Engineering Applications of Artificial Intelligence最新文献_第5页

Bayesian optimization interval type-3 fuzzy broad compensated intelligent control for flue gas oxygen content 烟气含氧量贝叶斯优化区间3型模糊广义补偿智能控制

IF 8 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2026-02-10 DOI: 10.1016/j.engappai.2026.114044

Weiwei Yang , Jian Tang , Wen Yu , Junfei Qiao

In industrial sites of municipal solid waste incineration (MSWI) processes in developing countries such as China, manual control modes based on domain experts' embodied intelligence are commonly used for stable operation. Flue gas oxygen content is a crucial controlled variable in the MSWI process, where traditional control methods often lack adaptability and robustness under nonlinear uncertainties. To achieve high-precision and robust oxygen content control, this study aims to develop a novel intelligent control strategy. We propose a Bayesian optimization (BO)-based interval type-3 fuzzy broad compensated control method. The core of this approach is a parallel control architecture, which integrates an interval type-3 fuzzy broad learning system (IT3FBLS) constructed from prior knowledge with a conventional proportion integration differentiation (PID) controller. Furthermore, the BO algorithm is introduced to automatically tune the numerous hyperparameters of the hybrid IT3FBLS-PID controller, ensuring optimal performance. Experimental validation using data from an actual MSWI power plant demonstrates that, compared to conventional PID and fuzzy PID controllers, the proposed method achieves smaller steady-state error, faster response speed, and exhibits superior disturbance rejection capability. This work introduces a novel parallel control paradigm that effectively combines the interpretability and adaptability of advanced fuzzy broad learning systems with the stability of classical control. It also offers a practical BO-driven solution for parameter optimization, aimed at enhancing intelligent applications in complex industrial control systems.

在中国等发展中国家的城市生活垃圾焚烧工业现场，为了稳定运行，通常采用基于领域专家具身智能的手动控制模式。烟气含氧量是城市污水处理厂过程中一个重要的控制变量，传统的控制方法在非线性不确定性下往往缺乏适应性和鲁棒性。为了实现高精度和鲁棒的氧含量控制，本研究旨在开发一种新的智能控制策略。提出了一种基于贝叶斯优化（BO）的区间3型模糊广义补偿控制方法。该方法的核心是将基于先验知识构建的区间3型模糊广义学习系统（IT3FBLS）与传统的比例积分微分（PID）控制器相结合的并行控制体系结构。在此基础上，引入BO算法对IT3FBLS-PID混合控制器的众多超参数进行自动整定，保证了控制器的最优性能。实际MSWI电厂数据的实验验证表明，与传统PID和模糊PID控制器相比，该方法的稳态误差更小，响应速度更快，抗干扰能力更强。本文介绍了一种新的并行控制范式，该范式有效地将先进模糊广义学习系统的可解释性和适应性与经典控制的稳定性相结合。它还为参数优化提供了实用的bo驱动解决方案，旨在增强复杂工业控制系统中的智能应用。

{"title":"Bayesian optimization interval type-3 fuzzy broad compensated intelligent control for flue gas oxygen content","authors":"Weiwei Yang , Jian Tang , Wen Yu , Junfei Qiao","doi":"10.1016/j.engappai.2026.114044","DOIUrl":"10.1016/j.engappai.2026.114044","url":null,"abstract":"<div><div>In industrial sites of municipal solid waste incineration (MSWI) processes in developing countries such as China, manual control modes based on domain experts' embodied intelligence are commonly used for stable operation. Flue gas oxygen content is a crucial controlled variable in the MSWI process, where traditional control methods often lack adaptability and robustness under nonlinear uncertainties. To achieve high-precision and robust oxygen content control, this study aims to develop a novel intelligent control strategy. We propose a Bayesian optimization (BO)-based interval type-3 fuzzy broad compensated control method. The core of this approach is a parallel control architecture, which integrates an interval type-3 fuzzy broad learning system (IT3FBLS) constructed from prior knowledge with a conventional proportion integration differentiation (PID) controller. Furthermore, the BO algorithm is introduced to automatically tune the numerous hyperparameters of the hybrid IT3FBLS-PID controller, ensuring optimal performance. Experimental validation using data from an actual MSWI power plant demonstrates that, compared to conventional PID and fuzzy PID controllers, the proposed method achieves smaller steady-state error, faster response speed, and exhibits superior disturbance rejection capability. This work introduces a novel parallel control paradigm that effectively combines the interpretability and adaptability of advanced fuzzy broad learning systems with the stability of classical control. It also offers a practical BO-driven solution for parameter optimization, aimed at enhancing intelligent applications in complex industrial control systems.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"169 ","pages":"Article 114044"},"PeriodicalIF":8.0,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175307","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Deep Potential Semantic-aware Hashing for Cross-modal Retrieval 跨模态检索的深度潜在语义感知哈希

IF 8 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2026-02-10 DOI: 10.1016/j.engappai.2026.114155

