Search Results (238)

Accurate thermal management is crucial for ensuring the safety, longevity, and performance of lithium-ion batteries, especially in compact embedded systems like USB chargers, power banks, and IoT nodes. Despite extensive research on predictive thermal models and intelligent control frameworks, their implementation in resource-constrained microcontroller-class devices has been limited. Existing strategies in the literature, such as threshold-based or PID logic, cloud-enabled analytics, machine learning models, and observer-based estimators, are often reactive, computationally intensive, or dependent on external infrastructure, making them unsuitable for low-power, standalone applications. This study introduces a novel Scalable Embedded Thermal Intelligence architecture designed for real-time battery thermal regulation in locally executable, without cloud dependency, low-cost platforms. Unlike conventional methods, the proposed system operates entirely on-device using closed-form models implemented on an ESP32 microcontroller. It combines two synergistic algorithms: a static preemptive model that calculates a safe C-rate at startup based solely on ambient and initial battery temperature, and a dynamic disturbance-aware model that monitors temperature rise per SOC step and adjusts airflow or current adaptively without requiring high memory, floating-point units, or supervisory control. The architecture achieves sub-second response times, <7% RAM, and <25% Flash usage, and does not need cloud connectivity, simulation backend, or complex thermal-management infrastructures such as liquid cooling circuits, phase-change systems, or cloud-supervised architectures. The significant contribution of this work is not the introduction of a new electrochemical–thermal formulation, but the effective integration and application of previously validated closed-form thermal predictors on low-cost microcontroller-class hardware, designed for anticipatory battery thermal regulation while adhering to strict computational limitations. Compared to traditional battery thermal management systems using PCM, liquid-cooling circuits, or cloud-based predictive estimators, the proposed approach eliminates the need for complex thermal hardware, fluidic systems, external computing infrastructure and resource-efficient edge operation. This makes the system suitable for deployment in real-world embedded applications like USB-C smart charging cables, compact IoT power banks, and portable medical devices, where form factors, energy efficiency, and cost are critical. The proposed SETI framework offers a firmware-integrated architecture and a firmware-integrated solution that provides a lightweight embedded alternative for predictive thermal regulation for distributed energy systems and miniaturized electronics. Full article

Examining how artificial intelligence industry agglomeration (AIIA) affects carbon and pollution reduction is crucial for China’s agricultural sustainability. Existing research mainly examines the effect of artificial intelligence (AI) on the reduction of single pollutants while overlooking how industry agglomeration influences the marginal cost of coordinated abatement, a key issue for the agricultural resource–environment–economy system. Using panel data for 30 Chinese provinces from 2016 to 2024, this study constructs a marginal cost-based indicator of agricultural pollution–carbon reduction synergy (APCRS) and examines the effect of AIIA. The full-sample results reveal that AIIA has a U-shaped relationship with APCRS. Technological progress partially mediates this relationship. Agricultural socialized services and rural industrial integration buffer the initial negative association, whereas agricultural labor productivity strengthens the curvature of the estimated nonlinear pattern. The effect of AIIA also varies with external conditions and is more pronounced in regions with higher levels of marketization and industrialization while remaining significantly U-shaped across grain strategic zones. This dynamic process is more likely to emerge when public innovation investment and rural household income exceed critical thresholds. These findings provide new evidence for understanding how AI-driven agglomeration can support green agricultural transformation. Full article

