Search Results (16,505)

To address the bottlenecks of low sample efficiency and poor control accuracy in traditional single-layer reinforcement learning during autonomous driving overtaking, this paper proposes an overtaking decision and control system based on hierarchical reinforcement learning to decouple complex tasks in spatial and temporal dimensions. A heterogeneous two-layer architecture is constructed, where the upper layer adopts the Proximal Policy Optimization algorithm to generate macroscopic discrete decisions, while the lower layer employs Twin Delayed Deep Deterministic Policy Gradient combined with Long Short-Term Memory to achieve smooth continuous control of steering and acceleration by perceiving temporal features of dynamic obstacles. A composite reward mechanism, integrating hard safety constraints and soft efficiency incentives, is designed to balance safety, efficiency, and comfort. Experimental results in complex scenarios with multiple interfering vehicles and random lane-changing behaviors demonstrate that the proposed system improves the training convergence speed by approximately 30% within 500,000 steps compared to single-layer algorithms. In tests across varying traffic densities, the system achieves a 98.3% success rate in medium-density scenarios with a collision rate of only 0.6%. In high-density challenges, the success rate remains above 95%, with the collision rate reduced by about 80% compared to baseline models. Furthermore, the lateral control deviation is strictly limited to within 0.2 m, and the longitudinal safety distance remains stable above 5 m. This system provides a robust, high-efficiency paradigm for autonomous overtaking. Full article

Addressing the lack of high-fidelity test cycles for fuel cell light-duty trucks, this paper proposes an automated driving cycle construction method that integrates unsupervised clustering and reinforcement learning. Firstly, based on large-sample real-world driving data, four libraries of typical driving pattern segments are extracted through dimensionality reduction via Principal Component Analysis (PCA) and K-means clustering. Subsequently, the cycle construction process is formulated as a sequential decision-making problem, and a framework based on the Proximal Policy Optimization (PPO) algorithm, incorporating an action masking mechanism, is designed. This framework innovatively injects macro-level time budget allocation as a hard constraint into the agent’s policy space via action masking, while utilizing micro-level Markov transition probabilities as a soft guide. This dual approach drives the agent to learn an optimal segment concatenation strategy, thereby simultaneously ensuring both the macro-level statistical representativeness and the micro-level driving logic coherence of the synthesized cycle. Validation results demonstrate that the cycle constructed by the proposed method achieves an average relative error of only 7.53% in key characteristic parameters, and its joint speed-acceleration distribution exhibits a similarity as high as 0.9886 with the original data, significantly outperforming traditional methods such as the clustering method, the Markov chain method, and standard driving cycles. This study provides an effective tool for generating high-fidelity driving cycles and testing energy management strategies for fuel cell commercial vehicles. Full article

Land cover classification (LCC) is a fundamental task in remote sensing, which enables effective environmental monitoring, agricultural planning, and disaster management. The existing approaches often rely on fine-tuning pre-trained models, which are not specifically designed for LCC, which lead to suboptimal performance in complex scenarios. To address these limitations, we propose the Cross-Module Attention Land Cover Network (CALCNet), a novel architecture developed from scratch. CALCNet follows a contracting and restoration backbone, where the contracting path extracts progressively abstract semantic features while reducing spatial resolution, and the restoration path recovers fine-grained spatial details through upsampling and skip connections. In addition, CALCNet integrates a cross-module attention mechanism that combines spatial attention and multi-scale feature selection to enhance feature representation. Furthermore, we applied a differential evolution-based neuron pruning strategy to create a compressed CALCNet variant, which retains high classification performance while reducing computational cost. The CALCNet is evaluated on four benchmark LCC datasets, AID, UCMerced_LandUse, NWPU_RESISC45, and EuroSAT, demonstrating strong performance across all benchmarks. Specifically, the model achieves classification accuracies of 98.09%, 99.47%, 99.19%, and 99.19%, respectively. The compressed CALCNet variant reduces computational cost to 78.55 million floating point operations (FLOPs) with a model size of 43 MB, while achieving improved inference speeds (38.32 frames/sec on CPU and 118.3 frames/sec on GPU), representing approximately 45–50% reduction in FLOPs and model storage. These results highlight that CALCNet is both highly accurate and computationally efficient, making it well suited for real-world LCC applications. Full article