Lei Wu , Qibing Qin , Jiangyan Dai , Lei Huang , Wenfeng Zhang

Hashing learning has moved into the mainstream for multimedia retrieval because it offers the advantages of low storage cost and high retrieval efficiency. Currently, most cross-modal hashing methods commonly explore the similarity relations between samples by constructing pair-wise or triplet-wise constraints. However, these methods focus on the relative correct ranking of samples, ignore the potential semantic similarity of raw sample distribution, and generate sub-optimal hash codes. To resolve this issue, the novel Deep Potential Semantic-aware Hashing framework (DPSaH) is proposed to mine the local semantic structure of heterogeneous samples, maintaining inter-modality-consistent and cross-modality-correlated semantic relationships. Specifically, by exploring the potential local structure of the data, the multi-modal quadruple loss is extended to the cross-modal hashing framework, thereby preserving the potential semantic neighborhoods among raw samples in Hamming space. During model training, based on the average semantic labels, the label-averaged balanced strategy is developed to quantify the frequency difference between positive and negative samples. Besides, by injecting noise information into the generated discrete codes, the binary-injection loss is introduced to alleviate the over-activation of specific bits, decorrelating different bits in the Hamming space. Extensive experiments are performed on three public datasets, and the results verify the superiority of the DPSaH framework compared to the current mainstream cross-modal hashing frameworks. The source code for DPSaH is available at https://github.com/QinLab-WFU/DPSaH.

哈希学习具有存储成本低、检索效率高等优点，已成为多媒体检索的主流。目前，大多数跨模态哈希方法通常通过构造成对或三重约束来探索样本之间的相似性关系。然而，这些方法侧重于样本的相对正确排序，忽略了原始样本分布的潜在语义相似性，并生成次优哈希码。为了解决这一问题，提出了一种新的深度潜在语义感知哈希框架（Deep Potential semantic -aware hash framework, dppah）来挖掘异构样本的局部语义结构，保持模态间一致和跨模态相关的语义关系。具体而言，通过探索数据潜在的局部结构，将多模态四重损失扩展到跨模态哈希框架，从而在汉明空间中保留原始样本之间潜在的语义邻域。在模型训练过程中，基于平均语义标签，提出了标签平均平衡策略来量化正负样本之间的频率差。此外，通过在生成的离散码中注入噪声信息，引入二进制注入损失来缓解特定位的过度激活，在汉明空间中解除不同位的相关。在三个公共数据集上进行了大量的实验，结果验证了dpah框架相对于当前主流的跨模态哈希框架的优越性。dpah的源代码可从https://github.com/QinLab-WFU/DPSaH获得。

{"title":"Deep Potential Semantic-aware Hashing for Cross-modal Retrieval","authors":"Lei Wu , Qibing Qin , Jiangyan Dai , Lei Huang , Wenfeng Zhang","doi":"10.1016/j.engappai.2026.114155","DOIUrl":"10.1016/j.engappai.2026.114155","url":null,"abstract":"<div><div>Hashing learning has moved into the mainstream for multimedia retrieval because it offers the advantages of low storage cost and high retrieval efficiency. Currently, most cross-modal hashing methods commonly explore the similarity relations between samples by constructing pair-wise or triplet-wise constraints. However, these methods focus on the relative correct ranking of samples, ignore the potential semantic similarity of raw sample distribution, and generate sub-optimal hash codes. To resolve this issue, the novel Deep Potential Semantic-aware Hashing framework (DPSaH) is proposed to mine the local semantic structure of heterogeneous samples, maintaining inter-modality-consistent and cross-modality-correlated semantic relationships. Specifically, by exploring the potential local structure of the data, the multi-modal quadruple loss is extended to the cross-modal hashing framework, thereby preserving the potential semantic neighborhoods among raw samples in Hamming space. During model training, based on the average semantic labels, the label-averaged balanced strategy is developed to quantify the frequency difference between positive and negative samples. Besides, by injecting noise information into the generated discrete codes, the binary-injection loss is introduced to alleviate the over-activation of specific bits, decorrelating different bits in the Hamming space. Extensive experiments are performed on three public datasets, and the results verify the superiority of the DPSaH framework compared to the current mainstream cross-modal hashing frameworks. The source code for DPSaH is available at <span><span>https://github.com/QinLab-WFU/DPSaH</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"169 ","pages":"Article 114155"},"PeriodicalIF":8.0,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175224","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Uncertainty-aware data-driven three-dimensional turbine aerodynamic design system with transformer and multi-fidelity neural networks 基于变压器和多保真度神经网络的不确定性感知数据驱动涡轮三维气动设计系统

IF 8 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2026-02-10 DOI: 10.1016/j.engappai.2026.114125

Peng Ren, Xiangjun Fang, Junfeng Chen

Gas turbines are widely used energy conversion devices, and secondary flows have a significant impact on their overall efficiency. Adjusting the stacking line through sweep and lean is an important method for controlling secondary flows. Traditional stacking line design methods typically rely on designers' experience and iterative processes, which are time-consuming, computationally expensive, and lack generalizable design guidelines. To address these challenges, this paper proposes a data-driven stacking line design method that integrates a transformer architecture with Deep Ensemble (DE) learning to model the relationship between optimal stacking lines and blade geometry under varying operating conditions. To reduce computational costs, a multi-fidelity network is employed to model the relationship between low- and high-fidelity data for predicting the intermediate physical feature of spanwise distributions of total pressure loss. Geometric and aerodynamic features are linearly transformed before being input into the transformer network to extract more informative representations, thereby enhancing the accuracy of a multilayer perceptron (MLP). Multiple transformer-based probabilistic neural networks are ensembled to estimate predictive uncertainty, which improves model robustness and extends its applicability to unseen data. Results show that the transformer-based models improve MLP performance in predicting both the spanwise distribution of total pressure loss and optimal stacking lines. The ensemble model exhibits high uncertainty in out-of-domain predictions, effectively capturing potential large prediction errors. Using a representative low-pressure turbine stage as a benchmark, the proposed method significantly reduces endwall secondary flows, resulting in a 0.61 ± 0.11% increase in stage efficiency compared to the baseline design, thereby validating the effectiveness of the approach.