High-dynamic amusement ride conditions involving impacts, rapid rotations, and abrupt posture changes introduce severe motion artifacts that degrade vital sign quality and destabilize physiological state recognition. This study aims to develop an engineering-ready closed-loop framework for robust passenger monitoring and intelligent diagnosis. A multimodal sensing and modeling pipeline was designed to jointly leverage physiological signals such as heart rate and SpO₂ and kinematic measurements, including acceleration, angular rate, velocity, and attitude. Inertial and PPG signals were preprocessed into supervised samples through wavelet multiresolution denoising and coordinate frame unification, while a strapdown inertial navigation system was used to propagate a 12-channel physical quantity sequence. To ensure interpretability and standards compliance, constraints from GB 8408-2018 were translated into executable threshold rules, enabling standards-driven auto-labeling and rule-based early warning. Building on this foundation, three learning modules were developed: a fusion model for high-dynamic heart rate estimation, a CNN–LSTM dynamic-threshold-enhanced network TAPNet for rapid kinematic anomaly screening, and an attention-augmented hybrid model HS-BANet integrating one-dimensional residual blocks, bidirectional LSTM, and multi-head attention for fine-grained arrhythmia classification. Experimental results demonstrated accurate and consistent heart rate estimation with RMSE of 1.18 bpm on HSSH-I and 1.24 bpm on the independent HSSH-II set, strong agreement with training and testing correlations of 0.9928 and 0.9865, and near-zero bias in Bland–Altman analysis. TAPNet achieved 96.9% validation accuracy and 98.2% test accuracy for kinematic anomaly recognition, maintaining robust generalization under class imbalance. HS-BANet enabled multi-class identification of PVC, PAC, VT, SVT, and AF, achieving an accuracy of 92.37%, an F1-score of 86.87%, a precision of 88.45%, a sensitivity of 88.14%, and a specificity of 89.42%. Overall, the proposed two-stage multimodal closed-loop—fast, interpretable early warning based on physical quantity thresholds followed by fine-grained diagnosis from physiological signals—supports stable feature extraction and reliable decision-making under strong motion artifacts and non-stationary dynamics, balancing responsiveness and diagnostic credibility, while showing potential for practical safety early warning and future deployment-oriented operational support in amusement ride scenarios. Full article

In the context of an increasingly complex and dynamic digital ecosystem, accurately assessing the value of Specialized, Refined, Differentiated, and Innovative (SRDI) enterprises is crucial for making effective decisions. Traditional valuation methods struggle to effectively address issues such as the high R&D expenditures and significant operational risks associated with these enterprises. This study proposes an interpretable intelligent decision-support framework for valuing SRDI enterprises listed on the Beijing Stock Exchange (BSE), constructing a multidimensional indicator system that encompasses solvency, profitability, and R&D capabilities. Feature importance screening using the XGBoost algorithm was conducted to identify key indicators as input variables for a backpropagation (BP) neural network. Concurrently, the Particle Swarm Optimization (PSO) algorithm was applied to the neural network to optimize initial weights and thresholds, thereby modeling nonlinear valuation relationships. Empirical analysis of 770 SRDI firms listed on the Beijing Stock Exchange from 2020 to 2024 indicates that the XGBoost-PSO-BPNN model achieved a coefficient of determination of 0.8083 on the test set, outperforming traditional linear models and benchmark models such as single-tree models. SHAP explainability analysis further reveals that current asset turnover, return on assets, and equity concentration are the primary value drivers. This study employs various clustering methods to further classify enterprises into three categories and proposes recommendations for differentiated regulatory policies, providing intelligent decision support for enterprises operating within complex digital ecosystems. Full article

The growing complexity of software-defined automotive systems generates massive heterogeneous sensor and ECU data during real and virtual validation, and conventional rule-based analysis of such multivariate time series struggles under dynamic operating conditions, noise, and diverse fault scenarios. Deep learning-based anomaly detection has shown promising performance, yet existing approaches remain limited by static thresholds, insufficient robustness, and reduced interpretability. This study proposes an adaptive framework for intelligent fault detection in test recordings of automotive software systems (ASSs), integrating deep denoising autoencoders (DAEs), adaptive Gaussian thresholding, and explainable artificial intelligence (XAI) techniques. Four DAE architectures (ANN-, RNN-, GRU-, and LSTM-DAE) are systematically evaluated under different noise levels, system versions, and fault conditions, with detection thresholds that adapt dynamically to the statistical behavior of the reconstructed signals, thereby reducing false alarms under varying operating conditions. The framework was evaluated using real-world test recordings from IAV and Hardware-in-the-Loop (HIL)-based digital test drives, where ANN-DAE achieved the most robust detection performance, with F1-scores of 93.91% and 96.39% on the real and virtual test-drive data, respectively. Furthermore, the integration of XAI improved the transparency of anomaly interpretation at the signal level. Overall, the proposed framework shows strong potential for intelligent anomaly detection and quality assurance in safety-critical automotive systems. Full article

