examined through ablation testing. Experimental results showed that the proposed LBABC obtained average improvements of 1.37% and 40.74% on the hypervolume and inverted generational distance metrics, respectively, compared with the results obtained by the corresponding algorithm with the second-best performance. These results demonstrated the effectiveness of the designed components and the superior performance of the proposed LBABC in solving multi-objective path planning problems.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"192 ","pages":"Article 114692"},"PeriodicalIF":6.6,"publicationDate":"2026-01-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146096144","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

Spectrum sparsity-driven learnable wavelet packet transform for denoising and its application in fault diagnosis 谱稀疏驱动的可学习小波包去噪及其在故障诊断中的应用

IF 6.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2026-01-25 DOI: 10.1016/j.asoc.2026.114713

Xin Kang, Jianxin Cui, Junsheng Cheng, Yu Yang, Qiyao Liu

Domain-specific noise is a key cause of cross-domain data distribution shifts and a major obstacle in cross-domain fault diagnosis. Most existing denoising methods are built on the assumptions that noise is weak and Gaussian, which often do not hold in the industrial context, thus limiting their effectiveness. To address this, we propose a spectrum sparsity-driven learnable wavelet packet transform (SSLWPT) model for unsupervised, small-sample mechanical signal denoising and feature enhancement. First, based on the vibration properties of rotating machinery, where stationary and cyclostationary components carry critical health-related information, we propose a spectrum sparsity-driven mechanism to actively capture these components for adaptive feature enhancement. Next, an autoencoder mimicking wavelet packet decomposition and reconstruction is proposed, with learnable wavelet bases and denoising parameters. Under the joint constraints of reconstruction loss, parameter regularization, and spectrum sparsity loss, the model achieves effective signal denoising. We further integrate the proposed SSLWPT as a preprocessing tool into a two-stage framework for cross-domain fault diagnosis. Both theoretical analysis and experimental results validate the superiority of the proposed model in denoising. Moreover, cross-domain diagnosis experiments conducted on bearing, gear and motor datasets confirm that our method achieves significantly better generalization than related denoising methods and SOTA domain generalization methods.

域特异性噪声是导致数据跨域分布偏移的主要原因，也是跨域故障诊断的主要障碍。大多数现有的去噪方法都是建立在假设噪声是弱的和高斯的基础上的，这在工业环境中往往不成立，从而限制了它们的有效性。为了解决这个问题，我们提出了一个频谱稀疏驱动的可学习小波包变换（SSLWPT）模型，用于无监督、小样本机械信号去噪和特征增强。首先，基于旋转机械的振动特性，其中静止和循环静止组件携带关键的健康相关信息，我们提出了一种频谱稀疏驱动机制来主动捕获这些组件以进行自适应特征增强。其次，提出了一种模拟小波包分解和重构的自编码器，具有可学习的小波基和去噪参数。在重构损失、参数正则化和频谱稀疏性损失的共同约束下，该模型实现了有效的信号去噪。我们进一步将提出的SSLWPT作为预处理工具集成到跨域故障诊断的两阶段框架中。理论分析和实验结果都验证了该模型在去噪方面的优越性。此外，在轴承、齿轮和电机数据集上进行的跨域诊断实验证实，我们的方法的泛化效果明显优于相关的去噪方法和SOTA域泛化方法。

