Engineering Applications of Artificial Intelligence最新文献

{"title":"A stress-dependent strength neural network model for predicting the true triaxial strength of rocks","authors":"Tianzhi Yao ,&nbsp;Yunpeng Gao ,&nbsp;Jianhai Zhang ,&nbsp;Ru Zhang ,&nbsp;Li Qian ,&nbsp;Qijun Hu ,&nbsp;Xianliang Wang ,&nbsp;Feng Jiang","doi":"10.1016/j.engappai.2025.112057","DOIUrl":"10.1016/j.engappai.2025.112057","url":null,"abstract":"<div><div>Understanding the true triaxial strength of rocks is essential for safe underground engineering, yet existing empirical and data-driven models often fail to capture the nonlinear effects of the intermediate principal stress <em>σ</em>₂. This study proposes a stress-dependent strength neural network (SDSNN) that integrates physically informed constraints, including monotonicity and boundary conditions, as well as an exponential adaptive weighting strategy to balance data and constraint losses. Cohesion and internal friction angle are used as input features, replacing conventional reliance on uniaxial strength. Compared with a purely data-driven neural network (DDNN) and a constraint-aware variant without adaptive weighting (SDSNN#0), SDSNN achieves significantly better predictive accuracy and robustness across test sets, representative rock types, and five-fold cross-validation. In particular, it maintains consistent strength trends and improved stability across six representative rocks with diverse mechanical properties. Importantly, the model maintains high performance even when trained only on low-to-mid <em>σ</em>₂ data and tested on high <em>σ</em>₂ conditions—demonstrating strong generalization under extrapolation. This capability has received relatively limited attention in data-driven models and is particularly valuable in practical scenarios where high-<em>σ</em>₂ test data are limited or difficult to obtain. Furthermore, ablation analysis demonstrates that removing physical constraints leads to a notable decrease in model accuracy, underscoring the importance of incorporating strength variation characteristics into the model. SDSNN shows particular advantage under boundary stress conditions and when facing noisy or sparse datasets, indicating its potential to serve as a robust and interpretable tool for true triaxial strength prediction in geotechnical applications.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"161 ","pages":"Article 112057"},"PeriodicalIF":8.0,"publicationDate":"2025-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144997307","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}

{"title":"Maritime supply chain optimization using robust adversarial reinforcement learning","authors":"Truong Ngoc Cuong ,&nbsp;Sam-Sang You ,&nbsp;Le Ngoc Bao Long ,&nbsp;Hwan-Seong Kim ,&nbsp;Duy Anh Nguyen ,&nbsp;Nguyen Duy Tan","doi":"10.1016/j.engappai.2025.112127","DOIUrl":"10.1016/j.engappai.2025.112127","url":null,"abstract":"<div><div>This study explores novel strategies for analyzing and managing port productivity in a multi-stage supply chain network by integrating synchronization and reinforcement learning (RL) techniques. Current port management systems face issues with nonlinearity, high interdependency, and vulnerability to market disruptions, which might destabilize port operations. To tackle these issues, seaport operations are examined in four stages of implementation: the terminal operator, inland carrier, inland terminal operator, and consignee. Brownian motion is applied to characterize stochastic disruptions in volatile markets, and the port performance under market disruptions is analyzed using nonlinear data analytics. The dynamical analysis reveals that port management systems exhibit highly coupled nonlinear dynamics with a tendency towards instability. The complex nature of maritime port logistics requires innovative strategies to analyze container handling volumes, optimize the strategic planning of port management, and improve overall efficiency. A novel optimal policy for port operations is realized by integrating a deep deterministic policy gradient into robust adversarial deep learning. A deep reinforcement learning algorithm is employed to learn adaptively from historical port-related data and real-time container handling feedback, enabling intelligent strategies to make informed decisions and dynamically adjust policies in response to stochastic disruptions or changing market conditions. Quantitative results demonstrate that the proposed strategy achieves up to 97.78 % operating efficiency despite disturbances, and the adversarial attacks are shown to decrease port productivity by up to 77.78 % in scenarios without robust deep learning support. This study contributes to the growing field of intelligent port operations by paving the way for more adaptive and smart solutions in the maritime supply chain network.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"161 ","pages":"Article 112127"},"PeriodicalIF":8.0,"publicationDate":"2025-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144997367","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}