燃气轮机是应用广泛的能量转换装置，二次流对燃气轮机的综合效率有重要影响。通过扫斜调节堆垛线是控制二次流的重要方法。传统的堆叠线设计方法通常依赖于设计师的经验和迭代过程，这是耗时的，计算昂贵的，并且缺乏通用的设计指南。为了解决这些挑战，本文提出了一种数据驱动的堆叠线设计方法，该方法将变压器架构与深度集成（DE）学习集成在一起，以模拟不同运行条件下最佳堆叠线与叶片几何形状之间的关系。为了降低计算成本，采用多保真度网络对低保真度数据和高保真度数据之间的关系进行建模，预测全压损失沿程分布的中间物理特征。几何和空气动力学特征在输入到变压器网络之前进行线性变换，以提取更多的信息表示，从而提高多层感知器（MLP）的精度。将多个基于变压器的概率神经网络集成来估计预测不确定性，提高了模型的鲁棒性，扩展了模型对未知数据的适用性。结果表明，基于变压器的模型在预测总压损失的展向分布和最优叠加线方面都提高了MLP的性能。集成模型在域外预测中具有很高的不确定性，可以有效地捕获潜在的较大预测误差。以一个具有代表性的低压涡轮级为基准，该方法显著减少了端壁二次流，与基线设计相比，该方法的级效率提高了0.61±0.11%，从而验证了该方法的有效性。

{"title":"Uncertainty-aware data-driven three-dimensional turbine aerodynamic design system with transformer and multi-fidelity neural networks","authors":"Peng Ren, Xiangjun Fang, Junfeng Chen","doi":"10.1016/j.engappai.2026.114125","DOIUrl":"10.1016/j.engappai.2026.114125","url":null,"abstract":"<div><div>Gas turbines are widely used energy conversion devices, and secondary flows have a significant impact on their overall efficiency. Adjusting the stacking line through sweep and lean is an important method for controlling secondary flows. Traditional stacking line design methods typically rely on designers' experience and iterative processes, which are time-consuming, computationally expensive, and lack generalizable design guidelines. To address these challenges, this paper proposes a data-driven stacking line design method that integrates a transformer architecture with Deep Ensemble (DE) learning to model the relationship between optimal stacking lines and blade geometry under varying operating conditions. To reduce computational costs, a multi-fidelity network is employed to model the relationship between low- and high-fidelity data for predicting the intermediate physical feature of spanwise distributions of total pressure loss. Geometric and aerodynamic features are linearly transformed before being input into the transformer network to extract more informative representations, thereby enhancing the accuracy of a multilayer perceptron (MLP). Multiple transformer-based probabilistic neural networks are ensembled to estimate predictive uncertainty, which improves model robustness and extends its applicability to unseen data. Results show that the transformer-based models improve MLP performance in predicting both the spanwise distribution of total pressure loss and optimal stacking lines. The ensemble model exhibits high uncertainty in out-of-domain predictions, effectively capturing potential large prediction errors. Using a representative low-pressure turbine stage as a benchmark, the proposed method significantly reduces endwall secondary flows, resulting in a 0.61 ± 0.11% increase in stage efficiency compared to the baseline design, thereby validating the effectiveness of the approach.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"169 ","pages":"Article 114125"},"PeriodicalIF":8.0,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175305","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 framework integrating data-driven and computational fluid dynamics simulation for continuous blast furnace monitoring 高炉连续监测数据驱动与计算流体力学模拟相结合的框架

IF 8 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2026-02-10 DOI: 10.1016/j.engappai.2026.114092

Xin Wang, Xiao-Yu Tang, Zheng Hao , Kunwei Lin, Chunjie Yang, Wenhai Wang

Due to the complex and nonlinear characteristics of blast furnace (BF) systems, conventional data/mechanism-driven modeling methods have been facing challenges in continuously on-field BF internal state monitoring. Mechanism-driven computational fluid dynamics (CFD) simulations, while interpretable, have high computational costs that prevent real-time application. Conversely, data-driven methods often ignore the coupling relationships between variables and fail to provide a comprehensive understanding of the BF's internal state, which hinders precise control. In response to the aforementioned issues, this paper proposes a novel state continuous monitoring framework that incorporates data-driven with offline pre-calculated CFD simulations. It decouples the intricate and nonlinear BF operation process into a set of interpretable sub-modes, modeling each sub-mode via numerical simulations, then reconstructs the real-time BF transient by properly selecting and fusing some sub-modes. In the offline stage, the particle swarm optimization (PSO) algorithm is employed to obtain the sub-modes that best represent BF operational states. The CFD simulation is then conducted for multiple physical field states corresponding to each sub-mode. For online monitoring, a multi-mode fusion strategy (MMFS) is designed to achieve the real-time BF transient modeling. Application in a BF in South China validates the effectiveness of the proposed framework.

