The current digital modeling and cross-platform dissemination of intangible cultural heritage elements face the challenges of low efficiency and accuracy. Existing methods exhibit significant limitations in terms of model fidelity, terminal adaptability, and the effectiveness of feedback loops. In this study, an intelligent modeling and cross-platform collaborative dissemination mechanism based on digital twins (DTs) is proposed to achieve high-fidelity reconstruction, lightweight optimization, and dynamic iteration. An initial DT model is constructed by integrating high-precision 3D scanning and computer vision technologies. A lightweight convolutional neural network, combined with a joint geometry−texture loss function, is designed to automatically inpaint model details and optimize quality. The GL transmission format encapsulation and an API gateway are used to establish data transmission channels, enabling seamless integration of mobile devices, virtual museums, and social media platforms. A personalized recommendation algorithm driven by a graph neural network is introduced, which integrates user behavior data to deliver precise content recommendations. A real-time feedback mechanism based on Kafka and Spark Streaming is constructed to transmit dissemination effect data back to the modeling and recommendation modules, forming a dynamic optimization loop. The experimental results show that this method reduces the Hausdorff distance of the paper-cutting intangible cultural heritage model from 0.858 to 0.631 mm and reduces the cross-platform synchronization delay on entry-level devices from 987 to 512 ms. The proposed method provides a scalable technical path for the intelligent inheritance and global dissemination of intangible cultural heritage.
Exposed to complex environments for extended periods, ultra-high-voltage transmission lines now face lightning strikes as the primary natural disaster responsible for unplanned outages. The ability to accurately estimate the risk of trip‒outs caused by lightning is of great importance for ensuring power grid safety. However, existing methods have problems such as poor adaptability in signal feature extraction and insufficient support for mixed variables in hyperparameter optimization. Therefore, this paper proposes a Bayesian simulation and parameter optimization method for lightning tripping risks in ultra-high-voltage power grids. First, an adaptive wavelet multi-scale feature selection mechanism is designed to achieve precise decoupling of transient impulse components and background noise. Second, a Bayesian Transformer prediction network is constructed to quantify aleatoric uncertainty and cognitive uncertainty. Finally, the hyperparameters of the prediction model are optimized based on the improved Bayesian optimization algorithm. Genetic algorithms and multi-fidelity Gaussian process surrogate models are introduced to reduce the simulation optimization cost. Experimental outcomes indicate that the prediction accuracy and F1 score of the suggested approach are at least 6.35% and 9.1% higher than those of the benchmark method, respectively, verifying its effectiveness in lightning tripping risk prediction.
Complex wind farm environments cause severe spatial aliasing and signal attenuation in blade acoustic signature monitoring. This paper presents an acoustic sensor array topology optimization method based on multi-physics simulation for high-fidelity acquisition of weak voiceprint features. A three-dimensional sound field model coupling aerodynamic noise and mechanical vibration quantifies sound propagation under varying wind speeds and yaw angles. A heuristic particle swarm algorithm discretely optimizes microphone array coordinates on tower and nacelle surfaces by maximizing the signal-to-noise ratio. A surrogate model accelerates sound field evaluation while eliminating nodes disturbed by strong wind vortices. Experiments on a public blade crack acoustic dataset show that the optimized array reduces normalized root mean square error by 44.4 percent compared to conventional spiral arrays, and the proposed attention-based multi-scale network achieves an area under the curve of 0.967 and recall of 0.913.
Accurate and reliable categorization of lung cancer histopathology is considered challenging due to incremental data arrival, distributional shifts, and the tendency of conventional convolutional neural networks to suffer from overfitting and catastrophic forgetting when new subclasses are introduced. To address these issues, a continual learning framework was developed to preserve previously acquired knowledge while maintaining high diagnostic performance. A ResNet50 backbone regularized with Elastic Weight Consolidation (EWC) was employed, in which the Fisher-weighted penalty coefficient (λ) was updated dynamically through a curvature-aware schedule. By doing so, stability and plasticity were balanced adaptively without the need for manual grid search. The framework was trained and evaluated on 15,000 pathologically confirmed lung tissue images comprising adenocarcinoma (aca), squamous cell carcinoma (scc), and benign (n) classes. The experimental setup was organized as a continual learning sequence over three phases of training, where new classes were introduced incrementally while earlier ones were revisited. The proposed approach achieved external test accuracy of 100%, 97.2%, and 98.96% across the three phases, demonstrating its capability to sustain high accuracy while effectively preventing catastrophic forgetting. These results indicate that the integration of adaptive hyperparameter tuning with a curvature-driven EWC schedule can enhance continual learning performance in lung cancer classification. It is concluded that the ResNet50-EWC framework provides a scalable and clinically relevant solution that eliminates the need for full retraining when data expands, and future validation on additional histological subtypes and external whole-slide cohorts is planned.