{"title":"A robust fuzzy twin support vector machine with kernel-target alignment for binary classification","authors":"Deepak Gupta ,&nbsp;Barenya Bikash Hazarika ,&nbsp;Umesh Gupta ,&nbsp;Witold Pedrycz","doi":"10.1016/j.engappai.2025.112189","DOIUrl":"10.1016/j.engappai.2025.112189","url":null,"abstract":"<div><div>Many algorithms similar to the twin version of the support vector machine and their variants have shown better results in the binary classification of nonlinear data points. But in the presence of outlier and noise, these algorithms exhibit low generalization efficiency. To alleviate this challenge, recently proposed, kernel-target alignment based fuzzy least square twin bounded support vector machine (KTA-FLSTBSVM) used fuzzy membership values and is solved using least squares. Inspired by this strategy, for further improvement, we propose a novel approach called decision support kernel-target alignment based fuzzy least square twin bounded support vector machine (DS-KFIFTBSVM). DS-KFIFTBSVM considers the kernelized fuzzy membership values with the regularized twin support vector machine and solves for linear and nonlinear data points using a functional iterative approach. In DS-KFIFTBSVM, the solution is obtained by solving a linearly convergent iterative scheme rather than solving quadratic programming problems. The proposed DS-KFIFTBSVM offers better generalization efficiency, which has been evaluated using both linear and Gaussian kernels, mostly on artificially developed and publicly accessible datasets with diverse dimensionalities. In terms of various performance evaluation metrics, including specificity, precision, false positive rate, rate of misclassification error, F_score, and geometric mean in linear and non-linear cases, DS-KFIFTBSVM outperforms various baseline approaches. It shows the highest accuracy in various datasets, including 98.1884 % for musk dataset (linear kernel) and 100 % for the glass dataset (Gaussian kernel). Further statistical analysis confirms its classification efficiency.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"161 ","pages":"Article 112189"},"PeriodicalIF":8.0,"publicationDate":"2025-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144997304","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}

{"title":"Deep bioinspired evolutionary stacking algorithm for unpaired multimodal cell classification calibration","authors":"Lili Zhao ,&nbsp;Di Xu ,&nbsp;Xueping Tan ,&nbsp;Jinzhao Yang ,&nbsp;Weiping Ding ,&nbsp;Hengde Zhu ,&nbsp;Lichi Zhang ,&nbsp;Qian Wang","doi":"10.1016/j.engappai.2025.112153","DOIUrl":"10.1016/j.engappai.2025.112153","url":null,"abstract":"<div><div>Single-modality deep learning approaches for cell image classification exhibit inherent limitations in informational diversity when processing cross-institutional datasets acquired under varied imaging protocols. In contrast, multimodal cell imaging has emerged as a promising alternative for addressing data heterogeneity through comprehensive information integration. This study introduces a novel multimodal alternate training decision-making architecture based on a stacking algorithm for unpaired multimodal cell classification calibration. The method leverages deep bioinspired evolutionary networks combined with kernel-based support vector machines. Specifically, deep base classifiers incorporating multimodal concepts are derived from heterogeneous cell datasets. Each base classifier employs a bioinspired strategy to perform alternate training between two evolutionary densely connected attention networks. To mitigate class imbalance, where diseased cells are significantly outnumbered by normal cells, we incorporate a Shannon entropy loss term. Finally, multiple kernel-based support vector machines serve as meta classifiers, transforming high-level multimodal concepts into a separable feature space for robust decision-making. Experimental results demonstrate the superiority of our approach over existing algorithms for unpaired multimodal cell image classification. Our findings emphasize the importance of alternate training intra-modality classifiers and inter-modality fusion calibration for accurate and reliable medical image classification. All source code for this work will be publicly available on GitHub.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"161 ","pages":""},"PeriodicalIF":8.0,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144933124","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}

{"title":"Enhanced neural-network-based iterative learning control considering iterative uncertainties for piezoelectric actuated micro-positioning platform","authors":"Miaolei Zhou ,&nbsp;Yulong Sun ,&nbsp;Xiuyu Zhang ,&nbsp;Wei Gao ,&nbsp;Chun-Yi Su","doi":"10.1016/j.engappai.2025.112134","DOIUrl":"10.1016/j.engappai.2025.112134","url":null,"abstract":"<div><div>This research aims to developing a new enhanced data-driven sliding-mode iterative learning control (E-DDSILC) strategy for piezoelectric actuated micro-positioning (PAMP) platforms. For the first time, the analysis demonstrating that errors converge to 0 in E-DDSILC is successfully extended from strictly repetitive systems to systems with non-strictly repetitive initial conditions. This generalization expands the practical application range of E-DDSILC. Simultaneously, iterative uncertainties are considered, which are the major factor affecting the performance of iterative learning control. To address these uncertainties, a diagonal recurrent neural network is employed to fit and compensate for them within a dynamic linearization model, thereby further enhancing the tracking accuracy and practicability of E-DDSILC. Finally, Several experiments are performed on a PAMP platform to compare the developed E-DDSILC method with both classical DDSILC and traditional E-DDSILC schemes. Comparative experimental results prove the superiority of the developed controller.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"161 ","pages":""},"PeriodicalIF":8.0,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144933230","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}