An automotive welding unit is a modular production cell within a welding workshop that integrates industrial manipulators, welding equipment, fixtures, and control systems to perform specific welding and assembly tasks. A large number of industrial manipulators are utilized in the automotive welding unit. The capability to quickly plan a short and collision-free path in the workspace of the manipulator is of great importance for improving the manipulator’s intelligence level and production efficiency. The RRT* algorithm, based on random sampling, has been widely applied in path planning for high-dimensional manipulators due to its probabilistic completeness and powerful exploration capabilities. However, the RRT* algorithm performs poorly in spaces containing narrow passages. Research on the practical application of path planning for 6-DOF manipulators is still insufficient, particularly in planning posture. To solve these two problems, an improved RRT* algorithm is proposed in this paper. New sampling and node connection strategies are designed to improve the expansion and convergence speed of the random tree in spaces containing narrow passages. A distance-constrained posture quaternion interpolation method is presented to generate smooth and continuous paths for manipulators of the automotive welding unit. Simulations and experiments are carried out to validate the proposed method, which confirms that the method can plan collision-free paths for manipulators more quickly compared to other methods. Full article

This paper proposes an advanced wind energy conversion and management framework for improving grid integration and mitigating frequency and power fluctuations caused by wind intermittency. The studied system combines a permanent magnet synchronous generator (PMSG), a unidirectional Vienna rectifier on the machine side, a Li-ion battery energy storage system, and a bidirectional Vienna rectifier on the grid side. The main scientific challenge addressed in this work is to ensure efficient wind power extraction, secure battery charging/discharging operation, and stable power exchange with the grid under variable operating conditions. To this end, a comprehensive nonlinear state-space model of the overall system is first established. Then, nonlinear controllers based on integral sliding mode principles are developed to guarantee rotor-speed tracking, DC-bus voltage regulation, battery charging current limitation, and active/reactive power control. In addition, an adaptive observer is designed to estimate the battery open-circuit voltage and support the supervision of the state of charge. An energy management strategy is further proposed to coordinate the operating modes according to grid conditions and battery constraints. Simulation results demonstrate that the proposed approach effectively smooths wind power fluctuations, improves grid support capability, and enhances the overall dynamic performance of the wind energy conversion system. Full article

In this study, Delrin^® (POM) polymer was reinforced with 15 wt.% chopped ramie fiber (RF) to develop a sustainable composite, which was injection-molded and machined using abrasive waterjet machining (AWJM). SEM revealed a skin-core morphology with flow-induced RF alignment and small voids at bundle crossovers, indicating interfacial adhesion. A Taguchi L₉ (3³) design evaluated waterjet pressure (WJP: 100–300 MPa), traverse speed (TS: 100–200 mm/min), and stand-off distance (SoD: 1–3 mm) on kerf width (KW) and surface roughness (SR). Increasing WJP from 100 to 300 MPa lowered mean SR from 6.23 to 4.80 µm (23% reduction) and KW from 1.31 to 1.07 mm (reduction of 18%); enlarging SoD from 1 to 3 mm raised SR from 4.98 to 5.55 µm (an 11% increase) and KW from 1.12 to 1.20 mm (a of 7% increase); and raising TS from 100 to 200 mm/min narrowed KW from 1.24 to 1.11 mm (a 10.5% reduction) with a modest SR decrease from 5.45 to 5.28 µm. ANOVA confirmed WJP as the dominant factor for SR (79.8%), as well as a significant SoD (18.3%). For KW, the influence of WJP (68.8%) was substantial, followed by TS (19.9%) and SoD (11%). Linear models captured the trends well (SR: R² = 88.29%; KW: R² = 93.36%). A desirability-based multi-response optimizer yielded ideal conditions for TS (200 mm/min), WJP (300 MPa), and SoD (1 mm), predicting a KW of 0.94 mm and an SR of 4.1567 µm. Confirmation tests produced a KW (0.970 ± 0.01 mm) and SR (4.27 ± 0.05 µm), which are within 3.19% and 2.73% of the predicted values, validating the DoE regression approach. Full article