Urban traffic noise significantly affects the acoustic comfort of landscape spaces, and the scattering interactions between sound waves and complex terrain are difficult to quantify precisely. In response, this study proposes a method oriented toward noise reduction for the optimization of acoustic simulation parameters in landscape spaces. A three-dimensional terrain sound field model is established adopting the multi-domain coupled boundary element approach to accurately solve the Helmholtz equation under complex terrain conditions. The terrain's impedance coefficient is extracted as a key parameter, and a high-precision Kriging response surface surrogate model is constructed using Latin hypercubic sampling to replace time-consuming physical acoustical calculations. By combining a multi-objective particle swarm optimization approach, the terrain elevation parameters are globally optimized under constraints that balance earthwork volume and visual permeability. The natural impedance characteristics of the terrain are utilized to achieve diffuse reflection and absorption of sound waves in space, thereby blocking noise propagation pathways at the source in the physical environment. Experimental outcome indicates that the suggested approach gains an average noise reduction of 4.53 dB while keeping earthwork volume within 195 m3, maintaining a visual permeability of 0.58, and achieving a comprehensive satisfaction score of 0.91, thereby exhibiting high practical value.
This study employs the C++ programming framework and Tecplot software to construct the initial fluidic molecular structure model. A molecular dynamics simulation method is utilized to modify the molecular structure during a prerun process, aiming to establish the natural equilibrium state of molecules that aligns with the Maxwell-Boltzmann distribution plot. The 3D model representing the natural equilibrium fluidic state, developed after the pre-run process, is used to explore fluidic nanojet ejections under various technological parameters. The outcomes of the fluidic nanojets obtained through the simulation process reveal disparities under different research scenarios such as pressing forces, system temperatures, and nozzle diameters. These acquired fluidic nanojets further validate the credibility of the established natural equilibrium state after the 10,000 femtosecond (fs) pre-run process, thus serving as a reliable foundation for subsequent investigations based on this fundamental equilibrium premise. Furthermore, this study provides significant insights into the dynamic ejection of fluidic nanojets through molecular dynamics simulations. Additionally, the generated data are invaluable for experimental endeavors and manufacturing processes incorporating this technology.
Double-corrugated steel shear walls (DCSSWs), made of two connected trapezoidal corrugated plates, were developed to address the limitations of single-CSSWs. In recent years, flat-corrugated steel shear walls (FCSSWs) built from flat and corrugated plates have been introduced to benefit from the advantages of flat and corrugated steel plates. In this exploration, the lateral behavior of the DCSSWs and FCSSWs with different corrugation angles and infill plate thicknesses is investigated under cyclic lateral loading in ABAQUS software. The outcomes indicate that FCSSWs exhibit superior performance compared to DCSSWs due to the interaction effect in the flat and corrugated plates. The maximum strength of the FCSSWs is increased by at least 1.7% and a maximum of 12.0% compared to the DCSSWs. Additionally, energy absorption in DCSSWs is not significantly different from that in FCSSWs, with variations ranging from 0.7% to 4.5%. In some samples, flat-corrugated walls absorb more energy, while in others, double-corrugated walls do. The flat-corrugated system uses 3.4%, 6.8%, and 12.04% less steel at corrugation angles of 30°, 45°, and 60°, respectively, and shows higher maximum strength, making it more economical than the double-corrugated system. The outcomes displayed that with the increase in plate thickness, the equivalent viscous damping increases.
