Many forensic methods have been developed for uncovering authors of threatening letters to get rid of their anonymous nature. With the development of new tools for the digital creation of threatening letters, such methods became obsolete, yet new options for their digital analysis also became available. This paper covers a method for digital analysis of threatening letters written in the Croatian language using machine learning algorithms. First the techniques for extraction of style elements of a text are covered, focusing on the most influential elements, and then transforming the extracted data into a vectorized form which can be used in machine learning models. The models are built upon the autoencoder architecture in combination with two text style vectorization methods. Multiple such models are implemented and their results are compared and analyzed, focusing on their effectiveness and the quality of their results. Finally, additional possibilities for further development of machine learning models for the analysis of threatening letters written in the Croatian language are covered and their possible effectiveness explained.
This study presents the development and validation of an intelligent error compensator based on Long Short-Term Memory (LSTM) recurrent neural networks for dynamic weighing systems in copper concentrate belt conveyors. Conventional weighing systems fail to capture nonlinear temporal dynamics, leading to measurement inaccuracies during container filling operations. The methodology comprised data acquisition from load cells, speed sensors, and inclinometers; systematic hyperparameter optimization; and evaluation using Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and coefficient of determination (R2). Hyperparameter optimization identified an optimal configuration with one LSTM layer (20 units, learning rate 0.001, window size 20 steps). Evaluation on an independent test set showed that the compensator reduced MAPE from 8.5% (uncompensated system) to 3.01%, representing a 64.6% improvement, and reduced RMSE from 12.3 to 4.2 tons (65.9% improvement), with an R2 of 0.95. Feature importance analysis confirmed physical consistency, with load cell voltage as the dominant predictor (42%). These results demonstrate that LSTM-based compensation significantly enhances weighing accuracy. The study provides a replicable framework for industrial metrology modernization, contributing to sustainable mining operations through material loss reduction and logistics optimization. While the proposed model has been validated offline using historical data, its deployment in the live production environment remains pending.
In short-duration power-support applications of energy storage stations, state of power (SOP) estimation should reflect the constant-power boundary over the target horizon, while constant-current extrapolation may misrepresent the current rise caused by voltage decline. This study proposes a 30 s constant-power SOP evaluation framework for portable inspection, decoupling parameter inversion from boundary propagation. The method uses a single-particle model with electrolyte dynamics (SPMe) with degradation factors for ohmic resistance, kinetics, and diffusion. The ohmic degradation factor is determined through time-zero voltage-drop hard calibration, while the kinetic and diffusion degradation factors are identified from 30 s constant-current pulse responses using physics-informed neural network (PINN)-based inversion, and the constant-power boundary is solved by Runge–Kutta integration and bisection search. In model-consistent closed-loop verification, which assesses numerical and inversion consistency under matched-model assumptions rather than independent physical accuracy, the method achieved a mean absolute error (MAE) of 0.100%, a 95th-percentile error of 0.503%, and a maximum error of 2.019%, below the constant-current approximation and first-order equivalent circuit model baselines within the matched-model synthetic setting. Its Jetson Nano-equivalent runtime was approximately 0.630 s. An external proxy comparison using 154 discharge pulses from a public HPPC dataset for an LCO-graphite cell showed an MAE of 0.41 W relative to the pulse-power proxy. This result measures agreement with the selected pulse-power proxy rather than accuracy against a strictly defined 30 s constant-power ground truth. The 10 mV-noise case increased the SOP MAE to 3.868%, indicating substantial sensitivity to voltage-measurement disturbance and the need for validated signal conditioning. These results indicate a physically interpretable and computationally feasible candidate framework for rapid battery power screening, while direct constant-power experiments, broader chemistry coverage, and measured-noise validation remain necessary before field deployment.
