Search Results (2,401)

In recent years, the demand for efficient neural networks in embedded contexts has grown, driven by the need for real-time inference with limited resources. While GPUs offer high performance, their size, power consumption, and cost often make them unsuitable for constrained or large-scale applications. FPGAs have therefore emerged as a promising alternative, combining reconfigurability, parallelism, and increasingly favorable cost–performance ratios. They are especially relevant in domains such as robotics, IoT, and autonomous drones, where rapid sensor fusion and low power consumption are critical. This work presents the full implementation of a neural network on a low-cost FPGA, targeting real-time image and video recognition for drone applications. The workflow included training and quantizing a YOLOv3-Tiny model with Brevitas and PyTorch, converting it into hardware logic using the FINN framework, and optimizing the hardware design to maximize use of the reprogrammable silicon area and inference time. A custom driver was also developed to allow the device to operate as a TPU. The resulting accelerator, deployed on a Xilinx Zynq-7020, could recognize 208 frames per second (FPS) when running at a 200 MHz clock frequency, while consuming only 2.55 W. Compared to Google’s Coral Edge TPU, the system offers similar inference speed with greater flexibility, and outperforms other FPGA-based approaches in the literature by a factor of three to seven in terms of FPS/W. Full article

This study tested whether classic and specific bioelectrical impedance vector analysis (BIVA) parameters could explain metabolic and mechanical performance in endurance-trained trail runners. Fifteen males ($\dot{\mathrm{V}}$O₂max 61.04 ± 6.91 mL·kg⁻¹·min⁻¹) completed a 60-min treadmill protocol at 70% $\dot{\mathrm{V}}$O₂max across randomized slopes (−7% to +7%), with continuous gas-exchange, heart-rate, and running-power recording; whole-body BIVA was obtained immediately pre- and post-exercise. Post-test, impedance and resistance increased (+2.73%, +2.84%), while reactance (Xc) and phase angle decreased (−2.36%, −4.91%); all were significant and mirrored by both classic and specific indices, consistent with acute fluid loss and altered cellular status. After Benjamini–Hochberg adjustment, baseline Xc/height correlated inversely with $\dot{\mathrm{V}}$CO₂peak and $\dot{\mathrm{V}}$CO₂mean, whereas exercise-induced changes in ΔXc/height and ΔXcspecific correlated positively with both metabolic variables and mean power. Stepwise regression retained ΔXc/h or ΔXcspecific as the only BIVA predictors for $\dot{\mathrm{V}}$CO₂peak, $\dot{\mathrm{V}}$CO₂mean, and mean power output, explaining ~31–36% and ~22–23% of the variance, respectively; classic and specific approaches performed similarly. No bioelectrical variable predicted $\dot{\mathrm{V}}$O₂max. These preliminary findings suggest that acute reactance shifts may provide a modest yet sensitive, non-invasive index of exercise-induced physiological responses, warranting confirmation in larger and more diverse cohorts. Full article

This study compared two physiological modeling approaches, the Peronnet-Thibault (P-T) model and the Minimal Power (MP) model, to identify key parameters representing individual physiological characteristics and to explore their applications in running training. Model parameters were estimated using nonlinear least squares fitting, and predictive performance was evaluated by the mean absolute error (MAE). Results from the World Running Records (WRR) indicated that the MP model generally outperformed the P-T model in linking running performance with physiological variables, demonstrating greater capability in extracting physiological parameters. Further validation using the British Runner Records (BRR) showed that the MP model achieved MAE values of 3.02% for males and 3.47% for females, reflecting strong generalization to real running performance. Furthermore, descriptive analyses of the relationships between MP model parameters and running performance further support its potential value in personalized training and performance prediction. Full article

The transition to cleaner, hydrogen-containing fuels is critical for reducing the environmental impact of marine infrastructure, yet their potential effects on the durability and mechanical reliability of engine components remain a significant engineering challenge. Although aluminum alloys are generally regarded as less susceptible to hydrogen-induced degradation and are widely applied in internal combustion engine components, experimental data obtained under real operating conditions with hydrogen-containing fuel mixtures remain insufficient to fully assess all potential risks. In the present study, two identical low-power gasoline engine–generators were operated for 220 h on fuels with and without hydrogen. Post-test analysis included mechanical testing and microstructural characterization of aluminum alloy pistons for comparative assessment. The measured values of ultimate tensile strength, elongation and deflection, maximum bending force, and effective stress concentration factor revealed pronounced property degradation in the piston operated on the gasoline–hydrogen mixture compared to both the new piston and the one run on pure gasoline. Microstructural analysis provided a plausible explanation for this degradation. The results of this preliminary study provide insights into the effects of hydrogen-containing fuel on the mechanical performance of engine component alloys, contributing to the development of safer and more reliable marine energy systems. Full article

