Search Results (342)

Adversarial attacks involve malicious actors introducing intentional perturbations to machine learning (ML) models, causing unintended behavior. This poses a significant threat to the integrity and trustworthiness of ML models, necessitating the development of robust detection techniques to protect systems from potential threats. The paper proposes a new approach for detecting adversarial attacks using a surrogate model and diagnostic attributes. The method was tested on 22 tabular datasets on which four different ML models were trained. Furthermore, various attacks were conducted, which led to obtaining perturbed data. The proposed approach is characterized by high efficiency in detecting known and unknown attacks—balanced accuracy was above 0.94, with very low false negative rates (0.02–0.10) for binary detection. Sensitivity analysis shows that classifiers trained based on diagnostic attributes can detect even very subtle adversarial attacks. Full article

In reconfigurable radio frequency (RF) microsystems, the interconnect structure critically affects high-frequency signal integrity, and the accuracy of electromagnetic (EM) modeling directly determines the overall system performance. Conventional neural network-based surrogate models mainly focus on minimizing numerical errors, while neglecting essential physical constraints, such as causality and passivity, thereby limiting their applicability in both time and frequency domains. This paper proposes a physics-constrained Neuro-Transfer surrogate model with a broadband output architecture to directly predict S-parameters over the 1–50 GHz range. Causality and passivity are enforced through dedicated regularization terms during training. Furthermore, a particle swarm optimization (PSO)-based multi-objective intelligent optimization framework is developed, incorporating fixed-weight normalization and a linearly decreasing inertia weight strategy to simultaneously optimize the S11, S21, and S22 performance of a quasi-coaxial TSV composite structure. Target values are set to −25 dB, −0.54 dB, and −24 dB, respectively. The optimized structural parameters yield prediction-to-simulation deviations below 1 dB, with an average prediction error of 2.11% on the test set. Full article

Background: Hepatocellular carcinoma (HCC) is a highly aggressive and heterogeneous malignancy which predominantly arises in the setting of cirrhosis, and there is lack of models to predict prognosis in cirrhotic HCC. This study aims to develop and validate a prediction model based on the pathomics signature and clinicopathological characteristics to predict the prognosis of HCC with cirrhosis. Methods: In this multicenter, retrospective study, 389 patients were enrolled (training cohort: 268; independent validation cohort: 121). A total of 351 pathomics features were extracted from digital H-E–stained images, and a pathomics signature (PS_HCC) was constructed using a least absolute shrinkage and selection operator Cox regression model. Then two nomograms were established by combining the PS_HCC and clinicopathological characteristics. Further validation was performed in the validation cohort. Results: This study included 389 patients. A 24 feature-based PS_HCC was constructed. A higher PS_HCC was significantly associated with worse OS and DFS in both the training (OS: hazard ratio [HR], 4.341 [95% CI, 3.109–6.062]; DFS: HR, 3.058 [95% CI, 2.223–4.207]) and validation (OS: HR, 4.145 [95% CI, 2.357–7.291]; DFS: HR, 3.395 [95% CI, 2.104–5.479]) cohorts (p < 0.001 for all comparisons). Multivariable analysis revealed that the PS_HCC was an independent factor associated with OS and DFS. Integrating the PS_HCC into pathomics nomograms resulted in better performance for prognosis prediction than the traditional model in both cohorts. Conclusions: The PS_HCC may serve as a reliable surrogate for prognosis, and the nomograms offer promising tools to predict individual outcomes, facilitating personalized management of HCC with cirrhosis. Full article

The credibility of Computational Fluid Dynamics (CFD) has been a topic of debate due to the significant uncertainties inherent in its modeling processes and numerical implementations. Uncertainty Quantification (UQ) offers a scientific framework for quantitatively assessing and mitigating uncertainties in CFD simulations. However, this procedure typically requires numerous CFD simulations and considerable manual effort for both simulation management and data analysis. To overcome these challenges, this work develops a platform called UQ4CFD, a browser–server software that provides automated and customized uncertainty quantification capabilities for CFD studies. The UQ4CFD platform integrates different kinds of methodologies to perform comprehensive uncertainty analysis, including uncertainty propagation, sensitivity analysis, surrogate modeling, numerical discretization uncertainty analysis, model validation, model calibration, etc. A tightly coupled CFD-UQ workflow is built to automate the complete analytical process, encompassing parameter sampling, simulation execution, and results analysis, which significantly improves computational efficiency while reducing risks associated with data processing errors. Comprehensive validation employing both analytical benchmark functions and practical CFD cases has been conducted to demonstrate the platform’s effectiveness and adaptability in diverse UQ scenarios. Full article