This study investigates the failure mechanisms and machining performance of 3D-printed H13 tool steel end mills driven by the creation of a Finite Element Analysis (FEA)-based digital twin. The primary objective is to assess the process capability of these tools by integrating CAD and FEA with product design simulation-based data acquisition within a digital manufacturing framework, thereby validating a physical model. This research begins by redesigning a 20 mm end mill into a 6 mm, four-flute configuration, and then FEA simulating H13 tool steel and tungsten carbide (WC) tools. This is carried out to machine Al-6082-T6 under spindle speeds of 5500 rpm and 1500 rpm, with a constant feed rate of 0.5 mm/tooth over 100,000 cycles. The process is integrated with the Siemens Insights hub via Node-RED to replicate the simulation to correlate the CPU signal spikes and enhanced processing capacity, especially in relation to CAD/CAE kernel activities. Based on the simulation insights, two H13 end mills are fabricated using Fused Filament Fabrication (FFF). The first tool, tested at 5500 rpm and a 1100 mm/min feed rate, fractured after 70 mm of cutting. The second trial, using a diamond-coated solid carbide tool at 1500 rpm and 300 mm/min, achieved successful machining with graphene-enhanced coolant. The cutting forces ranged from 300 to 500 N for 3D-printed tools, compared with 150–180 N for the carbide tool. The surface roughness varied between 0.6–1 µm and 4–6 µm for the printed tools, aligning closely with traditional tools (0.5–1 µm). Post-machining analysis using SEM and EDX confirmed tool wear and material changes. This work adopted a methodology to capture and monitor CPU signal spikes via the digital twin platform, enabling real-time comparison with failure thresholds. The results demonstrate the feasibility of using 3D-printed H13 tools for sustainable, customizable machining, offering a pathway for industries to adopt in-house tool design and manufacturing with integrated digital validation. Full article

Background/Objectives: A robust and scalable manufacturing framework for lipid-based nanocarriers remains a critical challenge, particularly for labile phytochemicals such as curcuminoids in turmeric. This study presents an integrated Quality by Design (QbD)-driven and Outcome-Based Design (ObD) strategy to establish a scalable, resource-efficient manufacturing process for curcuminoids-loaded nanostructured lipid carriers (NLCs). Methods: To overcome the limitations of conventional multivariate design of experiments (DOE), which require extensive experimental runs, a risk-based, knowledge-driven single-factor screening approach was employed. Guided by risk assessment tools, including Ishikawa diagrams and failure mode considerations, 12 representative processing conditions were selected to define the design space. Critical quality attributes (CQAs), namely, particle size, polydispersity index (PDI), and zeta potential, were predefined to establish a robust control strategy. A two-step homogenization process—high-shear homogenization (HSH) for pre-emulsification followed by high-pressure homogenization (HPH) for nanoscale refinement—was systematically optimized. Results: Multivariate data analysis using principal component analysis (PCA) and hierarchical cluster analysis (HCA) identified key critical process parameters (CPPs), particularly HSH speed, processing time, and HPH cycles, as dominant factors influencing nanoparticle characteristics. The optimized 1-h process enabled successful scale-up of NLCs from 100 g to 5000 g, demonstrating the capability to generate nanosized particles within 100–500 nm. The combined HSH–HPH approach produced smaller, more uniform nanoparticles with high encapsulation efficiency and physical stability, outperforming HSH alone. Conclusions: Overall, this study establishes a practical and industrially viable framework that integrates QbD principles with data-driven optimization tools, for enabling reliable translation from laboratories to semi-industrial production. Full article

The overall swing dynamics of rigid slender payloads lifted in a vertical orientation deviate significantly from the ideal point-mass pendulum model, leading to severe performance degradation of feedforward control strategies designed based on this simplified model. This paper focuses on the bridge crane system and establishes a double-pendulum dynamic model that accounts for the payload’s mass distribution effect. To compensate for the theoretical error of the linearized model, a data-driven payload swing frequency correction strategy is proposed. Based on this corrected model, a dual-mode Zero Vibration Derivative (Corrected-Dual-ZVD) input shaping feedforward controller is designed. Simulations under eight typical operating conditions were conducted using the Matlab/Simulink control system simulation software. The results show that compared to the controller designed based on the traditional single-pendulum model, the proposed Corrected-Dual-ZVD controller, based on the corrected double-pendulum model, can significantly reduce the maximum residual swing angle of the payload. The average swing angle suppression rate reaches 68.9% across seven valid operating conditions, and it can reach 98.9% under the extreme condition of high speed and short rope length. When model parameters are subjected to ±10% disturbances, the proposed method demonstrates good robustness. Full article