To address the challenges of integrating multi-source heterogeneous data, fragmented fault knowledge, and the limited capability of traditional rule engines in recognizing edge cases for ultra-high voltage (UHV) bushing fault diagnosis, this paper proposes a fault identification and decision-making method based on knowledge graph (KG) rule reasoning and inductive graph convolutional network (Inductive GCN). First, a triple-matching strategy is employed to perform entity extraction and relation mining from fault cases, constructing a fault knowledge graph that transforms unstructured fault case texts into a structured knowledge graph. Second, a rule engine based on a multi-source feature rule set is designed, utilizing the entropy weight method and the RETE algorithm to achieve interpretable symbolic reasoning. On this basis, a double-layer inductive graph convolutional network is introduced to learn implicit fault patterns by aggregating topological information from neighboring nodes, and a confidence-driven dynamic weighted fusion strategy is adopted to achieve complementary advantages between the two models. Finally, a large language model is introduced to generate operation and maintenance decision recommendations. Experimental results demonstrate that the proposed method achieves an identification accuracy of 98.1% on a test set of 159 samples, which is 10.7 percentage points higher than that of a single rule engine and 6.9 percentage points higher than that of a single inductive graph convolution network. The standard deviation of accuracy across different test batches is only 0.0029. These results demonstrate the effectiveness and stability of the proposed method, providing a practical technical solution for UHV bushing fault identification.
In the present work, graphene-modified zinc oxide (ZnO-rGO) and fluorine-doped tin oxide (FTO) thin films were successfully fabricated using a simple, low-cost pulsed spray pyrolysis technique. The structural, morphological, optical, electrical, and surface electronic properties of the deposited films were systematically characterized. X-ray diffraction (XRD) analysis confirmed the formation of polycrystalline ZnO- and SnO2-based phases with crystallite sizes in the nanometer range. The crystallographic parameters, microstrain, and dislocation density of the deposited films were found to be influenced by the incorporation of reduced graphene oxide (rGO) and fluorine dopants. Scanning electron microscopy (SEM) revealed compact and homogeneous surface morphologies with good film coverage and well-defined nanocrystalline features. Optical characterization demonstrated the wide-bandgap semiconducting behavior of the deposited films, with optical bandgap energies ranging from 3.262 to 3.312 eV for the ZnO-rGO films and from 3.91 to 4.01 eV for the FTO films. Kelvin probe measurements yielded work-function values in the range of approximately 5.0–5.2 eV, indicating favorable surface electronic characteristics suitable for optoelectronic applications. Furthermore, fluorine incorporation enhanced the dielectric response of the SnO2 films, particularly in the low-frequency region owing to increased interfacial polarization effects. The obtained results demonstrate that pulsed spray pyrolysis provides a simple, cost-effective, and efficient route for fabricating ZnO-rGO and FTO thin films with desirable structural, optical, electrical, and surface electronic properties. These findings highlight the considerable potential of the developed materials for transparent electrodes and a wide range of optoelectronic applications.
Ultra-high-voltage direct-current (UHVDC) transmission systems impose stringent requirements on the reliability of insulation materials used in converter transformer bushings. Epoxy resin systems are key insulating materials in resin-impregnated paper (RIP) capacitor bushings, and their processing characteristics, curing behavior, and electrical properties directly affect bushing performance. In this study, two epoxy insulation systems used for resin-impregnated paper (RIP) bushings, namely the imported Araldite LY1564/Aradur 3486 system and the domestic EP-2020/CA-3015 system, were systematically investigated through viscosity, curing, and electrical property tests. The results show that the viscosities of both resins decreased significantly with increasing temperature. At 60 °C, the viscosities of Resin A and Resin B were 151.6 mPa·s and 156.3 mPa·s, respectively. The mixed resin–hardener systems exhibited similar viscosity evolution and comparable pot life characteristics. DSC measurements revealed two-stage curing reactions for both materials, with first exothermic peak temperatures of 65.4 °C and 96.3 °C and second peak temperatures of 269.3 °C and 269.8 °C for Materials A and B, respectively. Electrical testing demonstrated that both materials exhibited similar temperature-dependent dielectric and resistivity behavior, with dielectric loss increasing at elevated temperatures and resistivity decreasing as temperature increased. The volume resistivity trends and dielectric characteristics of the two materials remained highly consistent throughout the investigated temperature range. The results indicate that Material B exhibits processing performance, curing characteristics, and electrical insulation properties comparable to those of Material A. Therefore, Material B demonstrates strong potential for application in UHVDC RIP bushing insulation systems and provides a promising alternative for the localization of key insulating materials.