由于高炉系统的复杂性和非线性特性，传统的数据/机制驱动建模方法在高炉内部状态的连续现场监测中面临挑战。机制驱动的计算流体动力学（CFD）模拟虽然具有可解释性，但计算成本高，不利于实时应用。相反，数据驱动的方法往往忽略变量之间的耦合关系，不能提供对BF内部状态的全面理解，这阻碍了精确的控制。针对上述问题，本文提出了一种新的状态连续监测框架，该框架将数据驱动与离线预计算CFD模拟相结合。该方法将复杂的非线性高炉运行过程解耦为一组可解释的子模式，通过数值模拟对每个子模式进行建模，然后通过适当选择和融合子模式重建高炉实时瞬态。在离线阶段，采用粒子群优化（PSO）算法获得最能代表高炉运行状态的子模式。然后对每个子模对应的多个物理场状态进行CFD模拟。针对在线监测，设计了一种多模融合策略，实现了高炉瞬态实时建模。在华南某高炉的应用验证了该框架的有效性。

{"title":"A framework integrating data-driven and computational fluid dynamics simulation for continuous blast furnace monitoring","authors":"Xin Wang, Xiao-Yu Tang, Zheng Hao , Kunwei Lin, Chunjie Yang, Wenhai Wang","doi":"10.1016/j.engappai.2026.114092","DOIUrl":"10.1016/j.engappai.2026.114092","url":null,"abstract":"<div><div>Due to the complex and nonlinear characteristics of blast furnace (BF) systems, conventional data/mechanism-driven modeling methods have been facing challenges in continuously on-field BF internal state monitoring. Mechanism-driven computational fluid dynamics (CFD) simulations, while interpretable, have high computational costs that prevent real-time application. Conversely, data-driven methods often ignore the coupling relationships between variables and fail to provide a comprehensive understanding of the BF's internal state, which hinders precise control. In response to the aforementioned issues, this paper proposes a novel state continuous monitoring framework that incorporates data-driven with offline pre-calculated CFD simulations. It decouples the intricate and nonlinear BF operation process into a set of interpretable sub-modes, modeling each sub-mode via numerical simulations, then reconstructs the real-time BF transient by properly selecting and fusing some sub-modes. In the offline stage, the particle swarm optimization (PSO) algorithm is employed to obtain the sub-modes that best represent BF operational states. The CFD simulation is then conducted for multiple physical field states corresponding to each sub-mode. For online monitoring, a multi-mode fusion strategy (MMFS) is designed to achieve the real-time BF transient modeling. Application in a BF in South China validates the effectiveness of the proposed framework.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"169 ","pages":"Article 114092"},"PeriodicalIF":8.0,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175223","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 and disentangled image editing by manipulating the semantic latent space in diffusion models 利用扩散模型中的语义潜在空间进行可解释和解纠缠的图像编辑

IF 8 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2026-02-10 DOI: 10.1016/j.engappai.2026.114149

Tian Qiu , Qianmu Li

Diffusion models (DMs) have gained prominence in image manipulation, surpassing traditional methods like Variational Autoencoders (VAEs) and Generative Adversarial Networks (GANs) in quality and realism. However, performing precise attribute manipulation and interpretability directly in their high-dimensional Gaussian noise space

X

remains challenging. This limitation motivates a more controllable and interpretable editing framework for practical attribute-level manipulation. To overcome this, we propose leveraging an intermediate semantic latent space

H

along with generated semantic attention masks, achieving efficient and high-fidelity image editing with enhanced disentanglement. Our method facilitates targeted editing in the spatial domain, enabling the modification of specific attributes without affecting unrelated regions of the image. Specifically, we develop a remapper network to map textual prompt embeddings into semantic latent representations within

H

, ensuring editing operations closely align with textual prompts. To further improve the disentanglement and editing efficiency, we design an attention module with three different attention mask strategies applied to the adjusted latent representation. The attention mask intuitively explains the area that DMs focus on during image editing at each time step. We conduct extensive experiments on a variety of datasets, including human faces, dogs, and oil paintings. Both qualitative and quantitative results demonstrate the superiority of our approach over state-of-the-art diffusion-based editing baselines in terms of editing quality, target alignment, and reduced non-target drift.

扩散模型（dm）在图像处理方面取得了突出的成就，在质量和真实感方面超越了传统的方法，如变分自编码器（VAEs）和生成对抗网络（GANs）。然而，在其高维高斯噪声空间X中直接执行精确的属性操作和可解释性仍然具有挑战性。这个限制激发了一个更可控和可解释的编辑框架，用于实际的属性级操作。为了克服这个问题，我们提出利用中间语义潜在空间H和生成的语义注意掩模，通过增强的解纠缠实现高效和高保真的图像编辑。我们的方法便于在空间域中进行有针对性的编辑，可以在不影响图像无关区域的情况下修改特定属性。具体来说，我们开发了一个重标注网络，将文本提示嵌入映射到H中的语义潜在表示，确保编辑操作与文本提示紧密一致。为了进一步提高解纠缠和编辑效率，我们设计了一个注意模块，将三种不同的注意掩模策略应用于调整后的潜在表征。注意遮罩直观地解释了dm在每个时间步的图像编辑过程中关注的区域。我们在各种数据集上进行了广泛的实验，包括人脸、狗和油画。定性和定量结果都证明了我们的方法在编辑质量、目标对准和减少非目标漂移方面优于最先进的基于扩散的编辑基线。