Frequent seasonal phase transitions in cold and arid lakes require different remote sensing indices for frozen and open-water periods, complicating the use of traditional empirical indices for automated monitoring. To address this challenge, this study proposes an intelligent indexing framework integrating the heuristic reasoning of Large Language Models (LLMs) with Random Forest (RF) feature selection. Leveraging the Google Earth Engine (GEE) and Landsat 8 data from Ulansuhai Lake, five LLMs such as Gemini and ERNIE were employed to generate candidate spectral indices based on typical sample spectra. Optimal band combinations were identified via RF importance, and Land Surface Temperature (LST) was incorporated as a physical constraint for unified cross-seasonal classification and determine the optimal threshold. Results show that the LLM-derived ERNIE-WISI and Gemini-WISI exhibit high robustness. During the freezing period, ERNIE-WISI significantly outperformed other indices, achieving an Overall Accuracy (OA) of 89% and a Kappa of 0.86. Spatially, it yielded snow and ice mapping with clear textures and low commission errors. During the non-freezing period, ERNIE-WISI achieved an OA of 95% with a Kappa of 0.84. While Gemini-WISI achieved an OA of 94% with a Kappa of 0.80, performing comparably to MNDWI. Notably, ERNIE-WISI effectively suppressed background interference in complex landscapes like narrow channels and aquaculture areas, maintaining high geometric fidelity and spatial continuity. A key advantage of ERNIE-WISI is its consistent performance without seasonal threshold adjustments. Aligned with the AI for Science paradigm, this methodology bridges AI-driven heuristic discovery and physical remote sensing, offering a robust, transferable solution for long-term dynamic lake monitoring in extreme environments, thereby facilitating sustainable water management. Full article

With the continuous growth of urban building density and elevator deployment, the reliability, maintenance, and degradation risk warning of elevator systems have attracted increasing attention. Conventional monitoring methods based on fixed thresholds or rule logic are easy to implement, but they often fail to identify progressive degradation and are sensitive to complex operating conditions and measurement noise. This paper proposes a physics-informed neural network (PINN)-based method for elevator health monitoring and early warning. First, multi-sensor data are processed through time alignment and feature reconstruction, and a dual-path acceleration estimation method is introduced to improve the stability of dynamic state calculation. Second, a simplified traction elevator dynamic model considering load variation, motor drive, and mechanical resistance is embedded into PINN training to identify hidden parameters. Electrical and dynamic residual indicators are then constructed to characterise system condition from different physical perspectives. Finally, a time-accumulated risk model combining anomaly magnitude and persistence duration is developed to detect progressive degradation trends. Results show stable parameter convergence and effective condition assessment. The proposed approach detects degradation trends earlier than conventional threshold-based monitoring methods and reduces false alarms caused by transient disturbances. It provides an interpretable and practical solution for predictive maintenance and intelligent operation of elevator systems. Full article

The You Only Look Once (YOLO) series has consistently advanced the field of object detection, evolving from YOLOv1 to the latest YOLO26, achieving remarkable improvements in detection accuracy and computational efficiency. However, deploying such high-performance models on resource-constrained edge devices remains challenging, particularly for tasks requiring real-time inference. A critical yet often overlooked factor affecting edge deployment is the trade-off between input image resolution and computational cost: while higher resolution preserves fine-grained details essential for detecting small defects, it proportionally increases energy consumption and latency. To address this issue, we propose a novel multi-resolution adaptive detection framework based on YOLO26, specifically optimized for Ascend NPU edge devices. Our method dynamically selects the most suitable input resolution for each inference instance via a jointly optimized scene complexity metric, where the feature weights and resolution thresholds are simultaneously calibrated through Bayesian multi-objective optimization to achieve an optimal balance between predictive accuracy and energy efficiency. The experiments on transmission line insulator defect detection demonstrate that our approach achieves favorable trade-offs, maintaining high detection precision while significantly reducing power consumption compared to fixed-resolution baselines. The proposed framework provides a viable solution for intelligent visual inspection in power grid infrastructure. Full article

To improve the efficiency, accuracy, and operational stability of postharvest potato tuber sorting in the presence of mineral impurities, mainly soil clods and stones, an intelligent sorting system for potato tubers and mineral impurities was designed and developed. The system employed YOLOv10n as the baseline network and incorporated a PSA module together with a dynamic blur augmentation strategy to establish a task-adapted detection model, termed YOLOv10n-PB. Rather than treating detection accuracy alone as the optimization objective, the proposed system jointly considered detection performance, inference-latency stability, temporal–spatial coordination, and pneumatic rejection reliability. In addition, a programmable logic controller and pneumatic actuators were integrated to enable online target identification and dynamic removal. Comparative experiments involving lightweight YOLO models and $L_{25}\left(5^3\right)$ orthogonal tests were conducted to evaluate the effects of conveyor belt speed, material spacing, and classification threshold on sorting performance. The results showed that YOLOv10n-PB achieved a mAP@0.5 of 98.9% on the test set. Among the investigated factors, conveyor belt speed had the greatest effect on overall sorting accuracy, followed by material spacing and classification threshold. Range analysis identified the optimal parameter combination as a conveyor belt speed of 0.2 m/s, a material spacing of 9 cm, and a classification threshold of 0.4. Validation experiments under these conditions yielded an overall sorting accuracy of 98.3%, a combined mineral-impurity removal accuracy of 98.3%, and a potato tuber false rejection rate of 1.7%. These results demonstrate the feasibility of the proposed system for accurate and stable automatic sorting of potato tubers and mineral impurities under postharvest operating conditions. Full article

