Most audio-based machine-health studies on the MIMII benchmark stop at simulation, and existing edge-oriented anomaly detectors remain too large (8–64 MiB) for microcontroller deployment—leaving open the question of whether an accurate detector can actually run on a low-cost ARM Cortex-M device. This paper addresses that gap with a complete development-to-deployment pipeline for a lightweight YAMNet-based detector on the STM32 B-U585I-IOT02A (Cortex-M33). Its distinct contribution is the joint integration of moving-average denoising, random under-sampling, and an abnormality-thresholding decision rule, validated with the same YAMNet backbone, preprocessing chain and training protocol on both the MIMII fan sub-dataset and a self-collected dataset from 16- and 18-inch Khind fans (normal at three speeds, friction, block); for the three-class real-environment case the output layer is widened from two to three neurons and the decision rule extended to a priority-based scheme, so the architecture is common to both datasets rather than strictly identical. The deployed int8 model occupies only 0.18 MiB Flash with 0.30 s inference, reaching 0.87 ROC-AUC in MIMII simulation and 0.81–0.97 balanced accuracy across 30–70 cm in on-device tests. The MIMII partition is made at file level within pooled machine IDs rather than by recording session or by ID, so the simulation figures are within-session estimates and are reported as such; the on-device tests with live fans, whose audio is generated at test time and is disjoint from every training file, provide the partition-independent evidence. Because comparison protocols differ across studies (supervised vs. unsupervised, dataset subset, hardware), we report results with explicit metric definitions and treat literature comparisons as indicative. The work demonstrates that supervised lightweight transfer learning is deployable on resource-constrained hardware for practical fan monitoring, while noting limitations in data partitioning and validation scale that motivate the planned recording-level cross-validation and power-measurement studies.
The radiation tolerance of mixed-halide perovskites is of considerable interest for next-generation space photovoltaics and nuclear-grade optoelectronics devices operating in harsh radiation environments. In this work, the structural, optical, and electrical evolution of Cs0.15FA0.85PbI2.55Br0.45, Cs0.15FA0.85PbCl0.45I2.55 and Cs0.15FA0.85PbI3 thin films subjected to gamma irradiation doses of 0–25 kGy were systematically investigated using X-Ray Diffraction (XRD), UV-Visible spectroscopy, Scanning Electron Microscopy (SEM), and Hall-Effect Measurements. A non-monotonic radiation response was observed, whereby moderate irradiation doses (10-15 kGy) resulted in improved crystallinity and charge transport properties while higher doses (> 20 kGy) induced structural degradation. Quantitative XRD analysis showed that the crystallite size of Cs0.15FA0.85PbI2.55Br0.45 increased from 51.50 nm to 136.98 nm (approximately 166
This study presents the theoretical design and analysis of a metamaterial-based one dimensional (1D) photonic crystal (PC) refractive index biosensor for detection of PSA. The proposed sensor design is distinguishable from typical 1D PC biosensors in that it comprises an arrangement of metamaterial end-cap layers symmetrically placed on both sides of a defect cavity within a periodic Si/SiO₂ multilayer structure. The metamaterial end-caps significantly improve the sensitivity of the resonance shift readout. The change in PSA concentration is modeled as a slight change in the refractive index of the defect cavity, leading to the observable shift of the photonic band-gap defect mode resonance wavelength. The optical transmission of the structure is computed using the impedance method. The optimal configuration for the biosensor achieves a sensitivity of 3628.9343 nm/RIU, a quality factor of 248.1131, and a figure of merit of 72.4625 RIU⁻¹. The findings indicate that the incorporation of metamaterial layers can significantly enhance the sensing capability of defect mode photonic crystal biosensors. The configuration proposed in this work holds excellent promise for the optical detection of biomolecular targets with high precision and provides an effective platform for further developments in refractive index-based biosensors.
This paper proposes a novel Double-Integral Super-Twisting Direct Power Control (ST-DISMC-DPC) scheme for Dual Active Bridge (DAB) converters. The controller is designed to enhance output voltage regulation under varying load conditions, offering improved robustness and reference tracking compared to conventional PID and standard Double-Integral Sliding Mode Control (DISMC) benchmarks. With a straightforward design that avoids complex modeling and linearization, the ST-DISMC-DPC scheme generates a reference power for direct power control (DPC), computes the phase shift ratio, and achieves fast transient response alongside stable steady-state performance. A rigorous stability analysis of the proposed controller is included to substantiate its theoretical foundation. By incorporating the integral of the voltage error, the controller significantly enhances tracking accuracy and transient mitigation. Quantitative evaluations reveal that the proposed strategy reduces overshoot/undershoot by 56–74
Abstract Accurate Time-to-Empty (TTE) prediction is a cornerstone for optimizing system energy efficiency and enhancing user experience in modern mobile computing. This paper presents a continuous-time multi-physics coupled framework that integrates second-order RC battery dynamics, temperature-dependent thermal feedback, and user-driven modular workloads. Unlike conventional Coulomb-counting or static equivalent-circuit approaches, the proposed framework explicitly captures transient polarization dynamics and temperature-induced impedance growth under realistic burst workloads and sub-zero environments. Validated against Panasonic 18650PF datasets (UW-Madison) at -20℃, the results demonstrate that the proposed framework achieves a Mean Absolute Percentage Error (MAPE) of 1.42% under dynamic UDDS workloads. This work demonstrates that accurate smartphone runtime prediction is strongly influenced by voltage-stability dynamics, providing a scalable and physically interpretable foundation for next-generation intelligent battery-management systems.