Landscape environments present substantial difficulties for autonomous systems because of issues like vegetation occlusion, rolling terrain, varying light levels, and confusing textures that hinder accurate 3D perception and lead path planners to settle upon local optima or to run slowly. To attack this problem, this paper drinks a stride towards proposing the Landscape Perception-Planning Framework (LPPF), an end-to-end lightweight architecture capable of optimizing perception and planning jointly. LPPF includes a MobileNetV3–Swin Transformer architecture integrated to provide robust monocular depth estimation, construction of StyleGAN2-ADA generated synthetic 3D point clouds in multiple weather conditions for the purposes of generalization, and Proximal Policy Optimization (PPO) planner that dynamically adjusts depth confidence into a cost map for error-aware navigation. LPPF is evaluated using 10,000 synthetic LiDAR frames and 500 real LiDAR frames, achieving an overall score of 0.93, an improvement of 19.2% over DPT using the LPPF framework to process under a 50 ms real-time constraint on an embedded platform. By applying channel pruning and INT8 quantization, the model reduces parameters by 85.2% and increases inference by a factor of 3.21 indicating strong accuracy, robustness, and efficiency for intelligent navigation in complex, resource-constrained landscape environments.
Based on CFD (Computational Fluid Dynamics) simulation technology, this article proposes three key innovations in CFD based aerodynamic analysis of renewable energy wind turbines: firstly, integrating real wind field data processed by WRT (Wind Field Data Processing Technology) into the numerical model, significantly improving the authenticity of the simulation; Secondly, it was verified through system comparison that the SST turbulence model has higher accuracy in simulating axial flow fans compared to the standard kappa − ε model, with an average error controlled within 12%; Finally, the quantitative relationship between terrain undulation and wind turbine performance was revealed through multi scenario simulation, providing important basis for optimizing wind turbine layout in this paper. By constructing a high-precision numerical model and combining with the actual wind field data, we successfully simulated the operation of the fan under different wind speeds, wind directions and terrain conditions, thus comprehensively evaluating its aerodynamic performance. In the simulation process, we focus on the key indicators such as power output, efficiency and wind energy utilization coefficient of the wind turbine. After a large number of data analysis and comparison, we found that CFD is used for numerical simulation and experimental verification of fan. Taking air volume and energy efficiency ratio as verification indexes, the simulated value of turbulence model SST (Shear Stress Transport) is closer to the experimental value, and the average relative error between the simulated value of air volume and the experimental value is 11.9%, and the average relative error between the simulated value of energy efficiency ratio and the experimental value is 12.7%. In addition, with the increase of wind speed, the power output of wind turbine shows an obvious upward trend, but the growth rate gradually slows down, which accords with the general law of wind energy conversion. In addition to the wind speed factors, the geometry of fan blades, the installation angle and the mutual interference between fans are also deeply studied. By adjusting these parameters, we successfully optimize the aerodynamic performance of the fan, so that it can maintain efficient and stable operation under a wider range of wind conditions. In addition, this paper also innovatively studies the aerodynamic performance of the fan under different terrain conditions. We found that the terrain relief, obstacles distribution and other factors will have a significant impact on the performance of the fan. Therefore, in the process of site selection and layout of fans, it is necessary to fully consider terrain factors to ensure the best performance of fans.
Rolling bearings are critical components in rotating machinery, directly influencing system reliability and performance. Bearing failures, driven by stochastic degradation processes, necessitate accurate reliability assessment and maintenance optimization to minimize costs and downtime. This study proposes an adaptive inspection and maintenance model for rolling bearings based on the Delay Time Model (DTM), which captures the two-stage failure process: a normal operating stage until a hidden defect emerges, followed by a delay time until failure. The DTM leverages the failure delay time to schedule preventive maintenance, preventing costly failures. By modeling the defect initiation and delay time distributions using Weibull distributions, a maintenance cost model is developed to determine optimal periodic inspection intervals that minimize the long-term expected cost per unit time. A parameter estimation framework is established for both continuous and discrete inspection data, ensuring robust model applicability. The proposed approach is validated using a real-world run-to-failure dataset, demonstrating its effectiveness in optimizing maintenance schedules. Key contributions include the application of DTM to rolling bearing lifetime modeling, the formulation of a cost-effective inspection and maintenance strategy, and empirical validation through a case study. This work provides a practical framework for enhancing rolling bearing reliability and reducing maintenance costs.