Complex engineering optimization problems are typically characterized by high dimensionality, multimodality, and strong constraints, posing significant challenges to traditional swarm intelligence algorithms in terms of convergence speed, solution accuracy, and robustness. The Catch Fish Optimization Algorithm (CFOA), a recently proposed swarm-based metaheuristic, exhibits promising global search capability; however, it still suffers from deficiencies in search direction stability, elite solution utilization, and exploitation performance in the later stages of optimization. To address these limitations, this paper proposes an Improved Catch Fish Optimization Algorithm, named Elite-Driven Reinforced Catch Fish Optimization Algorithm (EDR-CFOA). On the basis of the original CFOA framework, EDR-CFOA integrates three complementary elite-based enhancement strategies: an elite-enhanced search strategy, an elite differential evolution strategy, and an elite random local search strategy. Through a multi-level elite-guided mechanism, these strategies collaboratively improve the reliability of search directions, strengthen solution-space recombination, and enhance fine-grained exploitation of high-quality solutions, thereby significantly improving the overall optimization performance of the algorithm. The proposed EDR-CFOA is systematically evaluated on the CEC2020 and CEC2022 benchmark test suites under 10-dimensional and 20-dimensional settings and is compared with eight classical and recently developed high-performance metaheuristic algorithms. The Friedman mean ranking results demonstrate that EDR-CFOA achieves the lowest average rank in all four test scenarios (CEC2020: 1.30 for 10D and 2.20 for 20D; CEC2022: 1.17 for 10D and 1.08 for 20D), consistently ranking first overall and significantly outperforming the competing algorithms. Furthermore, Wilcoxon rank-sum tests confirm that EDR-CFOA exhibits statistically significant superiority on the majority of benchmark functions. In addition, EDR-CFOA is applied to the economic optimal scheduling problem of a grid-connected microgrid and several typical constrained engineering design problems, where experimental results verify its feasibility, robustness, and practical engineering applicability. Comprehensive numerical experiments and real-world engineering case studies indicate that EDR-CFOA is a highly effective swarm intelligence algorithm featuring high solution accuracy, strong stability, and excellent generalization capability, making it well suited for complex engineering optimization problems. Full article

Minimizing the impacts of coupling, randomness, time variation and uncertain nonlinearity to enhance real-time performance is critical for controlling complex industrial systems. This paper proposes an optimized adaptive filtering method (LAF-MTNF) for time-varying uncertain nonlinear systems with multiple-input multiple-output (MIMO) measurement noise, which integrates the multi-dimensional Taylor network (MTN) with Lyapunov stability theory (LST). Leveraging MTN’s inherent advantages—simple structure, linear parameterization, and low computational complexity—LAF-MTNF achieves efficient real-time filtering while avoiding the exponential computation burden of neural networks. The contributions of this work are threefold: (1) A novel integration of LST and MTN is proposed for MIMO filtering, in which an energy space is constructed with a unique global minimum to eliminate local optimization traps, addressing the stability deficit of traditional MTN filters using LMS/RLS algorithms. (2) Convergence performance is systematically quantified by deriving explicit expressions for the error convergence rate (regulated by a positive constant) and convergence region (a sphere centered at the origin) while modifying adaptive gain to avoid singularity, filling the gap of incomplete performance analysis in existing Lyapunov-based filters. (3) The design is disturbance-independent, relying only on input/output measurements and requiring no prior knowledge of noise statistics, thus enhancing robustness to unknown industrial disturbances. We systematically analyze the Lyapunov stability of LAF-MTNF, and simulations on a complex MIMO system verify that it outperforms existing methods in filtering precision (mean error 0.0227 vs. 0.0674 of RBFNN) and dynamic response speed, while ensuring asymptotic stability and real-time applicability. The proposed LAF-MTNF method achieves significant advantages over traditional adaptive filtering methods in filtering accuracy, convergence speed and anti-cross-coupling capability. This method has broad application prospects in high-precision industrial servo motion control, power system state monitoring and other multi-variable nonlinear industrial scenarios with complex noise environments. Full article