This paper presents an attitude filter that encodes both orientation and uncertainty in a single four-dimensional vector, requiring no covariance propagation or normalization constraints. The filter state is the natural parameter of the von Mises–Fisher (vMF) distribution on S3, whose exponential family structure reduces measurement updates to vector addition. Prediction is governed by a continuous-time ODE (Ordinary Differential Equation) that couples rotational kinematics with concentration decay. The QUEST-based construction of measurement natural parameters with a Fisher-information-matched concentration, an antipodal switching mechanism for the quaternion double cover, and a global exponential convergence analysis of the attitude error are described. The filter construction is left-invariant: it commutes with rotations of the reference frame, making the error dynamics trajectory-independent. The result is a filter with the computational simplicity of a complementary filter and the statistical grounding of Bayesian vMF fusion, operating entirely in unconstrained ℝ4 space. The filter is validated in simulation, on two recorded flights—including an evaluation against an EFIS attitude reference—and its computational cost is measured down to on-target microcontroller cycle counts. Gyroscope bias estimation is not included and is left to future work.
This article proposes an approach for acquiring HSQC (Heteronuclear Single Quantum Coherence) spectra using a high-resolution probehead originally designed for solutions. The method is suitable for studying organic substances with relatively low melting points, such as copolymers, composites, waxes, resins, and similar materials. The procedure involves preparing a melt of the substance directly inside the NMR (Nuclear Magnetic Resonance) sample tube prior to analysis. A comparison of HSQC spectra obtained from both the molten state and a solution of the same substance demonstrates that representative spectra can be acquired, enabling detailed analysis of their fine structure. The method has been successfully tested on a range of materials, including: paraffin, wax, honey, vanillin, polycaprolactone, a copolymer of lactide with phenol and maleic anhydride, composites of polycaprolactone and vanillin. This approach enables the identification of impurities in polymers and biological samples without requiring expensive deuterated solvents.
High-frequency (HF) RFID systems operating at 13.56 MHz are widely used in applications such as near-field communication, contactless identification, and smart sensing. Their performance strongly depends on the inductive coupling between the reader and tag antennas, which is influenced by antenna geometry, relative position, orientation, and environmental conditions. This work investigates the antenna component of passive HF RFID tags, represented by planar spiral inductors, without integrating an RFID microchip, allowing the electromagnetic coupling to be analyzed independently of chip-specific effects. A low-cost automated measurement platform was developed to experimentally evaluate the influence of antenna geometry, distance, orientation, and temperature on inductively coupled HF RFID antennas. The platform combined an automated positioning system with a mobile application for remote operation, minimizing the influence of the operator during measurements. A second experimental setup was designed to investigate the effect of temperature on antenna performance. Experimental results show that rectangular spiral antennas generally provided stronger inductive coupling than the other geometries investigated. Furthermore, varying the receiving antenna orientation improved the coupling between rectangular and octagonal antennas under specific configurations. Temperature variations within the investigated range had only a minor influence on antenna performance. The proposed platform provides a low-cost, portable, and reproducible solution for the experimental characterization of HF RFID antennas operating at 13.56 MHz.
This paper presents a novel method for designing manufacturable porous infill structures using a Voronoi-based topology optimization framework. By integrating discrete Voronoi representations into density-based topology optimization in a differentiable manner, the method enables variable-thickness edge structures, with Euclidean distance fields generated from seed points. The material distribution and structural shape are determined by the seed point locations and the distance tensor, which serve as the design variables in this work. As the seed points are directly associated with the dual graph of the Voronoi diagram (VD), namely the Delaunay triangulation (DT), a constraint is formulated based on the DT to ensure the manufacturability of the infill structure. This is achieved by constraining all edge angles of the DT to satisfy the overhang requirement. Since 3D printers can fabricate overhanging structures up to a certain length, VD edges shorter than this threshold are exempt from the self-supporting constraint. To reduce the number of design variables and simplify the manufacturability constraint, a merging strategy is introduced to combine seed points that are sufficiently close during the optimization process. To ensure manufacturability of the outer surface, a set of seed points is additionally sampled on the outer boundary and kept fixed throughout optimization. The proposed method is validated on both regular and irregular 2D design domains, and the results demonstrate its capability to generate manufacturable porous infill structures with satisfactory mechanical performance.