In megacities, where conventional mitigation strategies exhibit variable and environment-dependent performance, urban air pollution continues to be a significant public health concern. To methodically assess the operational reliability of urban smog mitigation systems under dynamic atmospheric conditions, this study proposes a data-driven failure analysis approach. A machine learning architecture based on Random Forest and XGBoost algorithms is developed using integrated meteorological and air quality metrics from Lahore, Pakistan, such as temperature, wind speed, and relative humidity. AQI is used as an integrated pollution indicator alongside meteorological variables to enhance the model’s ability to capture overall atmospheric pollution impact and improve the accuracy of smog mitigation failure prediction. This study presents a data-driven framework for predicting the failure of smog mitigation methods based on meteorological conditions. Unlike existing approaches that primarily focus only on air quality prediction, this work identifies specific environmental conditions, along with AQI as an input feature, to determine when mitigation strategies become ineffective. This enables proactive decision-making to maintain healthy indoor air quality. A threshold-controlled indoor air purification system that self-activates when the model predicts mitigation failure using real-time sensor inputs is introduced to address outdoor mitigation restrictions. PM_2.5 reduction efficiency, clean air delivery rate, and energy consumption indicators are used to evaluate the purifier’s optimized performance. Predicting mitigation failure rather than just pollution levels and connecting it with an intelligent interior reaction mechanism is what makes this research novel. In a comparative analysis, Random Forest outperforms XGBoost with an accuracy of 95.5% as opposed to 94.5%, as well as higher precision (96.9%), recall (96.1%), and F1-score (96.5%). The purifier lowered indoor AQI from dangerous to safe levels within 30–40 min. Full article

Safe and intelligent collision avoidance technology is essential for the autonomous underwater vehicle (AUV) to navigate in underwater environments. Most existing spike methods are constrained by a fixed static threshold and are unable to dynamically adjust to threshold changes reasonably, leading to difficulties in robustly adapting to external dynamic interference and thus resulting in insufficient homeostasis and generalization. To address these limitations, inspired by the dynamic threshold changes in biological neural systems, a bioinspired dynamic adaptive threshold (BDAT) is proposed. Combining the spiking neural network with deep reinforcement learning, a novel bioinspired dynamic adaptive threshold planner (BDAT-Planner) framework is constructed for underwater dynamic collision avoidance tasks performed by AUVs in complex, unknown environments. The proposed BDAT-Planner consists of the spiking dynamic adaptive actor network (SDAAN) and the deep critic normal network (DCNN). The BDAT is deployed to each spiking neuron in the SDAAN, dynamically adjusting the spike firing rate through threshold changes and avoiding excessive excitation or inhibition, thus maintaining homeostasis. The spiking encoder and spiking decoder are designed to convert continuous information and spiking sequences. Experimental results from both the training process and evaluation process (ablation studies, comparison experiments, and homeostasis experiments) demonstrate that the proposed BDAT-Planner has achieved superior performance in dynamic collision avoidance and model homeostasis compared to static threshold methods and existing comparison methods. The novel idea of bioinspired dynamic adaptive threshold can maintain model homeostasis and effectively enhance its adaptability to external dynamic interference, which offers significant development potential for promoting the efficient and stable operation of AUVs in marine environments. Full article