Convolutional neural networks (CNNs) have achieved remarkable performance in industrial image-based surface defect inspection in recent years. However, many state-of-the-art (SOTA) networks have become increasingly complex and computationally expensive, which limits their deployment in resource-constrained environments or high-throughput, real-time, practical industrial applications. To address this challenge, this paper proposes a novel approach, Light SDI-NAS, to automatically design lightweight CNN architectures for real-time industrial surface defect inspection through neural architecture search (NAS). First, a task-oriented search space for industrial image inspection is constructed by integrating prior knowledge of neural network architecture design with empirical observations. Second, a novel loss function is introduced to balance model accuracy and computational efficiency during the architecture search process. Finally, the lightweight networks generated by Light SDI-NAS demonstrate strong performance on three industrial image datasets. Experimental results show that the proposed models achieve comparable or superior accuracy to manually designed SOTA networks while significantly reducing the number of parameters and improving inference speed by 1.8 times, making them highly suitable for real-time industrial inspection applications. Full article

With the advent of 6G mobile communication, the demand for ultra-high bandwidth wireless communication has increased rapidly, drawing significant attention to visible light communication (VLC) as a promising emerging technology. GaN-based laser diodes (LDs) are regarded as high-speed light sources for VLC owing to their high modulation bandwidth and high optical power density. Apart from the active region design, the LD’s structure also plays a crucial role in determining their dynamic properties, which have yet to be thoroughly studied in III-nitride LDs. In this work, we systematically investigate InGaN/GaN laser diodes with three ridge waveguide configurations: a conventional single-ridge structure, a dual-ridge large-mesa structure, and a dual-ridge small-mesa structure. The threshold current, small-signal modulation bandwidth of devices with different structures are comparatively analyzed. Experimental results reveal that the double-ridge small mesa laser diode achieves a modulation bandwidth of −3 dB at 6.02 GHz. These results provide valuable insights into the structural optimization of GaN-based high-speed laser diodes and offer practical guidance for the development of high-performance, energy-efficient VLC transmitters. Full article

This paper presents a non-contact depth detection method for corrugated heat exchanger plates, aiming to improve measurement efficiency and accuracy. The system integrates a line laser sensor with a precision linear guide rail, enabling continuous acquisition of high-resolution 2D surface profiles as the sensor moves along the plate. To reduce data redundancy while preserving geometric features, a multi-stage data reduction strategy is proposed. This strategy combines the angle–chord height criterion with spline-based filtering to identify key regions of curvature and eliminate unnecessary point cloud data. For depth extraction, a two-stage feature recognition algorithm is designed. First, a coarse analysis locates candidate peaks and valleys by identifying local extrema in the reduced 2D data. Then, a fine detection process is applied: local B-spline fitting is performed near each candidate point, and a binary search algorithm is used to accurately determine the spline extrema. By computing the vertical distance between precisely located peaks and valleys, the system rapidly extracts the corrugation depth parameters. This method achieves a high balance between speed and precision, offering a practical and reliable solution for automated surface morphology inspection in heat exchanger manufacturing. Full article

With the growing simultaneous demands for miniaturization and high performance, thermal issues such as hotspots severely degrade the high-speed signal transmission performance of low temperature co-fired ceramic (LTCC) substrate in system-in-package (SiP) modules. This paper proposes a high-speed transmission line design for LTCC substrates, using a G-S (Ground-Signal) structure to ensure reliable signal transmission quality. Based on this structure, finite element simulations are performed to investigate the electromagnetic signal transmission characteristics under both uniform and non-uniform thermal fields, confirming that signal transmission efficiency exhibits strong temperature dependence. The results indicate that when the temperature exceeds 50 °C, non-uniform temperature distributions exert a significantly stronger influence on electromagnetic performance, leading to aggravated signal reflections and reduced transmission efficiency. At 300 °C, the transmission efficiency under non-uniform temperature drops to 35.0%, which is a 61.8% decrease compared with the optimal scheme obtained under ideal electric field conditions. Under electromagnetic-thermal coupling, a comparative study of different schemes shows that the optimal design derived from a single electric field is not suitable for electromagnetic-thermal coupled working conditions. The optimized Scheme 2 increases transmission efficiency to about 75.3%, with smoother S-parameter curves and smaller fluctuations. These findings provide valuable references for subsequent reliability-oriented design and experimental verification. Full article