Camouflage design for jungle environments has conventionally relied on the static optimization of color, texture, and edge features, presuming that the background remains visually stable. This presumption diverges from real conditions, in which wind continuously alters leaf orientation and vegetation texture, leaving a gap between static optimization and dynamic visual reality. To address this limitation, this study developed a systematic camouflage design process that integrates the Beaufort scale into a mimetic system for simulating vegetation sway. Dominant colors were extracted using the CIE L*a*b* color space and K-means clustering, and background maps were generated via Gaussian blur. Leaf textures from five plant species were arranged through seamless tiling and overlaid onto the backgrounds to form 15 camouflage samples. Validation employed a fuzzy logic questionnaire and eye-tracking measurements. Under the present experimental conditions, which used screen presentation under visible light, pattern A-13 performed best. Derived from the Terminalia mantaly leaf texture in the dark green variant, it achieved the most favorable balance between distinctiveness from the regional reference pattern and disruption of target–background segmentation, whereas C-15, the light green variant, consistently ranked last. The proposed process is reproducible and applicable to civilian equipment such as tents and backpacks.
Global shifts in energy policy have contributed to an increase in electricity generation from renewable sources, which introduces unique issues with volatility and grid reliability. Robust grid-scale energy storage methods must fill the gap between generation and consumption. Flywheel energy storage (FES) is a mechanical technology that utilizes the stored kinetic energy of a rotating body, but is typically only suited for shorter-term frequency regulation due to significant windage losses. In this work, a novel Python 3.13-based simulation and optimization tool is presented and used to optimize geometric design parameters for efficiency, energy density, and other metrics. The simulation utilizes a 1 degree-of-freedom, multi-regime fluid friction model with a time-marching algorithm. The optimization functionality utilizes pyswarms, a particle swarm optimization package, with adjustable search parameters and cost functions to evaluate simulation results. Optimization parameters include geometric parameters of rotor radius, shaft radius, airgap width, and airgap height; material properties of mass and moment of inertia; and initial angular velocity. An optimal initial angular velocity is found for a particular geometry, lasting 30 times longer until self-discharge versus the worst values. This work can inform the design of flywheel systems to minimize windage losses and promote the technology’s utility for longer-term energy storage.
Machine learning’s computational demands necessitate optimal performance and utilization across diverse hardware architectures. This research compares computing as spiking neural networks (CSNNs, or simulated neuromorphic computing) and regular CNNs on Apple Silicon M3 Pro with Metal Performance Shaders (MPS), and NVIDIA RTX 3070 GPU with CUDA. We run Convolutional Spiking Neural Networks (CSNNs) and traditional CNNs on two datasets (frame-based CIFAR-10; and sequential event-based DVS) to evaluate the suitability of neural net architectures and platforms for different data problems. For both CSNNs and traditional CNNs, Apple Silicon with MPS delivers better energy efficiency but longer processing times for training and inference. NVIDIA with CUDA offers faster computation in training and inference at higher energy costs for CNNs. For CSNNs, frame-based data (CIFAR-10) significantly degraded performance when proper temporal encoding was absent, while event-based data (DVS) proved more naturally suited to the CSNN architecture than frame-based inputs. Though CNNs still achieved higher empirical accuracy in the reported experiments. CSNNs also performed better on Apple Silicon (with MPS) for the sequential event-based data. RAM utilization patterns favored Apple Silicon (with MPS) across both data experiments. The CSNN architecture demanded higher memory resources than CNN, regardless of platform and dataset. NVIDIA (with CUDA) was less energy efficient for spiking neural networks (CSNNs) as compared to Apple Silicon (with MPS). We also compared how the number of time steps affects accuracy and energy consumption across hardware platforms, finding that higher accuracy correlates with energy costs as time steps increase; the accuracy-energy relation seems linear for frame-based data, while for event-based data the energy consumption remains stable increasing at higher time steps. Our cross-platform performance analysis of spiking and regular neural network architectures highlight the importance of matching platform-architecture combinations to a dataset and application requirements.