To address the issue of 2D laser-guided automated guided vehicles (AGVs) in industrial intelligent material handling scenarios being susceptible to interference from changes in lighting and complex obstacles, leading to abnormal positioning and mapping and frequent false stops, this paper designs a lightweight, multi-dimensional perception and anti-false-stop YOLOv8 anomaly recognition network, achieving accurate identification of various interferences in complex environments. An adaptive decision-making fault-tolerant control algorithm is proposed, introducing a temporal logic verification and dynamic threshold adjustment mechanism to achieve real-time dynamic switching of obstacle avoidance levels, ensuring efficient coordination between perception decision-making and control execution. An AGV anomaly detection sample set suitable for complex industrial scenarios is constructed, providing reliable data support for model optimization and accuracy evaluation. Finally, real-world deployment verification in a real electronics factory environment shows that this method reduces the vehicle false-stop rate and improves task handling efficiency. This research effectively solves the robust perception problem of AGVs in complex industrial environments and has significant engineering application value. Full article

To address the challenges of generating autonomous, high-quality control laws for high-performance Unmanned Aerial Vehicles (UAVs) performing long-horizon complex aerobatic maneuvers, specifically the difficulty of achieving energy-altitude closure, low-level control chattering, and low utilization of high-quality experience, this paper proposes an improved Soft Actor-Critic (SAC) algorithm incorporating a Quality-Aware Expert Pool (QAEP). Using the aerobatic loop maneuver as a representative scenario, this study explores the autonomous generation of control-surface manipulation policies comparable to those of skilled human pilots for agile fixed-wing UAVs. First, a singularity-free feature state representation and a rate-integrated action space are constructed. Combined with a symmetric shaping reward, these suppress control chattering at the physical level and achieve energy-altitude closure throughout the maneuver. Second, a dual-threshold expert pool driven by task reward and trajectory quality, together with a progressive mixed-sampling mechanism, is designed to effectively filter out low-quality samples and improve algorithmic convergence stability. Simulation experiments based on JSBSim with a high-fidelity F-16 model, which serves as a representative surrogate for a high-performance UAV, demonstrate that the proposed method generates maneuver strategies with manipulation quality comparable to that of skilled human pilots. The Dynamic Time Warping (DTW) similarity between the generated control commands and human expert demonstration data exceeds 0.97, the Manipulation Smoothness Index (MSI) is improved by 7.3%, and the loop completion rate under randomized initial conditions reaches 96.2%. These results suggest that the proposed framework enables human-like energy coordination and fine-grained control sequence generation in complex simulation environments, offering a promising approach to advancing maneuver intelligence and autonomous control capability in UAV systems. Full article

This research proposes a conceptual framework that employs generative artificial intelligence (AI) to automatically generate dynamic biomimetic façade designs for reducing building carbon emissions. Biomimetic façades show strong carbon-reduction potential; however, their application remains limited by interdisciplinary requirements and time-intensive optimization processes. Existing studies primarily rely on traditional multi-objective optimization for energy performance, while machine learning integration and carbon-oriented evaluation remain limited in biomimetic façade research. To address this gap, this study proposes an AI system for biomimetic façade generation in tropical climates by combining reinforcement learning–based multi-objective optimization with deep learning–based parameter prediction models. A carbon payback assessment method integrating operational and embodied carbon is further proposed to evaluate carbon reduction performance. Preliminary validation through pilot experiments and K-fold cross-validation achieved an average RMSE of 8.7% and an average R² value of 0.547, while façade parameter prediction for new building conditions could be completed within approximately 10 s. Simulated cases also indicated that the generated façade strategies generally remained within predefined carbon payback thresholds under different material configurations. The framework supports carbon-oriented biomimetic façade design and early-stage low-carbon design decision-making. Full article

Given the complex operating conditions and latent faults exhibited by electric vehicle charging infrastructure amid massive order volumes, traditional monitoring methods based on thresholds or single statistical metrics struggle to detect dynamic, time-varying anomalies. This paper proposes a method for identifying and tracing the operational status of charging facilities based on the k-shape time-series clustering algorithm. This method directly uses charging current time series as the research object, eliminating the cumbersome manual feature extraction process. By utilizing a shape-based distance (SBD) metric strategy, it overcomes common time-series data issues such as phase shifts and amplitude scaling while preserving the integrity of the time dimension. Through iterative calculation of cluster centroids, the algorithm successfully and adaptively classifies massive amounts of data into typical clusters such as “standard charging,” “deep oscillation,” and “power-limited.” Based on the clustering results, this paper further constructs a “shape-operating condition” mapping mechanism. Combined with a Bayesian posterior probability model, this enables the localization of high-risk “vehicle-charger” combinations statistically associated with abnormal waveforms. Empirical studies demonstrate that this method can effectively identify equipment performance degradation at the micro-level of waveforms and provide prioritized inspection clues for the intelligent operation and maintenance of charging networks. Full article