{"title":"Spectrum sparsity-driven learnable wavelet packet transform for denoising and its application in fault diagnosis","authors":"Xin Kang, Jianxin Cui, Junsheng Cheng, Yu Yang, Qiyao Liu","doi":"10.1016/j.asoc.2026.114713","DOIUrl":"10.1016/j.asoc.2026.114713","url":null,"abstract":"<div><div>Domain-specific noise is a key cause of cross-domain data distribution shifts and a major obstacle in cross-domain fault diagnosis. Most existing denoising methods are built on the assumptions that noise is weak and Gaussian, which often do not hold in the industrial context, thus limiting their effectiveness. To address this, we propose a spectrum sparsity-driven learnable wavelet packet transform (SSLWPT) model for unsupervised, small-sample mechanical signal denoising and feature enhancement. First, based on the vibration properties of rotating machinery, where stationary and cyclostationary components carry critical health-related information, we propose a spectrum sparsity-driven mechanism to actively capture these components for adaptive feature enhancement. Next, an autoencoder mimicking wavelet packet decomposition and reconstruction is proposed, with learnable wavelet bases and denoising parameters. Under the joint constraints of reconstruction loss, parameter regularization, and spectrum sparsity loss, the model achieves effective signal denoising. We further integrate the proposed SSLWPT as a preprocessing tool into a two-stage framework for cross-domain fault diagnosis. Both theoretical analysis and experimental results validate the superiority of the proposed model in denoising. Moreover, cross-domain diagnosis experiments conducted on bearing, gear and motor datasets confirm that our method achieves significantly better generalization than related denoising methods and SOTA domain generalization methods.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"191 ","pages":"Article 114713"},"PeriodicalIF":6.6,"publicationDate":"2026-01-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080107","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

FAHFD: Fractional-order adaptive Higuchi fractal dimension for signal recognition 分数阶自适应Higuchi分形维数用于信号识别

IF 6.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2026-01-25 DOI: 10.1016/j.asoc.2026.114721

Yuxing Li, Jingyi Li, Yonghong Zhao

Fractal dimension, as a classic nonlinear dynamic metric, can effectively characterize the complexity of signals and has been widely applied to signal recognition. However, existing fractal dimensions are limited to first-order information and suffer from the issue of parameter selection. This paper proposed the fractional-order adaptive Higuchi fractal dimension, where fractional-order calculus endows the metric with the ability to capture fractional-order information, and the design of a parameter-adaptive selection mechanism addresses the parameter selection problem. Simulation experiments show that the proposed fractional-order adaptive Higuchi fractal dimension achieves excellent classification performance on three simulation datasets involving colored noise discrimination, chaotic series analysis, and simulated bearing vibration; experimental tests on real-world datasets demonstrate that fractional-order adaptive Higuchi fractal dimension also exhibits strong generalizability, fully validating its reliability and effectiveness in addressing complex real-world engineering diagnosis tasks across different scenarios on three real-world engineering datasets including faulty bearings, faulty gears, and ship radiated noise signal.

分形维数作为一种经典的非线性动态度量，能有效表征信号的复杂程度，在信号识别中得到了广泛的应用。然而，现有的分形维数仅限于一阶信息，且存在参数选择问题。本文提出了分数阶自适应Higuchi分形维数，其中分数阶微积分赋予度量捕获分数阶信息的能力，并设计了参数自适应选择机制来解决参数选择问题。仿真实验表明，所提出的分数阶自适应Higuchi分形维数在有色噪声识别、混沌序列分析和模拟轴承振动三个仿真数据集上都取得了优异的分类性能；在实际数据集上的实验测试表明，分数阶自适应Higuchi分维也表现出较强的泛化能力，充分验证了分数阶自适应Higuchi分维在故障轴承、故障齿轮和船舶辐射噪声信号三种实际工程数据集上解决复杂实际工程诊断任务的可靠性和有效性。

{"title":"FAHFD: Fractional-order adaptive Higuchi fractal dimension for signal recognition","authors":"Yuxing Li, Jingyi Li, Yonghong Zhao","doi":"10.1016/j.asoc.2026.114721","DOIUrl":"10.1016/j.asoc.2026.114721","url":null,"abstract":"<div><div>Fractal dimension, as a classic nonlinear dynamic metric, can effectively characterize the complexity of signals and has been widely applied to signal recognition. However, existing fractal dimensions are limited to first-order information and suffer from the issue of parameter selection. This paper proposed the fractional-order adaptive Higuchi fractal dimension, where fractional-order calculus endows the metric with the ability to capture fractional-order information, and the design of a parameter-adaptive selection mechanism addresses the parameter selection problem. Simulation experiments show that the proposed fractional-order adaptive Higuchi fractal dimension achieves excellent classification performance on three simulation datasets involving colored noise discrimination, chaotic series analysis, and simulated bearing vibration; experimental tests on real-world datasets demonstrate that fractional-order adaptive Higuchi fractal dimension also exhibits strong generalizability, fully validating its reliability and effectiveness in addressing complex real-world engineering diagnosis tasks across different scenarios on three real-world engineering datasets including faulty bearings, faulty gears, and ship radiated noise signal.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"191 ","pages":"Article 114721"},"PeriodicalIF":6.6,"publicationDate":"2026-01-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080202","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 differential evolutionary algorithm improvement framework based on artificial potential fields 基于人工势场的差分进化算法改进框架