{"title":"Interpretable and disentangled image editing by manipulating the semantic latent space in diffusion models","authors":"Tian Qiu , Qianmu Li","doi":"10.1016/j.engappai.2026.114149","DOIUrl":"10.1016/j.engappai.2026.114149","url":null,"abstract":"<div><div>Diffusion models (DMs) have gained prominence in image manipulation, surpassing traditional methods like Variational Autoencoders (VAEs) and Generative Adversarial Networks (GANs) in quality and realism. However, performing precise attribute manipulation and interpretability directly in their high-dimensional Gaussian noise space <span><math><mi>X</mi></math></span> remains challenging. This limitation motivates a more controllable and interpretable editing framework for practical attribute-level manipulation. To overcome this, we propose leveraging an intermediate semantic latent space <span><math><mi>H</mi></math></span> along with generated semantic attention masks, achieving efficient and high-fidelity image editing with enhanced disentanglement. Our method facilitates targeted editing in the spatial domain, enabling the modification of specific attributes without affecting unrelated regions of the image. Specifically, we develop a remapper network to map textual prompt embeddings into semantic latent representations within <span><math><mi>H</mi></math></span>, ensuring editing operations closely align with textual prompts. To further improve the disentanglement and editing efficiency, we design an attention module with three different attention mask strategies applied to the adjusted latent representation. The attention mask intuitively explains the area that DMs focus on during image editing at each time step. We conduct extensive experiments on a variety of datasets, including human faces, dogs, and oil paintings. Both qualitative and quantitative results demonstrate the superiority of our approach over state-of-the-art diffusion-based editing baselines in terms of editing quality, target alignment, and reduced non-target drift.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"169 ","pages":"Article 114149"},"PeriodicalIF":8.0,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175221","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Long-term cooperative path planning for stratospheric airships based on hierarchical multi-agent reinforcement learning 基于分层多智能体强化学习的平流层飞艇长期协同路径规划

IF 8 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2026-02-10 DOI: 10.1016/j.engappai.2026.114156

Chao Lv , Ming Zhu , Xiao Guo , Jiajun Ou , Baojin Zheng , Liran Sun

Stratospheric airships are increasingly used for long-term collaborative tasks, requiring efficient path planning for multiple airships. Traditional methods struggle with collaborative optimization and state space explosion in such tasks. To address these issues, this paper presents a hierarchical cooperative airship path planning (HiCAPP). This HiCAPP employs a dual-layer control architecture, with the high-level controller responsible for task allocation and the low-level controller concentrating on path planning. Experimental results show that HiCAPP outperforms traditional multi-agent reinforcement learning methods in two critical metrics: average remaining energy and average distance to the task center. Additionally, through experiments with varying numbers of agents, task durations, and disturbances, HiCAPP has demonstrated robustness and scalability. These results confirm its effectiveness in long-term cooperative monitoring tasks and highlight the advantages of hierarchical decision-making in multi-agent systems.

平流层飞艇越来越多地用于长期协同任务，这需要多个飞艇进行有效的路径规划。在这类任务中，传统方法难以解决协同优化和状态空间爆炸问题。为了解决这些问题，本文提出了一种分层协同飞艇路径规划方法。该HiCAPP采用双层控制架构，高层控制器负责任务分配，低层控制器专注于路径规划。实验结果表明，HiCAPP在平均剩余能量和到任务中心的平均距离两个关键指标上优于传统的多智能体强化学习方法。此外，通过不同数量的代理、任务持续时间和干扰的实验，HiCAPP已经证明了鲁棒性和可扩展性。这些结果证实了该方法在长期协同监测任务中的有效性，突出了分层决策在多智能体系统中的优势。

{"title":"Long-term cooperative path planning for stratospheric airships based on hierarchical multi-agent reinforcement learning","authors":"Chao Lv , Ming Zhu , Xiao Guo , Jiajun Ou , Baojin Zheng , Liran Sun","doi":"10.1016/j.engappai.2026.114156","DOIUrl":"10.1016/j.engappai.2026.114156","url":null,"abstract":"<div><div>Stratospheric airships are increasingly used for long-term collaborative tasks, requiring efficient path planning for multiple airships. Traditional methods struggle with collaborative optimization and state space explosion in such tasks. To address these issues, this paper presents a hierarchical cooperative airship path planning (HiCAPP). This HiCAPP employs a dual-layer control architecture, with the high-level controller responsible for task allocation and the low-level controller concentrating on path planning. Experimental results show that HiCAPP outperforms traditional multi-agent reinforcement learning methods in two critical metrics: average remaining energy and average distance to the task center. Additionally, through experiments with varying numbers of agents, task durations, and disturbances, HiCAPP has demonstrated robustness and scalability. These results confirm its effectiveness in long-term cooperative monitoring tasks and highlight the advantages of hierarchical decision-making in multi-agent systems.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"169 ","pages":"Article 114156"},"PeriodicalIF":8.0,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175310","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Green supplier evaluation based on ISO14001 standard via entropy-integrated proximity indexed value method under Linear Diophantine fuzzy sets 基于ISO14001标准的线性丢番图模糊集下熵积分接近指标值法绿色供应商评价

IF 8 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2026-02-10 DOI: 10.1016/j.engappai.2026.114049

Sait Gül , Ali Aydoğdu , Umut Hulusi İnan

The linear Diophantine fuzzy set (LDFS) incorporates two reference parameters, thereby enabling a more comprehensive representation of human judgment. This structure provides flexibility, as the decision-maker can adjust the meaning of the reference parameters to reflect changes in the decision context. Among the many applications of fuzzy sets, information measures such as distance and entropy are particularly significant. Entropy is widely employed in objective attribute-weighting procedures of Multiple Attribute Decision Making (MADM) applications, as it captures the intrinsic information content of attributes. So, the first contribution of this study is the development of a new entropy measure for LDFS. Besides, the second contribution is the extension of the Proximity Indexed Value (PIV) method into the LDFS framework, marking the first proposal of LDF-oriented PIV in the literature. PIV was selected due to its flexibility, ease of application, and the proven strength against the rank reversal phenomenon. The proposed entropy measure is integrated into the attribute-weighting procedure of this new LDF-En-PIV extension. The third contribution is an application-oriented decision model for the selection of green suppliers with respect to their performance in building and employing an ISO14001 Environmental Management System. In this case study, four green supplier alternatives were evaluated across the main components of ISO14001 by a panel of experienced industry experts, with rankings obtained through the LDF-En-PIV approach. The robustness of the proposed approach was presented through comparative analyses with crisp PIV and LDF-ARAS. All comparisons yield consistent rankings, demonstrating the reliability of the proposed approach.