In the context of low carbon, achieving accurate forecasting of electrical energy is critical for power management with the continuous development of power systems. For the sake of improving the performance of load forecasting, an improved MNet-Atten electric vehicle charging load forecasting based on composite decomposition and the evolutionary predator–prey and strategy model is proposed. In this light, through the data decomposition theory, each subsequence is processed using complementary ensemble empirical mode decomposition and filters out high-frequency white noise by using singular value decomposition based on matrix operation, which improves the anti-interference ability and computational efficiency of the model. In the model construction stage, the MNet-Atten prediction model is developed and constructed. The convolution module is used to mine the local dependencies of the sequences, and the long term and short-term features of the data are extracted through the loop and loop skip modules to improve the predictability of the data itself. Furthermore, the evolutionary predator and prey strategy is used to iteratively optimize the learning rate of the MNet-Atten for improving the forecasting performance and convergence speed of the model. The autoregressive module is used to enhance the ability of the neural network to identify linear features and improve the prediction performance of the model. Increasing temporal attention to give more weight to important features for global and local linkage capture. Additionally, the electric vehicle charging load data in a certain region, as an example, is verified, and the average value of 30 running times of the combined model proposed is 117.3231 s, and the correlation coefficient PCC of the CEEMD-SVD-EPPS-MNet-Atten model is closer to 1. Furthermore, the CEEMD-SVD-EPPS-MNet-Atten model has the lowest MAPE, RMSE, and PCC. The results show that the model in this paper can better extract the characteristics of the data, improve the modeling efficiency, and have a high data prediction accuracy. Full article

Early train detection is vital for ensuring the safety of railway personnel, particularly in remote locations where fixed signaling infrastructure is unavailable. Unlike many existing solutions that rely on high-power, high-cost sensors and compute platforms, this work presents a lightweight, low-cost, and portable framework designed to run entirely on resource-constrained microcontrollers with just kilobytes of Random Access Memory (RAM). The proposed system uses vibration data from low-cost accelerometers and employs a simple yet effective Linear Regression (LR) model for almost real-time prediction of train arrival times. To ensure feasibility on low-end hardware, a parallel-processing framework is introduced, enabling continuous data collection, Machine Learning (ML) inference, and wireless communication with strict timing and energy constraints. The decision-making process, including data preprocessing and ML prediction, completes in under 10 ms, and alerts are transmitted via LoRa, enabling kilometer-range communication. Field tests on active railway lines confirm that the system detects approaching trains 15 s in advance with no false negatives and a small number of explainable false positives. Power characterization demonstrates that the system can operate for more than 6 days on a 10 Ah battery, with potential for months of operation using wake-on-vibration modes. Full article

LoRa is a long-range, low-power wireless communication technology widely used in Internet of Things (IoT) applications. However, its conventional implementation through Long Range Wide Area Network (LoRaWAN) presents operational constraints due to its centralized topology and reliance on gateways. To overcome these limitations, this work designs and validates a gateway-free mesh communication system that operates directly on commercially available commodity microcontrollers, implementing lightweight self-healing mechanisms suitable for resource-constrained devices. The system, based on ESP32 microcontrollers and LoRa modulation, adopts a mesh topology with custom mechanisms including neighbor-based routing, hop-by-hop acknowledgments (ACKs), and controlled retransmissions. Reliability is achieved through hop-by-hop acknowledgments, listen-before-talk (LBT) channel access, and duplicate suppression using alternate link triggering (ALT). A modular prototype was developed and tested under three scenarios such as ideal conditions, intermediate node failure, and extended urban deployment. Results showed robust performance, achieving a Packet Delivery Ratio (PDR), the percentage of successfully delivered DATA packets over those sent, of up to 95% in controlled environments and 75% under urban conditions. In the failure scenario, an average Packet Recovery Ratio (PRR), the proportion of lost packets successfully recovered through retransmissions, of 88.33% was achieved, validating the system’s self-healing capabilities. Each scenario was executed in five independent runs, with values calculated for both traffic directions and averaged. These findings confirm that a compact and fault-tolerant LoRa mesh network, operating without gateways, can be effectively implemented on commodity ESP32-S3 + SX1262 hardware. Full article

The paper finds optimal power flows and optimal placement of wind power plants (WPPs), static var compensators (SVCs), and thyristor-controlled series capacitors (TCSCs) in the IEEE 30-bus transmission power network by applying three high-performance algorithms, such as the equilibrium optimizer (EO), the Coot optimization algorithm (COOT), and the marine predators algorithm (MPSA). The three algorithms are run for the system without any added electric components and with three single objectives, including active power losses, total fuel cost, and total voltage deviation, for comparison with other previous algorithms. The three algorithms can reach better results than many algorithms and suffer worse results than a few algorithms. EO is more effective than MPSA and COOT in all cases. For simulation cases with SVCs, TCSCs, and WPPs, the losses are significantly reduced compared to the base case. The power loss of the base case is 3.066 MW, and the best loss is 2.869 MW for two cases with two SVCs and one TCSC. When applying the obtained solution and optimizing the placement of one, two, and three WPPs, the power loss is, respectively, 2.053, 1.512, and 1.112 MW. By optimizing two SVCs, one TCSC, and WPPs simultaneously, the power loss is, respectively, 2.041, 1.508, and 1.093 MW for one, two, and three WPPs. So, the optimal placement of TCSCs, SVCs, and WPPs can result in high benefits for power systems. Full article