{"title":"Evaluating reinforcement learning-based neural controllers for quadcopter navigation in windy conditions","authors":"Alain Andres ,&nbsp;Aritz D. Martinez ,&nbsp;Sümer Tunçay ,&nbsp;Ignacio Carlucho","doi":"10.1016/j.engappai.2025.112090","DOIUrl":"10.1016/j.engappai.2025.112090","url":null,"abstract":"<div><div>Accurate quadcopter navigation under windy conditions remains challenging for traditional control methods, especially in the presence of unpredictable wind gusts and strict navigational constraints. This paper evaluates Deep Reinforcement Learning (DRL) based controllers under such conditions, analysing the impact of wind domain randomisation, multi-goal training, enhanced state representations with explicit wind information, and the use of temporal data to capture affecting dynamics over time. Experiments in the AirSim simulator across four trajectories — evaluated under both no-wind and windy conditions — demonstrate that DRL-based controllers outperform classical methods, particularly under stochastic wind disturbances. Moreover, we show that training a DRL agent with domain randomisation improves robustness against wind but reduces efficiency in no-wind scenarios. However, incorporating wind information into the agent’s state space enhances robustness without sacrificing performance in wind-free settings. Furthermore, training with stricter waypoint constraints emerges as the most effective strategy, leading to precise trajectories and improved generalisation to wind disturbances. To further interpret the learned policies, we apply Shapley Additive explanations analysis, revealing how different training configurations influence the agent’s feature importance. These findings underscore the potential of DRL-based neural controllers for resilient autonomous aerial systems, highlighting the importance of structured training strategies, informed state representations, and explainability for real-world deployment.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"161 ","pages":""},"PeriodicalIF":8.0,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144933126","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}

{"title":"Energy management system scheduling optimization based on an improved generative adversarial network deep reinforcement learning algorithm","authors":"Weipeng Chao ,&nbsp;Yuanbo Shi ,&nbsp;Yushuai Li ,&nbsp;Meng Liu ,&nbsp;Xiaoling Leng","doi":"10.1016/j.engappai.2025.112129","DOIUrl":"10.1016/j.engappai.2025.112129","url":null,"abstract":"<div><div>Households, as electricity consumers, play a critical role in achieving the carbon peak and carbon neutrality goals. The development of smart grids provides technical support for the efficient integration and distribution of renewable energy, gradually extending to household users. This has led to higher demands for the stability of electricity supply to address the growing electricity demand and the uncertainties associated with renewable energy. To address this, this paper proposes an improved generative adversarial network and an enhanced deep reinforcement learning algorithm to improve the scheduling capability of the energy management system. First, we introduce an improved wasserstein generative adversarial network that combines stochastic differential equations and autocorrelation penalty terms with the generator. The experimental results demonstrate that the proposed method can generate high-quality time series data. The generated data were used to train our scheduling model, effectively enhancing its generalization capability. Secondly, We introduced the Minmax mechanism to address Q-value estimation bias by utilizing multiple Q-networks. This mechanism first divides the target Q-values into several groups, selects the maximum value from each group, and then takes the minimum among these maxima as the final target Q-value. We applied this mechanism to improve deep reinforcement learning algorithms based on multi-Q-value evaluation. Comparison experiments show that this improvement significantly enhances the algorithm’s performance, outperforming traditional algorithms in terms of convergence, volatility, and final rewards. The energy management system demonstrates stronger adaptability when handling uncertainties arising from renewable energy variations, ensuring reliable power supply and achieving balanced energy management.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"161 ","pages":""},"PeriodicalIF":8.0,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144933229","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}

{"title":"Wind power prediction based on hybrid deep learning and Monte Carlo simulation","authors":"Zhiyong Guo ,&nbsp;Qiaoli Han ,&nbsp;Fangzheng Wei ,&nbsp;Wenkai Qi","doi":"10.1016/j.engappai.2025.112082","DOIUrl":"10.1016/j.engappai.2025.112082","url":null,"abstract":"<div><div>This study proposes a hybrid deep learning model combining Variational Mode Decomposition (VMD), Convolutional Neural Networks (CNN), and a Three-Dimensional Gated Neural Network (TGNN) to enhance wind power prediction accuracy. VMD decomposes wind power time series into intrinsic mode functions, CNN extracts deep features, and TGNN captures temporal dependencies with error feedback control. A Monte Carlo simulation based on the Central Limit Theorem is introduced to evaluate predictive uncertainty and interval coverage. Experimental results from three wind farms in China demonstrate that the proposed model significantly outperforms baseline hybrid models. Specifically, it achieves a coverage probability (CP) of 0.887, an average interval width (AIW) of 200.870, and a normalized mean square error (NMSE) of 0.057. Compared with the VMD-CNN-LSTM model, the proposed VMD-CNN-TGNN reduces root mean error (RMSE) by approximately 50%, indicating its superior accuracy, robustness, and practical value in wind power forecasting.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"161 ","pages":""},"PeriodicalIF":8.0,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144933151","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}