The performance of single-photon avalanche diodes (SPADs) is highly dependent on the operating temperature, while traditional SPAD models neglect the self-heating effect induced by avalanche current during long-term device operation, leading to insufficient prediction accuracy. This paper proposes an electro-thermal coupled SPAD simulation model that self-consistently integrates the transient thermal effects of the avalanche process with temperature-dependent electrical parameters, including junction capacitance, breakdown voltage, impact ionization coefficients, and Shockley–Read–Hall (SRH) recombination rates. The complete electro-thermal coupled model is constructed based on Sentaurus-TCAD thermal simulation and Virtuoso circuit simulation and implemented via the Verilog-A language. Simulation results demonstrate that after the device operates for 100 μs under repeated avalanche-quenching processes, the self-heating effect causes a 0.34 V shift in breakdown voltage, increases the device dead time by 3.34 ps, and simultaneously reduces the photon detection probability and elevates the dark count rate. This study conducts a systematic investigation into the performance degradation mechanism of SPAD devices induced by the self-heating effect, laying a theoretical foundation at the device self-heating level for subsequent research on the electrothermal interaction between quenching circuits and device bodies.
Ischemic stroke is a major cause of death and disability and thus requires specialized treatment. The present work describes the design and control-oriented simulation of a smart steerable microcatheter tip based on Nitinol superelastic alloy for thrombectomy. The proposed framework allows for predictive and safe catheter navigation by combining experimental material characterization, electromechanical modeling, and control design. Experimental validations of key material properties, such as hemocompatibility, corrosion resistance, and full superelastic behavior, were incorporated into an environment created in MATLAB/Simulink. The bending curvature of a safe blood vessel was exactly followed by means of delay-guaranteed bandwidth-limited dynamical feedforward and feedback regulation. Simulation-based results validate steering and dynamic response, as well as safe interaction with blood vessel walls. Ultimately, from the work described in this paper, we hope to present a proposal for an entire framework for relating biomaterial properties with control performance that could stimulate safer and more efficient robot-assisted procedures in combating thromboembolic diseases.
With the goal of achieving more accurate wind power predictions by accounting for meteorological influences comprising wind speed, together with wind direction and air pressure, this thesis proposes a method combining fuzzy C-means (FCM) clustering with a TCN–Transformer hybrid model. After preprocessing the data to remove outage and missing records, we apply the Pearson correlation coefficient to identify average wind speed and wind direction that are suitable to serve as input features for the model, together with the atmospheric pressure, as key input features. FCM clustering is then applied to partition the data into low- and high-wind-speed operating conditions, mitigating the accuracy loss caused by uniform modeling. A TCN–Transformer model is subsequently constructed, integrating local temporal feature extraction with global dependency modeling to perform prediction under each condition. The experimental results demonstrate that the proposed FCM–TCN–Transformer framework consistently achieves superior forecasting performance under both low-wind-speed and high-wind-speed conditions. Compared with benchmark models, including TCN, LSTM, GRU, BiGRU, and Transformer, the proposed method achieves lower prediction errors and higher prediction accuracy across different forecasting horizons. Furthermore, repeated experiments with multiple random seeds verify the robustness and stability of the proposed framework. These results indicate that FCM-based wind regime classification effectively reduces data heterogeneity, while the hybrid TCN–Transformer architecture successfully captures both local temporal patterns and long-range temporal dependencies. Therefore, the proposed framework provides an effective and reliable solution for short-term wind power forecasting and contributes to the secure integration of wind energy into modern power systems.
This study proposes a novel simulation-driven intelligent framework for the performance and reliability assessment of renewable energy-integrated pavement systems by unifying coupled multiphysics finite element modeling, structured dataset generation, and graph-based artificial intelligence within a single computational paradigm. The proposed pavement is formulated as a seven-layer multifunctional infrastructure system comprising the asphalt surface, intermediate binder, base layer, thermoelectric energy layer, piezoelectric insert zone, subbase, and subgrade soil, thereby enabling simultaneous consideration of structural load transfer, thermal gradient-driven energy harvesting, moisture-sensitive support behavior, and reliability-oriented performance interpretation. A three-dimensional thermo-hydro-mechanical Abaqus model was developed to simulate the concurrent effects of moving wheel load, solar heat flux, rainfall infiltration, and internal moisture diffusion, and it was subsequently used to construct an AI-ready dataset containing 6000 simulation cases and 68 variables spanning geometric, material, environmental, traffic, uncertainty, structural, thermal, hydraulic, renewable-energy, and probabilistic reliability descriptors. To preserve the physical hierarchy of the layered pavement within the learning process, a Layer-Coupled Reliability Graph Operator Network (LaRGO-Net) was proposed, in which pavement layers are represented as interacting graph nodes linked through adaptive interlayer coupling and optimized through multi-task, physics-aware, and coupling-consistent learning. Experimental evaluation across nine progressive configurations demonstrated a monotonic improvement from baseline dense and graph-convolution models to the full LaRGO-Net formulation. The final model achieved the best overall performance with mean RMSE = 0.040, mean MAE = 0.028, mean R2=0.994, and reliability prediction accuracy characterized by F1 = 99.21 and AUC = 99.53. These results confirm that the proposed framework provides a highly accurate, physically interpretable, and reliability-aware surrogate for next-generation pavement systems capable of simultaneously supporting structural serviceability, renewable-energy functionality, and intelligent decision-making.