IF 6.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2026-01-24 DOI: 10.1016/j.asoc.2026.114714

Zhe Yang , Libao Deng , Xianxin Mao , Lili Zhang

The Artificial Potential Field (APF) method, widely utilized in path planning, uses virtual environmental forces derived from potential field gradients for guidance. While Differential Evolution (DE) algorithms are recognized for their simplicity and robust global optimization capabilities, they often suffer from premature convergence and reduced search efficiency in complex scenarios. Motivated by the similarity between APF pathfinding and DE optimization, this study proposes an APF-inspired framework that utilizes virtual forces to guide the population. Attractive forces guide individuals toward potential optima, while repulsive forces generated by treating inferior solutions as virtual obstacles mitigate the risk of stagnation. This mechanism effectively steers the population toward promising areas, thereby enhancing search efficiency. Furthermore, an integrated adaptive strategy enhances performance by dynamically adjusting key parameters based on problem characteristics. Comprehensive experiments on the CEC2020 benchmark suite and 36 real-world engineering problems validate the effectiveness of the framework. The results indicate performance improvements over baseline algorithms and competitiveness in comparative studies.

人工势场法（Artificial Potential Field， APF）是一种利用势场梯度产生的虚拟环境力进行路径引导的方法，在路径规划中得到广泛应用。差分进化（DE）算法以其简单性和鲁棒的全局优化能力而闻名，但在复杂场景中，它们往往存在过早收敛和搜索效率降低的问题。基于APF寻路与DE优化之间的相似性，本研究提出了一种基于APF的框架，该框架利用虚拟力来引导种群。吸引力引导个体走向潜在的最佳状态，而将次等解决方案视为虚拟障碍所产生的排斥力则减轻了停滞的风险。这一机制有效地引导人们转向有希望的领域，从而提高搜索效率。此外，集成自适应策略通过根据问题特征动态调整关键参数来提高性能。在CEC2020基准测试套件和36个实际工程问题上的综合实验验证了该框架的有效性。结果表明性能优于基线算法和竞争力的比较研究。

{"title":"A differential evolutionary algorithm improvement framework based on artificial potential fields","authors":"Zhe Yang , Libao Deng , Xianxin Mao , Lili Zhang","doi":"10.1016/j.asoc.2026.114714","DOIUrl":"10.1016/j.asoc.2026.114714","url":null,"abstract":"<div><div>The Artificial Potential Field (APF) method, widely utilized in path planning, uses virtual environmental forces derived from potential field gradients for guidance. While Differential Evolution (DE) algorithms are recognized for their simplicity and robust global optimization capabilities, they often suffer from premature convergence and reduced search efficiency in complex scenarios. Motivated by the similarity between APF pathfinding and DE optimization, this study proposes an APF-inspired framework that utilizes virtual forces to guide the population. Attractive forces guide individuals toward potential optima, while repulsive forces generated by treating inferior solutions as virtual obstacles mitigate the risk of stagnation. This mechanism effectively steers the population toward promising areas, thereby enhancing search efficiency. Furthermore, an integrated adaptive strategy enhances performance by dynamically adjusting key parameters based on problem characteristics. Comprehensive experiments on the CEC2020 benchmark suite and 36 real-world engineering problems validate the effectiveness of the framework. The results indicate performance improvements over baseline algorithms and competitiveness in comparative studies.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"192 ","pages":"Article 114714"},"PeriodicalIF":6.6,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146096146","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

Quantum-enhanced recurrent models for cognitive–motor assessment 认知-运动评估的量子增强循环模型

IF 6.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2026-01-24 DOI: 10.1016/j.asoc.2026.114696