Forecasting outcomes in high-stakes competitive spectacles like the Olympic Games, World Cups, and professional league championships has grown increasingly vital, directly impacting strategic planning, resource allocation, and performance optimization across a multitude of fields. However, accurate forecasting remains challenging due to complex, nonlinear interactions inherent in high-dimensional time-series data, further complicated by socioeconomic indicators, historical influences, and host-country advantages. In this study, we propose a comprehensive forecasting framework integrating structured time-series modeling with ensemble learning. We extract key structural features via two novel indices: the Advantage Index (measuring a competitor’s dominance in specific areas) and the Herfindahl Index (quantifying performance outcome concentration). We also evaluate host-country advantage using a Difference-in-Differences (DiD) approach. Leveraging these insights, we develop a dual-branch predictive model combining an Attention-augmented Long Short-Term Memory (Attention-LSTM) network and a Random Forest classifier. Attention-LSTM captures long-term dependencies and dynamic patterns in structured temporal data, while Random Forest handles predictions for unrecognized contenders, addressing zero-inflation issues. Extensive stability and comparative analyses demonstrate that our model outperforms traditional and state-of-the-art methods, exhibiting strong resilience to input perturbations, consistent performance across multiple runs, and appropriate sensitivity to key features. Our key contributions include the development of a novel integrated forecasting framework, the introduction of two innovative structural indices for competitive dynamics analysis, and the demonstration of robust predictive performance that bridges technical innovation with practical strategic application. Finally, we transform our modeling insights into actionable strategic insights. This translation is powered by interpretable feature importance rankings and stability analysis that rigorously validate the robustness of key predictors. These insights apply across multiple dimensions—encompassing advantage assessment, resource distribution, strategic simulation, and breakthrough potential identification—providing comprehensive decision support for strategic planners and policymakers navigating competitive environments. Full article

This study provides a comprehensive empirical examination of the drivers and dynamics of Juglar cycles in China from 1981 to 2024. We develop a unified framework that integrates investment, institutional, productivity, and structural factors, and employ a Vector Error Correction Model to analyze the long-run equilibrium and short-run adjustment mechanisms linking fixed asset investment (FAI), government fiscal expenditure (GFE), total factor productivity (TFP), industrial structure upgrading (ISU), and gross domestic product (GDP). Our results confirm a stable cointegration relationship and identify FAI as the most influential long-run driver of output, with a 1% increase in FAI leading to a 0.88% rise in GDP. Industrial upgrading also exerts a positive long-run influence on growth, whereas government spending exhibits a significant negative effect, potentially indicating crowding-out or efficiency losses. In the short run, we find unidirectional Granger causality from FAI to GDP, suggesting that changes in investment contain meaningful predictive power for future output fluctuations. Furthermore, impulse response and variance decomposition analyses illustrate the temporal evolution of these effects, highlighting that the contribution of TFP gains importance over the medium term. Overall, this study deepens our understanding of business cycle transmission mechanisms in emerging economies and offers valuable insights for policymakers seeking to balance investment-driven growth with structural reforms for sustainable and robust economic development. Full article

This paper presents the final implemented design and performance evaluation of the ground-based C-band Cheia radar system, developed to enhance Romania’s contribution to the EU Space Surveillance and Tracking (EU SST) network. All data used for performance analysis are real-time, real-life measurements of true spatial test objects orbiting Earth. The radar is based on two decommissioned 32 m satellite communication antennas already present at the Cheia Satellite Communication Center, that were retrofitted for radar operation in a quasi-monostatic architecture. A Linear Frequency Modulated Continuous Wave (LFMCW) Radar design was implemented, using low transmitted power (2.5 kW) and advanced software-defined signal processing for detection and tracking of Low Earth Orbit (LEO) targets. System validation involved dry-run acceptance tests and calibration campaigns with known reference satellites. The radar demonstrated accurate measurements of range, Doppler velocity, and angular coordinates, with the capability to detect objects with radar cross-sections as low as 0.03 m² at slant ranges up to 1200 km. Tracking of medium and large Radar Cross Section (RCS) targets remained robust under both fair and adverse weather conditions. This work highlights the feasibility of re-purposing legacy satellite infrastructure for SST applications. The Cheia radar provides a cost-effective, EUSST-compliant performance solution using primarily commercial off-the-shelf components. The system strengthens the EU SST network while demonstrating the advantages of LFMCW radar architectures in electromagnetically congested environments. Full article