{"title":"Real-time cooperative target tracking in cluttered environments using multiple drone swarms with adaptive fuzzy emotional learning","authors":"Lucas William Page ,&nbsp;Vu Phi Tran ,&nbsp;Duy Luan Nguyen","doi":"10.1016/j.engappai.2025.112125","DOIUrl":"10.1016/j.engappai.2025.112125","url":null,"abstract":"<div><div>This paper presents a real-time adaptive trajectory prediction framework for cooperative unmanned aerial vehicle (UAV) swarms engaged in dynamic target tracking within cluttered environments. The proposed system introduces a novel artificial intelligence (AI)-based control architecture combining fuzzy inference, neuro-emotional learning, and distributed multi-agent coordination. At the core of the approach is a Bidirectional Fuzzy Brain Emotional Learning Prediction (BFBEL-P) model, which integrates fuzzy logic and an online adaptive neural structure to enable trajectory forecasting without pre-training or prior knowledge of the environment. From an engineering perspective, this AI model is deployed in UAV swarm navigation, where robust decision-making, predictive coordination, and obstacle avoidance are essential for target interception missions. In contrast to conventional prediction methods, such as curve fitting, nonlinear model predictive control (MPC), and deep learning-based Long Short-Term Memory (LSTM) networks, the BFBEL-P framework offers fast convergence, low computational cost, and high adaptability. The system incorporates multi-threaded data fusion across the swarm to achieve consensus-driven predictions and maintain situational awareness, even under sensor failures or occlusions. Simulation results show that BFBEL-P improves short-term prediction accuracy by 82.2%, reduces prediction time by 15%, and achieves a 100% tracking success rate across benchmark scenarios. These results establish BFBEL-P as a reliable AI technique for distributed control in real-world UAV swarm applications, offering a promising tool for search-and-rescue, surveillance, and defense operations.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"161 ","pages":""},"PeriodicalIF":8.0,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144933123","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}

{"title":"Employing dual-path structure and soft attention mechanism to enhance recognition and classification of wild medicinal licorice in Xinjiang","authors":"Yuan Qin,&nbsp;Jianguo Dai,&nbsp;Guoshun Zhang,&nbsp;Miaomiao Xu,&nbsp;Jing Yang,&nbsp;Jinglong Liu","doi":"10.1016/j.engappai.2025.112126","DOIUrl":"10.1016/j.engappai.2025.112126","url":null,"abstract":"<div><div>Licorice is highly valued in traditional Chinese medicine for its anti-inflammatory, antiviral, and immunomodulatory properties, and is widely used in the pharmaceutical, food, and cosmetic industries. Xinjiang, the largest licorice-producing region in China, faces severe overharvesting of wild licorice due to increasing market demand, threatens natural populations and fragile ecosystems. Accurate identification and classification of licorice species are crucial for environmental protection and sustainable resource utilization, as traditional methods relying on experience are inefficient, subjective, and prone to errors. This study builds on the Inception-Residual Network-Version 2 (Inception-ResNet-V2) architecture and proposes an advanced licorice recognition model called Inception-ResNet-V2-Soft Attention, Dual-path Structure, and Focal Loss (IRV2-SDF). The IRV2-SDF model integrates a soft attention mechanism that focuses on key regions, a dual-path structure for multi-scale feature extraction, and a focal loss function to address class imbalance. It aims to improve the identification and classification of three wild licorice species (<em>Glycyrrhiza glabra</em>, <em>Glycyrrhiza inflata</em>, and <em>Glycyrrhiza uralensis</em>) and associated weeds in complex environments. Trained on 3,653 images collected from Xinjiang, the model achieves an average recognition accuracy of 91.79%, surpassing traditional models, with accuracy improvements of 4.27%, 2.08%, 2.76%, and 6.36% for <em>G. glabra</em>, <em>G. inflata</em>, <em>G. uralensis</em>, and weeds, respectively. By effectively reducing background noise and enhancing detection capabilities, the model overcomes the limitations of traditional methods and provides a robust solution for wild licorice recognition. This research offers a technical foundation for licorice conservation and sustainable utilization and can serve as a reference for identifying other medicinal plants in complex environments.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"161 ","pages":""},"PeriodicalIF":8.0,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144933125","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}