Basma Jalloul, Bassem Bouaziz, Walid Mahdi

Accurate assessment of cognitive and motor function is fundamental for the early detection of neurodegenerative and cerebrovascular conditions such as Mild Cognitive Impairment (MCI), stroke, and Parkinson’s disease. While clinical evaluations are often subjective, recent advances in wearable sensing and pose estimation have enabled objective gait analysis across these disorders. However, the inherent noise, artifacts, and inter-subject variability of real-world clinical data remain challenging for conventional deep learning models, which tend to overfit synthetic data and generalize poorly. In this work, we propose a hybrid soft-computing framework that integrates parameterized quantum circuits with recurrent neural networks (QE-RNNs) to enhance feature representation and robustness. Quantum-enhanced embeddings are executed on real quantum hardware and combined with classical temporal modeling to capture complex brain–motor dynamics. Experimental results across synthetic and clinical gait datasets show that, whereas classical RNNs lose significant accuracy in noisy environments, QE-RNNs maintain strong generalization, achieving up to 99.25% accuracy in stroke-related gait analysis and 81.43% in Parkinson’s motor-state classification. These findings highlight the potential of quantum-inspired soft computing for developing resilient, explainable, and clinically relevant tools for motor assessment.

准确评估认知和运动功能是早期发现神经退行性和脑血管疾病（如轻度认知障碍（MCI）、中风和帕金森病）的基础。虽然临床评估往往是主观的，但最近在可穿戴传感和姿势估计方面的进展使这些疾病的客观步态分析成为可能。然而，现实世界临床数据固有的噪声、伪影和学科间的可变性仍然是传统深度学习模型面临的挑战，这些模型往往会过度拟合合成数据，泛化能力差。在这项工作中，我们提出了一种混合软计算框架，该框架将参数化量子电路与递归神经网络（qe - rnn）集成在一起，以增强特征表示和鲁棒性。量子增强嵌入在真实的量子硬件上执行，并结合经典的时间建模来捕获复杂的脑运动动力学。合成和临床步态数据集的实验结果表明，经典rnn在噪声环境中失去了显著的准确性，而qe - rnn保持了很强的一般化，在卒中相关步态分析中准确率高达99.25%，在帕金森运动状态分类中准确率高达81.43%。这些发现突出了量子启发的软计算在开发有弹性的、可解释的和临床相关的运动评估工具方面的潜力。

{"title":"Quantum-enhanced recurrent models for cognitive–motor assessment","authors":"Basma Jalloul, Bassem Bouaziz, Walid Mahdi","doi":"10.1016/j.asoc.2026.114696","DOIUrl":"10.1016/j.asoc.2026.114696","url":null,"abstract":"<div><div>Accurate assessment of cognitive and motor function is fundamental for the early detection of neurodegenerative and cerebrovascular conditions such as Mild Cognitive Impairment (MCI), stroke, and Parkinson’s disease. While clinical evaluations are often subjective, recent advances in wearable sensing and pose estimation have enabled objective gait analysis across these disorders. However, the inherent noise, artifacts, and inter-subject variability of real-world clinical data remain challenging for conventional deep learning models, which tend to overfit synthetic data and generalize poorly. In this work, we propose a hybrid soft-computing framework that integrates parameterized quantum circuits with recurrent neural networks (QE-RNNs) to enhance feature representation and robustness. Quantum-enhanced embeddings are executed on real quantum hardware and combined with classical temporal modeling to capture complex brain–motor dynamics. Experimental results across synthetic and clinical gait datasets show that, whereas classical RNNs lose significant accuracy in noisy environments, QE-RNNs maintain strong generalization, achieving up to 99.25% accuracy in stroke-related gait analysis and 81.43% in Parkinson’s motor-state classification. These findings highlight the potential of quantum-inspired soft computing for developing resilient, explainable, and clinically relevant tools for motor assessment.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"191 ","pages":"Article 114696"},"PeriodicalIF":6.6,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080041","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

Reinforcement learning-driven nonsingular fast terminal sliding mode control for pipe crack sealing manipulator 管道缝封机械手的强化学习驱动非奇异快速终端滑模控制

IF 6.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2026-01-24 DOI: 10.1016/j.asoc.2026.114708