The increasing adoption of electric delivery fleets introduces significant challenges related to uneven energy utilization and suboptimal scheduling efficiency. Vehicle-to-Vehicle (V2V) energy sharing presents a promising solution, but its effectiveness critically depends on precise matching and co-optimization within dynamic urban traffic environments. This paper proposes a hierarchical optimization framework to minimize total fleet operational costs, incorporating a comprehensive analysis that includes battery degradation. The core innovation of the framework lies in coupling high-level path planning with low-level real-time speed control. First, a high-fidelity energy consumption surrogate model is constructed through model predictive control simulations, incorporating vehicle dynamics and signal phase and timing information. Second, the spatiotemporal longest common subsequence algorithm is employed to match the spatio-temporal trajectories of energy-provider and energy-consumer vehicles. A battery aging model is integrated to quantify the long-term costs associated with different operational strategies. Finally, a multi-objective particle swarm optimization algorithm, integrated with MPC, co-optimizes the rendezvous paths and speed profiles. In a case study based on a logistics network, simulation results demonstrate that, compared to the conventional station-based charging mode, the proposed V2V framework reduces total fleet operational costs by a net 12.5% and total energy consumption by 17.4% while increasing the energy utilization efficiency of EV-Ps by 21.4%. This net saving is achieved even though the V2V strategy incurs a marginal increase in battery aging costs, which is overwhelmingly offset by substantial savings in logistical efficiency. This study provides an efficient and economical solution for the dynamic energy management of electric fleets under realistic traffic conditions, contributing to a more sustainable and resilient urban logistics ecosystem. Full article

Traffic safety is undergoing a profound transformation, driven by advances in data science, sensing technologies, and computational modeling. Proactive approaches are enabling the early identification of potential hazards, real-time decision-making, and the development of smarter, safer transportation systems. This Special Issue summarizes recent progress in traffic safety assessment, highlighting the application of emerging tools such as machine learning, explainable artificial intelligence, and computer vision. These innovations are used to predict crash risks, evaluate surrogate safety measures, and automate the analysis of behavioral data, contributing to more inclusive and adaptive safety frameworks, particularly for vulnerable road users such as pedestrians and cyclists. The research also addresses key challenges, including data integration across diverse sources, aligning safety metrics with human perception, and ensuring the scalability of models in complex environments. By advancing both technical methodologies and human-centered evaluation, these developments signal a shift toward more intelligent, transparent, and equitable approaches to traffic safety assessment and policy-making. Full article

Homocysteine has long been studied as a potential cardiovascular risk factor due to its biochemical role in endothelial dysfunction, oxidative stress, inflammation, and thrombogenesis. Despite strong epidemiological and mechanistic support, the translation of homocysteine-lowering interventions into clinical benefit remains controversial. This non-systematic review aims to clarify the current understanding of homocysteine in the cardiovascular setting by distinguishing between well-established facts, clinically relevant interventions, and persistent misconceptions. We first revisit the historical emergence of homocysteine as a cardiovascular biomarker and explore its pathophysiological mechanisms, including endothelial damage, atherosclerosis progression, and prothrombotic effects—supported by in vitro and animal model studies. Subsequently, we evaluate evidence-based interventions such as B-vitamin supplementation (folate, B6, B12), lifestyle modifications, and the clinical relevance of homocysteine monitoring in specific populations (e.g., MTHFR mutations, chronic kidney disease). We then discuss common pitfalls, including the overinterpretation of genetic variants, the inappropriate use of supplementation, and the overreliance on surrogate biomarkers in clinical trials. Although elevated homocysteine remains a reproducible biomarker of cardiovascular risk, current evidence does not support routine intervention in unselected populations. A precision medicine approach—targeting high-risk subgroups and integrating homocysteine into broader cardiometabolic management—may help unlock its therapeutic relevance. Future pharmacological strategies should prioritize mechanistic insight, patient stratification, and clinically meaningful endpoints. Full article

Angle crack defects significantly affect compressor blade radial deformation characteristics, posing critical challenges for reliability assessment under operational uncertainties. This study proposes a novel osprey optimization algorithm (OOA)-optimized Kriging and radial basis function (RBF) method (OOA-KR) for the efficient reliability evaluation of blade radial clearance with angle crack defects. The approach integrates Kriging’s uncertainty quantification capabilities with RBF neural networks’ nonlinear mapping strengths through an adaptive weighting scheme optimized by OOA. Multiple uncertainty sources including crack geometry, operational temperature, and loading conditions are systematically considered. A comprehensive finite element model incorporating crack size variations and multi-physics coupling effects generates training data for surrogate model construction. Comparative studies demonstrate superior prediction accuracy with RMSE = 0.568 and R² = 0.8842, significantly outperforming conventional methods while maintaining computational efficiency. Reliability assessment achieves 97.6% precision through Monte Carlo simulation. Sensitivity analysis reveals rotational speed as the most influential factor (S = 0.42), followed by temperature and loading parameters. The proposed OOA-KR method provides an effective tool for blade design optimization and reliability-based maintenance strategies. Full article