Stroke patients with hemiplegia require personalized upper-limb rehabilitation, yet designing safe and effective robot-assisted trajectories that mimic natural human movement remains a significant challenge. This paper proposes a trajectory planning and optimization method to address this need by leveraging multi-objective constrained reinforcement learning. The method involves dynamically capturing motion data from the patient's healthy limb to define personalized Activities of Daily Living (ADL). A reinforcement learning algorithm, guided by a specially designed reward-punishment function, then optimizes the trajectory with objectives for smoothness, jerk minimization, and accurate tracking of key points. The approach was validated on a 4-degree-of-freedom (4-DOF) upper limb rehabilitation robot, which successfully achieved multi-joint coordinated trajectory tracking based on the learned ADL movements. The experiments confirm the method's effectiveness in designing personalized rehabilitation trajectories that improve the continuity and smoothness of robot-assisted movements, offering a promising solution for patient-specific therapy.
Software-Defined Networking (SDN) reveals a significant progression in networking technology, offering improved management and operational oversight of network infrastructures. Even though the control plane offers benefits, it is still susceptible to Denial of Service (DoS) attacks, and this poses a significant threat to system security. By taking advantage of the network's centralized architecture, these attacks pose serious dangers and can overload controllers, leading to severe packet loss and significant downtime in the network. To address this challenge, we propose a novel approach that efficiently detects DoS attacks by implementing a packet inspection process using a queuing mechanism, followed by machine learning classification using SVM and KNN algorithms. These algorithms were rigorously evaluated using the CICDoS 2017 dataset and integrated into an SDN threat-detection framework. The results of extensive testing in SDN environment demonstrated higher efficiency measures, such as enhanced network performance by reducing latency and resource consumption, maintaining a false-positive rate under 5%, and achieving a detection accuracy of 99%. These results demonstrate how well our proposed approach works to successfully detect DoS attacks in SDN systems. Moreover, the novel approach, the thorough end-to-end solution exhibited, and the importance of the experimental outcomes all work together to establish a solid basis for future studies in this area.
With the promotion of bidding policies in the wind power industry, the demand for improving profits and reducing costs is becoming increasingly urgent. In the past, manual layout was mostly used for wind turbines and cable layout, and subjective factors had a significant impact, making it difficult to achieve true optimization of machine placement. This study introduces heuristic optimization, fully considers the environmental conditions and factors of wind farms, automatically searches for the optimal wind turbine layout through fuzzy genetic algorithm, fully utilizes wind farm resources, and improves economic benefits. The study also adopted a particle swarm optimization algorithm based on penalty functions to optimize cable layout. This study conducted experiments to verify the effectiveness of the proposed algorithm. The results show that the fitness of fuzzy genetic algorithm reaches 2.827 at around 35 iterations, while genetic algorithm and adaptive differential evolution algorithm reach 1.427 and 1.685 at around 38 and 41 iterations, indicating that fuzzy genetic algorithm can converge faster and has advantages. The optimized cable temperature is approximately 90°C, and the total load of the optimized cable is 6136.57A. Notably, the current carrying capacity of the cables has increased by 15.92%, demonstrating a significant improvement compared to pre-optimization values. These improvements highlight the effectiveness of the proposed fuzzy genetic algorithm in optimizing both wind turbine layouts and cable configurations. The results confirm that the optimization method enhances power generation efficiency and cable load distribution, providing a valuable reference for the design of wind turbine and cable layouts in wind farms.
Current humanistic design of urban public spaces focuses on specific design elements while ignoring the conflicts and couplings between multiple user needs. This leads to spatial strategies stuck in local optima and lacking overall balance and adaptability. This paper constructs a multi-objective optimization model that integrates user preferences, multidimensional spatial indicators, and behavioral simulation. This model collects field data such as heat maps, path trajectories, and dwell time, identifies user types through K-means clustering, and models their spatial preferences using fuzzy membership functions. Design variables are set in Grasshopper; an optimization function is constructed; the optimal solution is searched using NSGA-III. Finally, pedestrian simulation is performed in AnyLogic, and the optimization results are corrected for function deviation to improve the coordination and adaptability of the design. Experimental results show that this strategy framework significantly improves spatial coordination, increasing weighted average satisfaction from 0.61 to 0.81 (+32.8%), reducing safety risks by 30.8% to 63.2%, and increasing interaction promotion by 71.2%. Multi-dimensional indicators verify the effectiveness of the optimization strategy in balancing user needs, alleviating local conflicts, and enhancing spatial adaptability, providing a quantitative basis and practical path for systematically solving the local optimal problem of humanized design of public spaces.