Santosh Kumar, S.K. Dwivedy

This study explores control challenges associated with the Pipe Crack Sealing Manipulator (PCSM), a tree-shaped robot designed to seal cracks within concrete pipes. Equipped with five specialized branches, the PCSM adeptly navigates both vertical and horizontal pipe environments. A detailed CAD model was developed in SolidWorks, and data were meticulously exported to Simulink for refined control operations. SolidWorks simulations revealed that the PCSM’s fifth branch excelled in extensive crack repairs. The control of pipe crack sealing manipulators presents significant challenges due to uncertain dynamics, random disturbances, signal delay, and sensor noise inherent in confined environments. To overcome these issues, this study proposes a reinforcement learning–driven nonsingular fast terminal sliding mode control (RLNFTSMC) framework that enhances robustness, precision, and adaptability for crack sealing operations. The implemented RLNFTSMC strategy, includes an equivalent part utilizing actor-critic neural networks for dynamic estimation and a switching part that treats the mass matrix as a lumped value. Comparative analysis with nonsingular fast terminal sliding mode control (NFTSMC), adaptive NFTSMC, radial basis function neural network control (RBFNNC), and neural network-enhanced nonsingular fast terminal sliding mode control (NENFTSMC) shows that RLNFTSMC significantly improves the PCSM’s reliability and effectiveness in environments marked by random disturbances and unknown dynamics.

本研究探讨了与管道裂缝密封机械手（PCSM）相关的控制挑战，PCSM是一种设计用于密封混凝土管道裂缝的树形机器人。PCSM配备了五个专门的分支机构，可以熟练地导航垂直和水平管道环境。在SolidWorks中开发了详细的CAD模型，并将数据精心导出到Simulink中进行精细控制操作。SolidWorks模拟显示，PCSM的第五分支机构在广泛的裂缝修复方面表现出色。由于不确定动力学、随机干扰、信号延迟和受限环境中固有的传感器噪声，管道裂缝密封机械手的控制面临着重大挑战。为了克服这些问题，本研究提出了一种强化学习驱动的非奇异快速终端滑模控制（RLNFTSMC）框架，该框架增强了裂缝密封操作的鲁棒性、精度和适应性。实现的RLNFTSMC策略包括一个等效部分，该部分利用行动者批评神经网络进行动态估计，以及一个将质量矩阵视为集总值的切换部分。与非奇异快速终端滑模控制（NFTSMC）、自适应非奇异终端滑模控制（NFTSMC）、径向基函数神经网络控制（RBFNNC）和神经网络增强非奇异快速终端滑模控制（NENFTSMC）的对比分析表明，RLNFTSMC显著提高了PCSM在随机干扰和未知动态环境中的可靠性和有效性。

{"title":"Reinforcement learning-driven nonsingular fast terminal sliding mode control for pipe crack sealing manipulator","authors":"Santosh Kumar, S.K. Dwivedy","doi":"10.1016/j.asoc.2026.114708","DOIUrl":"10.1016/j.asoc.2026.114708","url":null,"abstract":"<div><div>This study explores control challenges associated with the Pipe Crack Sealing Manipulator (PCSM), a tree-shaped robot designed to seal cracks within concrete pipes. Equipped with five specialized branches, the PCSM adeptly navigates both vertical and horizontal pipe environments. A detailed CAD model was developed in SolidWorks, and data were meticulously exported to Simulink for refined control operations. SolidWorks simulations revealed that the PCSM’s fifth branch excelled in extensive crack repairs. The control of pipe crack sealing manipulators presents significant challenges due to uncertain dynamics, random disturbances, signal delay, and sensor noise inherent in confined environments. To overcome these issues, this study proposes a reinforcement learning–driven nonsingular fast terminal sliding mode control (RLNFTSMC) framework that enhances robustness, precision, and adaptability for crack sealing operations. The implemented RLNFTSMC strategy, includes an equivalent part utilizing actor-critic neural networks for dynamic estimation and a switching part that treats the mass matrix as a lumped value. Comparative analysis with nonsingular fast terminal sliding mode control (NFTSMC), adaptive NFTSMC, radial basis function neural network control (RBFNNC), and neural network-enhanced nonsingular fast terminal sliding mode control (NENFTSMC) shows that RLNFTSMC significantly improves the PCSM’s reliability and effectiveness in environments marked by random disturbances and unknown dynamics.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"191 ","pages":"Article 114708"},"PeriodicalIF":6.6,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080293","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