In addressing the challenges of low pelletization qualification rates, poor uniformity between seeds and powder, and difficulties in optimizing equipment parameters for agropyron seed coating, this paper integrates numerical simulation with experimental verification to optimize the working parameters of pelletization coating. The study investigates the impact of vibration frequency, vibration direction, vibration amplitude, rotational speed, and inclination angle on seed-powder mixing uniformity and single-seed pellet qualification rates through both physical experiments and simulation tests. The study found that the coefficient of variation obtained through discrete element simulation can serve as a reliable surrogate indicator for evaluating pelletization coating quality, with its variation trend highly consistent with the single-seed pellet qualification rate observed in physical experiments. A secondary regression orthogonal design experiment used these indicators to establish a second-order regression equation, thereby performing single-objective optimization of the regression model. The results showed that the relative errors between simulation and physical test parameters were 1.24% for vibration frequency, 1.08% for coating pan rotational speed, and 0.17% for coating pan inclination angle. This demonstrates the high reliability of the coefficient of variation as a surrogate indicator for pellet qualification. With the optimized parameters, the qualification rate of single-seed pellets for agropyron seeds reached 95.3%, and the relative error between model predictions and physical tests was 1.7%. These findings validate the use of the second-order regression equation for predicting and analyzing single-seed pellet qualification rates and provide valuable insights for designing small-grain forage seed pelletization coating machines and optimizing coating parameters. Full article

This review critically examines the intersection of machine learning (ML), plate heat exchangers (PHEs), and latent heat thermal energy storage (LHTES) using phase-change materials (PCMs)—a combination not comprehensively addressed in the existing literature. Covering more than 120 peer-reviewed studies published between 2015 and 2025, we analyze the deployment of ML methods—including artificial neural networks, ensemble models, physics-informed neural networks, and hybrid optimization techniques—for modeling, performance enhancement, and real-time control of PCM-integrated PHE systems. Particular attention is given to ML-driven geometry optimization, flow prediction, and surrogate modeling for computational fluid dynamics (CFD) simulations. The review also explores digital twin development and nanofluid-enhanced storage strategies. By addressing key gaps in dataset availability, model interpretability, and integration challenges, we provide a structured roadmap for future research, emphasizing hybrid ML–physics models, explainable AI, and standardized benchmarking. This work offers a data-driven and focused perspective on advancing the design of intelligent and sustainable thermal systems. Full article

Flow field design in proton exchange membrane fuel cells (PEMFCs) is a critical issue, as it plays an important role in governing reactant transport dynamics and cell performance. In this work, numerical studies of a single Tesla-valve flow field were conducted. The influence of loop radius, channel angle, and channel height on the performance of PEMFCs were fully explored. Then, aiming to maximize the output current density, this study optimized the Tesla-valve flow field configuration through a framework that integrates Gaussian process modeling with a Genetic Algorithm (GA). The approach efficiently identifies the optimal geometric parameters, highlighting effective synergy between the surrogate model and intelligent evolutionary optimization for enhanced performance. Simulation results show that the current density at 0.4 V and the highest power density have been improved by more than 10% compared to the baseline design for both forward and reverse flow. The optimized Tesla valve design has been compared with conventional parallel and serpentine flow fields of the same flow area. Results show that, despite the larger pressure drop for the single channel case—which is due to the insufficient length of the serpentine channel—the Tesla-valve flow field demonstrated superior performance in other metrics, including current and power density, under both flow directions. Full article

Reducing hydrodynamic resistance remains a central concern in modern ship design. The Simulation-Based Design technique offers high-fidelity optimization through computational fluid dynamics, but this comes at the cost of computational efficiency. In contrast, surrogate models trained offline can accelerate the process but often compromise on accuracy. To address this issue, this study proposes a hybrid optimization framework connecting a computational fluid dynamics solver and a convolutional neural network surrogate model within a dual-mode mechanism. By comparing selected computational fluid dynamics evaluations with surrogate predictions during each iteration, the system is able to balance the precision and efficiency adaptively. The framework integrates a particle swarm optimizer, a free-form deformation modeler, and a dual-mode solver. Case studies on three benchmark hulls including KCS, KVLCC1, and JBC have shown 3.40%, 3.95%, and 2.74% resistance reduction, respectively, with computation efficiency gains exceeding 44% compared to the traditional Simulation-Based Design process using full computational fluid dynamics. This study provides a practical attempt to enhance the efficiency of hull form optimization while maintaining accuracy. Full article