Traditional rule-based manual bridge inspection methods often suffer from low efficiency and poor accuracy, making them inadequate for the demands of industrial-scale production. This study aims to achieve rapid recognition and localization of virtual assembly components within bridge 3D point clouds by constructing an intelligent analytical framework that integrates supervoxel clustering with a Transformer architecture. Specifically, an improved supervoxel clustering algorithm is developed, deeply integrating geometric morphology, density distribution, and structural response features to generate multimodal voxel units, thereby enhancing the semantic representation of local features. A graph-based Transformer module is introduced to model spatial relationships and semantic associations among supervoxel nodes through a self-attention mechanism, effectively integrating global contextual information. Additionally, a voxel voting strategy within a pose estimation module is employed to optimize component localization accuracy, forming an end-to-end recognition and localization system. The proposed model demonstrates excellent performance across multiple datasets, including Stanford Large-Scale 3D Indoor Spaces Dataset, ETH Zurich Building Dataset, International Society for Photogrammetry and Remote Sensing Benchmark Dataset, and National Building Museum Point Cloud Dataset. Compared to baseline models, the proposed approach achieves improvements of over 21.5% in semantic segmentation Mean Intersection over Union, instance recognition accuracy, and pose regression precision. In complex multi-box girder bridge scenarios, the recognition accuracy for small-scale connectors improves by up to 37.1%. Computational efficiency increases by more than 18.7%, with inference time reductions of up to 31.5% when processing large-scale data. Overall improvements in bridge component recognition exceed 22.4%, with recognition accuracy for critical connection components increasing by up to 37.4%, and localization accuracy improving by over 26.2%, reaching up to 35.9% for key node localization. The results demonstrate that the proposed model effectively addresses critical challenges in processing bridge point cloud data through multimodal feature fusion and global structural reasoning, significantly enhancing component recognition accuracy and localization precision in complex scenes while maintaining a balance between algorithmic efficiency and model performance. This study provides an efficient solution for the digital delivery and quality control of intelligent bridge construction. By integrating finite element analysis with deep learning, the model enhances semantic understanding of bridge structural functions, contributing significantly to the advancement of intelligent bridge engineering.
To address the comprehensive control challenges arising from the coupled effects of model uncertainties, parameter perturbations, and external disturbances in electro-hydraulic position servo systems, this study proposes an optimized fuzzy active disturbance rejection control strategy based on the Newton-Raphson-Based Optimizer (NRBO). A fuzzy-compensated active disturbance rejection controller (Fuzzy-ADRC) is developed. This controller introduces fuzzy logic to dynamically compensate for nonlinear disturbances in real time and enhances the system's robustness against external disturbances and uncertainties. To overcome the challenges associated with the numerous parameters and tuning difficulties of ADRC, the NRBO optimization algorithm is integrated to leverage its fast convergence and avoidance of local optima, enabling systematic parameter optimization for the Fuzzy-ADRC (NRBO-Fuzzy-ADRC). Simulation results demonstrate that compared to conventional ADRC control, the proposed NRBO-Fuzzy-ADRC reduces the step response time by 74.3% and decreases the average tracking error in sine responses by 70.1%. This algorithm significantly enhances control performance and provides a novel optimization framework for electro-hydraulic position servo system applications.