Portable and wearable sensors have gained attention in recent years to perform measurements in many different applications. Sensors based on Electrical Impedance Spectroscopy (EIS) are particularly promising, because they can make accurate measurements with minimum perturbation to the sample under test. Electrochemical biosensors are devices that use electrochemical techniques to measure a target analyte. In the case of electrochemical biosensors based on EIS, the measured impedance spectrum is fitted to that of an equivalent electrical circuit, whose component values are then used to estimate the concentration of the target analyte. Fitting EIS data is usually carried out by sophisticated algorithms running on a PC. In this paper, we have evaluated the feasibility to perform EIS data fitting using simple Artificial Neural Networks (ANNs) that can be run on resource constrained microcontrollers, which are typically used for portable and wearable sensors. We considered a typical case of an impedance spectrum in the range 0.1 Hz–10 kHz, modeled by using the simplified Randles equivalent circuit. Our analyses have shown that simple ANNs can be a low power alternative to perform EIS data fitting on low-cost microcontrollers with a memory occupation in the order of kilo bytes and a measurement accuracy between 1% and 3%. Full article

Investing in large-scale hydropower is on the rise in Ethiopia in accordance with the country’s climate-resilient green economy strategy. Rural electrification is a top priority on the development agenda of the country, with very limited off-grid interventions. Although small-scale hydropower can bring various social and economic benefits compared to other off-grid solutions, it is hardly localized in the country. The motivation for this research is to break this technological bottleneck by synergizing and strengthening the local capacity. Accordingly, this paper presents the full-scale crossflow turbine design and development process of a power plant constructed to give electricity access to about 450 households in a rural village called Amentila. Based on a site survey and the resource potential, the power plant was designed for a 125 kW peak at 0.3 m³/s of discharge with a 53 m head. The crossflow was selected based on the head, discharge, and simplicity of development with the available local capacities. The detailed design of the turbine and its auxiliary components was developed and simulated using SolidWorks and CFD ANSYS CFX. The power plant has a run-of-river design, targeting provision of power during peak hours. This study demonstrates an off-grid engineering solution with applied research on the water–energy–food–environment nexus. Full article

This work suggests a new use for the Non-Monopolize Optimization (NO) method to improve the dynamic stability and robustness of PID and PIDA controllers in Automatic Voltage Regulator (AVR) systems when there are load disruptions. The NO algorithm is a new search method that does not use metaphors and only looks for one answer. It utilizes adaptive dimension modifications to strike a balance between exploration and exploitation. Its addition to AVR control makes parameter tweaking more efficient, without relying on random metaphors or population-based heuristics. MATLAB/Simulink R2025a runs full simulations to check how well the system works in both the time domain (step response, root locus) and the frequency domain (Bode plot). We compare the results to those of well-known optimizers like WOA, TLBO, ARO, GOA, and GA. The suggested NO-based PID and PIDA controllers always show less overshoot, faster rise and settling periods, and higher phase and gain margins, which proves that they are more stable and responsive. A robustness test with a load change of ±50% shows that NO-tuned controllers are even more reliable. The results show that using NO to tune different controllers could be a good choice for real-time AVR controller tuning in modern power systems because it is lightweight and works well. Full article

Additively manufactured thermoplastic polyurethane (TPU 95A) is widely used in engineering, yet its machining behavior remains insufficiently explored. This study investigates the post-processing machinability of FDM-fabricated TPU 95A using CNC milling, with a particular focus on material removal rate (MRR) and surface roughness (Ra). A full factorial design of experiments (81 runs) is conducted, considering four input parameters such as spindle speed (N; 2000, 4000, 6000 rpm) and feed rate (F; 100, 200, 300 mm/min) on the CNC vertical machining center, together with infill density (ϕ; 33%, 66%, 100%) and layer thickness (LT; 1.0, 1.5, 2.0 mm). MRR is modeled and optimized across all densities, achieving strong fit (R² = 0.94; Adj-R² = 0.93). The optimum conditions are found to be MRR ≈ 1251 mm³/min at F = 300 mm/min, ϕ = 100%, N ≈ 3500 rpm and LT ≈ 1.05 mm. Ra can only be measured for 100% infill specimens, as lower infill surfaces violate profile measurement requirements. Its regression model shows weak explanatory power (R² = 0.14; Adj-R² = 0.03) and is excluded from optimization. Instead, Ra is reported descriptively: milling reduced roughness from ≈25–30 μm (as-printed) to ≈13.8 μm under favorable conditions. Overall, the study highlights machining’s role in the hybrid manufacturing practice. Full article