Saturation and pressure change estimation from time-lapse seismic data using vision transformers 利用视觉变压器估计时移地震数据的饱和度和压力变化

IF 6.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2026-01-24 DOI: 10.1016/j.asoc.2026.114715

Marcos Cirne , Masoud Maleki , Michael Diniz , Denis Schiozer , Alessandra Davolio , Anderson Rocha

Deep learning-based algorithms have been used in oil and gas industry applications to optimize the performance of four-dimensional (4D) seismic inversion methods. These pose an important problem in reservoir management since they convey substantial information about fluid recovery and distribution on the reservoir over time. In this work, we propose a novel deep learning architecture based in Vision Transformers (ViT) to estimate saturation and pressure changes from 4D seismic data using data collected from a pre-salt oil field. The proposed method is first trained with an ensemble of simulation models (synthetic domain) and the resulting architecture is used to predict the actual changes using real 4D seismic data (observed domain). Due to the absence of data points around the reservoir area conveyed in our dataset, we redesigned the ViT architecture so that it only processes the existing data points in a 1-D fashion, making the architecture capable of processing different reservoir shapes other than rectangular ones. The quantitative results against recent deep learning methods corroborate the effectiveness of our proposed architecture. Finally, a qualitative analysis by a specialist and the use of an Explainable Artificial Intelligence (XAI) method helped assess the key differences in the decision-making processes taken by each analyzed algorithm.

基于深度学习的算法已被应用于油气行业，以优化四维（4D）地震反演方法的性能。这些数据在油藏管理中构成了一个重要的问题，因为它们传达了油藏随时间变化的流体采收率和分布的大量信息。在这项工作中，我们提出了一种基于视觉变压器（ViT）的新型深度学习架构，利用从盐下油田收集的数据，从四维地震数据中估计饱和度和压力变化。该方法首先使用模拟模型集合（合成域）进行训练，然后使用实际四维地震数据（观测域）预测实际变化。由于我们的数据集中没有传达库区周围的数据点，我们重新设计了ViT架构，使其仅以1-D的方式处理现有的数据点，使该架构能够处理除矩形以外的不同油藏形状。针对最近深度学习方法的定量结果证实了我们提出的架构的有效性。最后，由专家进行的定性分析和可解释人工智能（XAI）方法的使用有助于评估每种分析算法在决策过程中所采取的关键差异。

{"title":"Saturation and pressure change estimation from time-lapse seismic data using vision transformers","authors":"Marcos Cirne , Masoud Maleki , Michael Diniz , Denis Schiozer , Alessandra Davolio , Anderson Rocha","doi":"10.1016/j.asoc.2026.114715","DOIUrl":"10.1016/j.asoc.2026.114715","url":null,"abstract":"<div><div>Deep learning-based algorithms have been used in oil and gas industry applications to optimize the performance of four-dimensional (4D) seismic inversion methods. These pose an important problem in reservoir management since they convey substantial information about fluid recovery and distribution on the reservoir over time. In this work, we propose a novel deep learning architecture based in Vision Transformers (ViT) to estimate saturation and pressure changes from 4D seismic data using data collected from a pre-salt oil field. The proposed method is first trained with an ensemble of simulation models (synthetic domain) and the resulting architecture is used to predict the actual changes using real 4D seismic data (observed domain). Due to the absence of data points around the reservoir area conveyed in our dataset, we redesigned the ViT architecture so that it only processes the existing data points in a 1-D fashion, making the architecture capable of processing different reservoir shapes other than rectangular ones. The quantitative results against recent deep learning methods corroborate the effectiveness of our proposed architecture. Finally, a qualitative analysis by a specialist and the use of an Explainable Artificial Intelligence (XAI) method helped assess the key differences in the decision-making processes taken by each analyzed algorithm.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"191 ","pages":"Article 114715"},"PeriodicalIF":6.6,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080101","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