This paper surveys the application of machine learning in fiber composite manufacturing, highlighting its role in adaptive process control, defect detection, and real-time quality assurance. First, the need for ML in composite processing is highlighted, followed by a review of data-driven approaches—including predictive modeling, sensor fusion, and adaptive control—that address material heterogeneity and process variability. An in-depth analysis examines six case studies, among which are XPBD-based surrogates for RL-driven robotic draping, hyperspectral imaging (HSI) with U-Net segmentation for adhesion prediction, and CNN-driven surrogate optimization for variable-geometry forming. Building on these insights, a hybrid AI model architecture is proposed for natural-fiber composites, integrating a physics-informed GNN surrogate, a 3D Spectral-UNet for defect segmentation, and a cross-attention controller for closed-loop parameter adjustment. Validation on synthetic data—including visualizations of HSI segmentation, graph topologies, and controller action weights—demonstrates end-to-end operability. The discussion addresses interpretability, domain randomization, and sim-to-real transfer and highlights emerging trends such as physics-informed neural networks and digital twins. This paper concludes by outlining future challenges in small-data regimes and industrial scalability, thereby providing a comprehensive roadmap for ML-enabled composite manufacturing. Full article

With the rapid advancement of intelligent technologies, permanent magnet synchronous motor (PMSM) servo systems have seen increasing applications in industrial fields, accompanied by continuously rising control performance demands. Moreover, the adjustment of controller parameters is pivotal for the performance optimization of servo systems. This paper presents an optimization method for PMSM servo systems based on the coupling technique of the neural network surrogate model and intelligent optimization algorithm. A hybrid model is constructed by the proposed method, integrating a mathematical model based on transfer functions with an artificial neural network surrogate model, which is employed to compensate for the discrepancies between the mathematical model and the actual measured values. The accuracy and superiority of the hybrid model are comprehensively validated through training and validation loss analysis, fitting plot construction, and ablation experiments. Subsequently, based on the hybrid model, the qualitative and quantitative comparative analysis of the Pareto fronts of five commonly used multi-objective intelligent optimization algorithms is conducted. The optimal algorithm is determined through experimental validation of the optimization results to obtain the optimal result. The optimal result demonstrates that, compared to the initial result before optimization, the overshoot is reduced by 89.7%, and the settling time is reduced by 80.1%. Additionally, several other non-dominated solutions are available for selection, and all optimized results are superior to the initial result. This study provides a novel idea and method for the performance optimization of PMSM servo systems, contributing to the field with a robust and effective approach to enhance system control performance. Full article

This work introduces a unified Artificial Intelligence-based framework for the optimal tuning of gains in a neural discrete-time sliding mode controller (SMC) applied to a two-degree-of-freedom robotic manipulator. The novelty lies in combining surrogate-assisted optimization with normalized search spaces to enable a fair comparative analysis of three metaheuristic strategies: Bayesian Optimization (BO), Particle Swarm Optimization (PSO), and Genetic Algorithms (GAs). The manipulator dynamics are identified via a discrete-time recurrent high-order neural network (NN) trained online using an Extended Kalman Filter with adaptive noise covariance updates, allowing the model to accurately capture unmodeled dynamics, nonlinearities, parametric variations, and process/measurement noise. This neural representation serves as the predictive plant for the discrete-time SMC, enabling precise control of joint angular positions under sinusoidal phase-shifted references. To construct the optimization dataset, MATLAB^® simulations sweep the controller gains $(k_0^{*},k_1^{*})$ over a bounded physical domain, logging steady-state tracking errors. These are normalized to mitigate scaling effects and improve convergence stability. Optimization is executed in Python^® using integrated scikit-learn, DEAP, and scikit-optimize routines. Simulation results reveal that all three algorithms reach high-performance gain configurations. Here, the combined cost is the normalized aggregate objective $\tilde{J}$ constructed from the steady-state tracking errors of both joints. Under identical experimental conditions (shared data loading/normalization and a single Python pipeline), PSO attains the lowest error in Joint 1 ($7.36\times {10}^{-5}$ rad) with the shortest runtime (23.44 s); GA yields the lowest error in Joint 2 ($8.18\times {10}^{-3}$ rad) at higher computational expense (≈69.7 s including refinement); and BO is competitive in both joints ($7.81\times {10}^{-5}$ rad, $8.39\times {10}^{-3}$ rad) with a runtime comparable to PSO (23.65 s) while using only 50 evaluations. Full article