To solve the problem of user visual attention distribution and spatial misalignment of interface elements in current user interface layout design due to reliance on subjective experience, this paper proposes an optimization method for fashion brand e-commerce user interface design that uses eye tracking simulation and integrates brand characteristics. This paper builds a fashion-specific visual behavior database through multimodal eye movement data collection, and uses the spatiotemporal attention mechanism of the Transformer-XL (Transformer with Extra Long Context) model to predict the gaze hotspots and scanning paths of users in dynamic tasks; then, this paper designs a multi-agent reinforcement learning optimizer, encodes aesthetic rules such as brand logo size and main color ratio as hard constraints, and generates an interface space distribution plan through element competition-collaboration game; finally, this paper develops a real-time interactive prototype based on the Unity engine to achieve dynamic layout closed-loop optimization driven by the collaborative efforts of brand constraints and eye movement simulation heat maps. Experimental results show that this method reduces the task completion time by 32% in the optimization of the light luxury clothing homepage, the overlap between the simulated and real eye movement hot spots reaches 88%, and the brand consistency score increases by 46% (from 3.2 to 4.7), significantly reducing the user's cognitive load and improving the user experience. The conclusion confirms that by integrating eye movement behavior quantification with brand characteristics, it is possible to break through the traditional design's reliance on static aesthetics, provide a “cognitive adaptation-brand expression” dual-goal collaborative intelligent design paradigm for fashion interfaces, and promote the simultaneous improvement of user conversion rate and brand value.
Marine high-salt spray particles significantly accelerate vessel corrosion, with chloride ions being the primary corrosive component. This study presents a funnel-shaped intelligent sensor system integrating ion-selective electrode technology with enhanced neural network algorithms. The proposed design employs an improved Sparrow Search Algorithm-optimized Back Propagation Neural Network for temperature compensation, addressing the critical challenge of thermal drift in marine environments. Experimental results demonstrate the system's superior performance: maximum relative error of 1.786% across 0.001–1 mol/L chloride concentrations, with average error reduced to 0.972%–59.3% lower than conventional compensation methods. The sensor maintains near-theoretical sensitivity while achieving 0.02 mV/°C temperature coefficient through the proposed compensation mechanism. This advancement enables precise real-time monitoring of salt spray corrosion factors, providing a technical foundation for extending marine vessel service life through proactive maintenance strategies. A new funnel-shaped intelligent sensor is developed, which combines ion-selective electrode technology with an improved neural network algorithm. By introducing the sparrow search algorithm for optimization, the detection accuracy of chloride ions and the temperature compensation capability are enhanced, providing an efficient solution for the real-time monitoring of Marine high salt spray particles.
This study focuses on the site selection problem of fresh cold chain logistics warehouses, using a site selection model to minimize operating costs, improve distribution efficiency, and meet customer needs. The model covers warehouse location selection, special requirements for the fresh and cold chain, and organization of delivery routes. At the same time, a solution algorithm combining genetic algorithm and particle swarm optimization algorithm is proposed to solve the problem of location selection model. This hybrid algorithm encodes the layout problem of logistics points into a chromosome problem in genetic algorithms and utilizes particle swarm optimization to improve the efficiency of the search process and avoid early convergence difficulties. The results indicated that the improved algorithm was more effective than the ordinary genetic algorithm. After only 100 iterations, the objective function value of the algorithm decreased to approximately 31,500. The average total delivery cost of the model was reduced to 75.8369 million yuan, and the calculation was completed within 75 s, with significant efficiency. Therefore, this model can effectively assist logistics enterprises in accurately formulating economically effective distribution plans, achieving the optimal balance between cost and efficiency.
The increasing growth of urbanization has created significant prospects for the advancement of architectural landscape design. However, the existing machine learning and image processing methods provide partial solutions, because they struggle with noise, overlapping landscape features, and poor segmentation accuracy. To address these limitations, we propose a hybrid simulation and classification model that integrates the advantages of computer learning with immersive virtual reality (VR) environments. First, wavelet-based denoising and intensity normalization are applied to enhance 360° landscape image quality. A multi-orientation segmentation method is then used to accurately classify the complex visual features. Texture features are extracted using a combination of Grey Level Co-occurrence Matrix (GLCM) and Bayesian-optimized Gauss Markov Random Field (GMRF), which helps to capture both spatial and statistical relationships. These features are classified using a hybrid approach combining logistic regression (LR) and K-nearest neighbor (KNN), which allows strong observation of landscape features. Simulation of the model is conducted using real world immersive VR studies. The experiments demonstrate the superiority of the model in terms of accuracy (98%), precision (99.5%), and sensitivity (98.5%), respectively.