线性丢芬图模糊集（LDFS）包含两个参考参数，从而能够更全面地表示人类的判断。这种结构提供了灵活性，因为决策者可以调整参考参数的含义以反映决策上下文中的变化。在模糊集的众多应用中，距离和熵等信息度量尤为重要。熵是多属性决策（MADM）应用中广泛应用的客观属性加权过程，它捕获属性的内在信息内容。因此，本研究的第一个贡献是为LDFS开发了一种新的熵测度。此外，第二个贡献是将邻近索引值（PIV）方法扩展到LDFS框架中，标志着文献中首次提出面向ldf的PIV。PIV之所以被选中，是因为它的灵活性，易于应用，以及对等级反转现象的证明强度。提出的熵测度被整合到新的LDF-En-PIV扩展的属性加权过程中。第三个贡献是基于绿色供应商在建立和采用ISO14001环境管理体系方面的表现，为选择绿色供应商提供了面向应用的决策模型。在本案例研究中，由经验丰富的行业专家组成的小组对ISO14001的主要组成部分进行了四个绿色供应商替代方案的评估，并通过LDF-En-PIV方法获得了排名。通过与crisp PIV和LDF-ARAS的对比分析，证明了该方法的鲁棒性。所有的比较产生一致的排名，证明了所提出的方法的可靠性。

{"title":"Green supplier evaluation based on ISO14001 standard via entropy-integrated proximity indexed value method under Linear Diophantine fuzzy sets","authors":"Sait Gül , Ali Aydoğdu , Umut Hulusi İnan","doi":"10.1016/j.engappai.2026.114049","DOIUrl":"10.1016/j.engappai.2026.114049","url":null,"abstract":"<div><div>The linear Diophantine fuzzy set (LDFS) incorporates two reference parameters, thereby enabling a more comprehensive representation of human judgment. This structure provides flexibility, as the decision-maker can adjust the meaning of the reference parameters to reflect changes in the decision context. Among the many applications of fuzzy sets, information measures such as distance and entropy are particularly significant. Entropy is widely employed in objective attribute-weighting procedures of Multiple Attribute Decision Making (MADM) applications, as it captures the intrinsic information content of attributes. So, the first contribution of this study is the development of a new entropy measure for LDFS. Besides, the second contribution is the extension of the Proximity Indexed Value (PIV) method into the LDFS framework, marking the first proposal of LDF-oriented PIV in the literature. PIV was selected due to its flexibility, ease of application, and the proven strength against the rank reversal phenomenon. The proposed entropy measure is integrated into the attribute-weighting procedure of this new LDF-En-PIV extension. The third contribution is an application-oriented decision model for the selection of green suppliers with respect to their performance in building and employing an ISO14001 Environmental Management System. In this case study, four green supplier alternatives were evaluated across the main components of ISO14001 by a panel of experienced industry experts, with rankings obtained through the LDF-En-PIV approach. The robustness of the proposed approach was presented through comparative analyses with crisp PIV and LDF-ARAS. All comparisons yield consistent rankings, demonstrating the reliability of the proposed approach.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"169 ","pages":"Article 114049"},"PeriodicalIF":8.0,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175220","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Time-optimal path planning for robots via Deep Lagrangian Networks 基于深度拉格朗日网络的机器人时间最优路径规划

IF 8 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2026-02-10 DOI: 10.1016/j.engappai.2026.114174

Pu Yang , Jiaheng Zhang , Zhiman Duan , Donghao Shi , Shaoping Bai , Qinchuan Li

This paper presents a time-optimal path planning (TOPP) method based on Deep Lagrangian Networks (DeLaN) for robot systems, overcoming the complexities of dynamic modeling. In artificial intelligence, we extend the DeLaN model by incorporating Coulomb friction, significantly enhancing dynamic parameter identification, and propose an excitation trajectory optimization method for training. In engineering, we introduce the data-driven DeLaN-TOPP algorithm, which generates time-optimal paths without the need for explicit dynamic modeling. Experimental validation on a 6-prismatic-spherical-spherical (6-PSS) parallel robot and a Franka serial manipulator demonstrates that the DeLaN-TOPP algorithm achieves time-optimal paths comparable to traditional physics-based methods, while substantially reducing modeling complexity and improving scalability across various platforms, showcasing its efficiency and flexibility for real-world robotic applications. Notably, for high-degree-of-freedom systems, generalization may be constrained by sample efficiency in sparse data regimes.