The integration of wind energy into power systems relies on forecasting technologies to address operational challenges caused by its volatility and intermittency. This paper proposes a computing architecture for ultra-short-term wind power forecasting. The methodology integrates an adaptive dual-stage signal processing technique with an optimized deep learning model. To manage the non-stationarity of meteorological variables, the Pearson and Maximal Information Coefficient (MIC) analyses are employed for feature selection. The ICEEMDAN algorithm is then used for initial decomposition, followed by sample entropy and K-Means clustering to assess component complexity. Variational Mode Decomposition (VMD) is applied only to the high-frequency component to further separate stochastic fluctuations while preserving relatively stable trend components. A Convolutional Neural Network-Bidirectional Long Short-Term Memory (CNN-BiLSTM) network is constructed to forecast the resulting multi-scale components. To reduce reliance on manual empirical tuning, the Crested Porcupine Optimizer (CPO) is used to fine-tune key network hyperparameters. Evaluations using operational wind-farm data indicate that the developed hybrid method captures the temporal dynamics of wind power and yields lower prediction errors than the tested benchmark models. This research provides a data-driven computing framework for renewable-energy forecasting and related operational analysis.
Intermittent faults in electrical connectors refer to cases in which contact resistance exceeds a specified threshold for microseconds or less, causing transient power or signal interruptions. Accurate detection and quantitative recording of these events are important for connector reliability evaluation. In this work, an eight-channel monitoring system with nanosecond resolution for intermittent faults in electrical connectors is developed, enabling quantitative recording of intermittent events together with dynamic contact resistance (DCR) waveform acquisition. Two-stage programmable amplification is used for DCR measurement, while threshold comparison and FPGA-abased quadrature multiphase oversampling are combined to capture intermittent events. The system supports DCR measurement over 1 mΩ–10,000 mΩ with a maximum relative error of 0.41%, and provides 1.25 ns equivalent time resolution for intermittent event monitoring, with an expanded uncertainty of 0.28 ns–0.54 ns over 20 ns–10 μs. Vibration tests on high-speed connectors further demonstrate that the system captures real intermittent events under mechanical excitation and measures their durations with a maximum relative error of 0.66% relative to oscilloscope results.
Ultra-fast electric vehicle (EV) charging systems are among the most demanding converter-dominated applications due to their high power levels, wide battery-voltage range, strict thermal constraints, and the need for adaptive charging control. Conventional design and tuning approaches often rely on fixed control policies and computationally expensive iterative optimization, which limits their ability to address nonlinear multi-objective trade-offs across the full charging envelope. This paper proposes a hybrid AI–quantum co-design framework for a SiC-based dual active bridge (DAB) converter intended for ultra-fast EV charging applications. The proposed approach combines a physical converter model, an AI surrogate-learning layer for rapid prediction of converter performance, and a quantum-assisted optimization layer for multi-objective exploration of design and control variables. To demonstrate the framework, a representative modular 350 kW ultra-fast charging case study is considered, implemented by four parallel 87.5 kW SiC-based DAB modules and including converter-level optimization and adaptive charging-policy refinement. The revised manuscript introduces a complete system schematic, an explicit DAB converter topology, a clarified methodological workflow, and a simulation-based proof-of-concept evaluation. Representative results indicate improved design-space exploration and more balanced trade-offs between efficiency, thermal stress, ripple, and dynamic response compared with a conventional baseline tuning approach. Although the study does not claim hardware-level quantum advantage, it provides a structured and practically interpretable computational framework for intelligent co-design of high-power charging converters.