提出了一种基于深度拉格朗日网络（DeLaN）的机器人系统时间最优路径规划（TOPP）方法，克服了动态建模的复杂性。在人工智能方面，我们通过引入库仑摩擦对DeLaN模型进行扩展，显著增强了动态参数辨识能力，并提出了一种用于训练的激励轨迹优化方法。在工程中，我们引入了数据驱动的DeLaN-TOPP算法，该算法无需显式动态建模即可生成时间最优路径。在6-棱镜-球-球（6-PSS）并联机器人和Franka系列机械手上的实验验证表明，DeLaN-TOPP算法实现了与传统基于物理的方法相当的时间最优路径，同时大大降低了建模复杂性，提高了跨各种平台的可扩展性，展示了其在实际机器人应用中的效率和灵活性。值得注意的是，对于高自由度系统，泛化可能受到稀疏数据体系中样本效率的限制。

{"title":"Time-optimal path planning for robots via Deep Lagrangian Networks","authors":"Pu Yang , Jiaheng Zhang , Zhiman Duan , Donghao Shi , Shaoping Bai , Qinchuan Li","doi":"10.1016/j.engappai.2026.114174","DOIUrl":"10.1016/j.engappai.2026.114174","url":null,"abstract":"<div><div>This paper presents a time-optimal path planning (TOPP) method based on Deep Lagrangian Networks (DeLaN) for robot systems, overcoming the complexities of dynamic modeling. In artificial intelligence, we extend the DeLaN model by incorporating Coulomb friction, significantly enhancing dynamic parameter identification, and propose an excitation trajectory optimization method for training. In engineering, we introduce the data-driven DeLaN-TOPP algorithm, which generates time-optimal paths without the need for explicit dynamic modeling. Experimental validation on a 6-prismatic-spherical-spherical (6-PSS) parallel robot and a Franka serial manipulator demonstrates that the DeLaN-TOPP algorithm achieves time-optimal paths comparable to traditional physics-based methods, while substantially reducing modeling complexity and improving scalability across various platforms, showcasing its efficiency and flexibility for real-world robotic applications. Notably, for high-degree-of-freedom systems, generalization may be constrained by sample efficiency in sparse data regimes.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"169 ","pages":"Article 114174"},"PeriodicalIF":8.0,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175225","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 novel rotating machinery fault diagnosis method based on multi-channel correlation strategy image information enhancement 一种基于多通道相关策略的旋转机械故障诊断方法

IF 8 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2026-02-10 DOI: 10.1016/j.engappai.2026.114102

Jiayao Hu , Wennian Yu , Zixu Chen , Quanyi Luo , Qiang Zeng , Xiaoxi Ding

To address the problem of blurred feature information caused by color mapping during the conversion of one-dimensional signals of industrial rotating machinery to pseudo-color image matrices, an image enhancement method based on matrix information expansion via a multi-channel correlation strategy is proposed. Building on the traditional Markov transition field (MTF) pseudo-color image generation method, the study develops a multi-channel correlation Markov transition field (MCMTF) image coding method. Without introducing additional variables, the multi-channel correlation strategy enhances the information representation capability of images through correlation calculations based on the inherent matrix information, while alleviating feature blurring from pseudo-color conversion. In addition, a max pooling and average pooling parallel convolutional neural network (MAPCNN) is developed to strengthen the network model's global feature extraction capability. The proposed method attains 99.51% and 99.88% accuracy on the self-collected sliding bearing and gear datasets, respectively. Comparisons with state-of-the-art image generation methods and classical models prove its superior performance. The study also explains the mechanism behind the improved performance from the perspective of image information entropy. Scalability experiments of the multi-channel correlation strategy applied to various image generation methods demonstrate the universality of this concept. Experiments on the rolling bearing dataset further confirm the performance of the proposed method. For industrial scenarios involving wind turbines, machine tools, and gas turbines, the proposed method provides both an effective diagnostic approach and new insights for the diagnosis of rotating machinery.

针对工业旋转机械一维信号转换为伪彩色图像矩阵时由于颜色映射导致特征信息模糊的问题，提出了一种基于多通道相关策略的矩阵信息展开的图像增强方法。在传统马尔可夫过渡场（MTF）伪彩色图像生成方法的基础上，研究开发了一种多通道相关马尔可夫过渡场（MCMTF）图像编码方法。多通道相关策略在不引入额外变量的情况下，通过基于固有矩阵信息的相关计算增强了图像的信息表示能力，同时减轻了伪色变换带来的特征模糊。此外，开发了一种最大池化和平均池化并行卷积神经网络（MAPCNN），增强了网络模型的全局特征提取能力。该方法在自采集的滑动轴承和齿轮数据集上分别达到99.51%和99.88%的准确率。通过与现有图像生成方法和经典模型的比较，证明了该方法的优越性。本研究还从图像信息熵的角度解释了性能提升的机制。应用于各种图像生成方法的多通道相关策略的可扩展性实验证明了该概念的普遍性。在滚动轴承数据集上的实验进一步验证了该方法的有效性。对于涉及风力涡轮机、机床和燃气轮机的工业场景，所提出的方法为旋转机械的诊断提供了有效的诊断方法和新的见解。

{"title":"A novel rotating machinery fault diagnosis method based on multi-channel correlation strategy image information enhancement","authors":"Jiayao Hu , Wennian Yu , Zixu Chen , Quanyi Luo , Qiang Zeng , Xiaoxi Ding","doi":"10.1016/j.engappai.2026.114102","DOIUrl":"10.1016/j.engappai.2026.114102","url":null,"abstract":"<div><div>To address the problem of blurred feature information caused by color mapping during the conversion of one-dimensional signals of industrial rotating machinery to pseudo-color image matrices, an image enhancement method based on matrix information expansion via a multi-channel correlation strategy is proposed. Building on the traditional Markov transition field (MTF) pseudo-color image generation method, the study develops a multi-channel correlation Markov transition field (MCMTF) image coding method. Without introducing additional variables, the multi-channel correlation strategy enhances the information representation capability of images through correlation calculations based on the inherent matrix information, while alleviating feature blurring from pseudo-color conversion. In addition, a max pooling and average pooling parallel convolutional neural network (MAPCNN) is developed to strengthen the network model's global feature extraction capability. The proposed method attains 99.51% and 99.88% accuracy on the self-collected sliding bearing and gear datasets, respectively. Comparisons with state-of-the-art image generation methods and classical models prove its superior performance. The study also explains the mechanism behind the improved performance from the perspective of image information entropy. Scalability experiments of the multi-channel correlation strategy applied to various image generation methods demonstrate the universality of this concept. Experiments on the rolling bearing dataset further confirm the performance of the proposed method. For industrial scenarios involving wind turbines, machine tools, and gas turbines, the proposed method provides both an effective diagnostic approach and new insights for the diagnosis of rotating machinery.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"169 ","pages":"Article 114102"},"PeriodicalIF":8.0,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175306","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 domain knowledge and cognitive law driven approach to anti-vibration hammer defect detection 领域知识和认知规律驱动的防震锤缺陷检测方法

IF 8 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2026-02-10 DOI: 10.1016/j.engappai.2026.114100

Hang Niu, Xinyu Ge, Xiaoyu Zhao, Ke Yang, Qianming Wang, Yongjie Zhai, Zhedong Hu

The intelligent detection of anti-vibration hammer defects in transmission lines via computer vision is confronted with challenges due to the limited number of defect samples and the high similarity between defect classes. To this end, a domain knowledge and cognitive law driven approach to anti-vibration hammer defect detection is proposed, which integrates a Structural Knowledge and Geometric Feature-driven image generation method (SKGF) with a Cognitive Law-guided Multilevel Progressive target Detection framework (CLMP-Det). The imposition of morphological and tilt angle constraints is incorporated into the SKGF, based on prior knowledge of the anti-vibration hammer’s structure and its tilt angle distribution characteristics. These constraints can guide the generation of artificial anti-vibration hammer samples semantically consistent with the real physical structure and solve the problem of insufficient defective samples. Secondly, CLMP-Det is designed to simulate the human visual cognitive law through a progressive strategy, progressing from ease to difficulty. This strategy includes two sequential phases: preliminary perception and in-depth discrimination, which enhance the model’s capacity to distinguish between the challenging normal and tilt defect categories. The results of the experiment demonstrate that the proposed method significantly improves the overall detection performance of several widely-used detectors. Compared to the baseline model, our approach achieves a 7.1% improvement in mean average precision. Thus, the method’s robust generalization capability and potential for engineering applications are fully validated.

由于缺陷样本数量有限，且缺陷类别之间具有较高的相似性，利用计算机视觉对传输线防震锤缺陷进行智能检测面临着挑战。为此，提出了一种领域知识和认知规律驱动的抗振锤缺陷检测方法，该方法将结构知识和几何特征驱动的图像生成方法（SKGF）与认知规律指导的多层次渐进目标检测框架（CLMP-Det）相结合。基于对抗振锤结构及其倾斜角分布特性的先验知识，将形态和倾斜角约束的施加纳入到SKGF中。这些约束条件可以指导生成语义上与真实物理结构一致的人工抗振锤试样，解决缺陷试样不足的问题。其次，CLMP-Det通过从简单到困难的递进策略来模拟人类视觉认知规律。该策略包括两个连续的阶段：初步感知和深度识别，这增强了模型区分具有挑战性的正常和倾斜缺陷类别的能力。实验结果表明，该方法显著提高了几种常用检测器的整体检测性能。与基线模型相比，我们的方法在平均精度上提高了7.1%。从而充分验证了该方法的鲁棒泛化能力和工程应用潜力。

{"title":"A domain knowledge and cognitive law driven approach to anti-vibration hammer defect detection","authors":"Hang Niu, Xinyu Ge, Xiaoyu Zhao, Ke Yang, Qianming Wang, Yongjie Zhai, Zhedong Hu","doi":"10.1016/j.engappai.2026.114100","DOIUrl":"10.1016/j.engappai.2026.114100","url":null,"abstract":"<div><div>The intelligent detection of anti-vibration hammer defects in transmission lines via computer vision is confronted with challenges due to the limited number of defect samples and the high similarity between defect classes. To this end, a domain knowledge and cognitive law driven approach to anti-vibration hammer defect detection is proposed, which integrates a Structural Knowledge and Geometric Feature-driven image generation method (SKGF) with a Cognitive Law-guided Multilevel Progressive target Detection framework (CLMP-Det). The imposition of morphological and tilt angle constraints is incorporated into the SKGF, based on prior knowledge of the anti-vibration hammer’s structure and its tilt angle distribution characteristics. These constraints can guide the generation of artificial anti-vibration hammer samples semantically consistent with the real physical structure and solve the problem of insufficient defective samples. Secondly, CLMP-Det is designed to simulate the human visual cognitive law through a progressive strategy, progressing from ease to difficulty. This strategy includes two sequential phases: preliminary perception and in-depth discrimination, which enhance the model’s capacity to distinguish between the challenging normal and tilt defect categories. The results of the experiment demonstrate that the proposed method significantly improves the overall detection performance of several widely-used detectors. Compared to the baseline model, our approach achieves a 7.1% improvement in mean average precision. Thus, the method’s robust generalization capability and potential for engineering applications are fully validated.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"169 ","pages":"Article 114100"},"PeriodicalIF":8.0,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175301","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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