This work proposes a synchronization framework that combines multi-switching synchronization with penta compound combination synchronization for nonlinear chaotic systems. The scheme is formulated to realize synchronization between two driving systems and ten response systems by means of suitably designed nonlinear controllers. The developed approach provides stable and efficient synchronization under different switching arrangements among multiple chaotic systems. For demonstrating the validity of the proposed method, the Plasma Torch Jerk chaotic system is employed as a representative example. Numerical investigations are carried out in MATLAB to verify the analytical results and to illustrate the accuracy and stability of the synchronization behavior. Moreover, the study discusses the usefulness of the proposed synchronization model in secure communication, especially for improving data security and encrypted signal transmission in modern communication networks.
Orthogonal Frequency Division Multiplexing (OFDM) is the core physical-layer modulation technique used in modern wireless communication systems due to its high spectral efficiency and robustness against frequency-selective fading. In fifth-generation (5G) systems, channel estimation for OFDM is mainly performed using classical pilot-based methods such as Least Squares (LS) and Minimum Mean Square Error (MMSE) estimation. Although these techniques are computationally efficient and well standardized, their performance degrades in practical fading environments, particularly under high mobility, sparse pilot configurations, and higher-order modulation schemes. This survey reviews the OFDM physical layer and traditional channel estimation techniques in 5G systems and analyzes their limitations using performance results over AWGN and Rayleigh fading channels. The paper further examines recent artificial intelligence and machine learning (AI/ML)-based channel estimation approaches proposed for sixth-generation (6G) systems, highlighting their improved robustness and adaptability in dynamic channel conditions. A comparative analysis between classical 5G and AI/ML-based 6G channel estimation methods is presented in terms of bit error rate performance, adaptability, and complexity, followed by a discussion on the advantages and limitations of intelligent channel estimation for future wireless networks.
This study investigates the operating principles and design of an electro-optic directional coupler (EODC) based on the electro-optic effect for optical logic applications. A comprehensive theoretical framework is presented, incorporating the fundamental mathematical relationships and coupled-mode analysis requires to characterize the switching activity of proposed optical logic devices. Utilizing this methodology, several optical digital logic circuits have been designed and analyzed, including Inverter (NOT), XOR/XNOR, OR/NOR, and NAND/AND gates. Extending this work, the integrated operation of these fundamental gates has also been investigated to realize an optical AND-OR-Inverter (AOI) cell. The proposed EODC is based on a GaAlAs material platform and employs a 3μ m × 3μ m modulator structure with a coupling length of 1 cm . At an operating wavelength of 900 nm , efficient switching is achieved under an applied electric field of approximately 3× 10^4 V/cm, resulting in a refractive-index change of approximately n≅ 1× 10^-4 . In addition to logic implementation, the performance of the proposed structure has been evaluated in terms of extinction ratio, contrast ratio, and amplitude modulation characteristics. The obtained results confirm that electro-optically controlled directional couplers can effectively perform optical signal processing and serve as promising building blocks for future optical computing and high-speed digital photonic systems.
This study evaluates the performance of Global Positioning System (GPS)-only, BeiDou-only, and hybridized GPS+BeiDou Global Navigation Satellite System (GNSS) constellations in Nigeria, with emphasis on positional stability and the benefits of multi-constellation integration. A dual-frequency u-blox ZED-F9P GNSS receiver was deployed at a static reference station at the Department of Pure and Applied Physics, Ladoke Akintola University of Technology, Ogbomoso, Nigeria. Continuous 24-hour GNSS observation data were collected at 4 Hz over a 14-day period in May 2025. The raw data were converted to Receiver Independent Exchange (RINEX) Format using the Real-Time Kinematic Library (RTKLIB) suite and processed with RTKPOST to obtain positioning solutions. The data were analyzed in terms of carrier-to-noise density ratio (C/N₀), number of satellites visible (NSV), satellite elevation angle, positional stability, and Single Point Positioning (SPP) performance. The results showed that GPS provided more stable satellite visibility, with visible satellites ranging from 6 to 11 per epoch and a daily average NSV of 8.22 ± 0.05, while BeiDou ranged from 4 to 8 visible satellites per epoch, with a daily average NSV of 6.05 ± 0.11. GPS maintained 31 visible Pseudo-Random Noise (PRN) satellites throughout the observation period, whereas BeiDou varied between 28 and 31 PRNs. The C/N₀ analysis showed that GPS L1 provided the strongest signal quality, with values ranging from 36.26 dB-Hz to 39.05 dB-Hz and a mean value of 38.10 ± 0.55 dB-Hz. In terms of positional stability, GPS recorded lower standard deviations in latitude, longitude, and altitude, ranging from 1.15 m to 1.75 m, 1.38 m to 1.94 m, and 4.16 m to 5.69 m, respectively. BeiDou showed higher variability, with corresponding values ranging from 2.07 m to 3.39 m, 2.07 m to 4.12 m, and 6.36 m to 9.85 m. However, BeiDou provided a complementary elevation-angle advantage, with a mean elevation angle of 29.06 ± 2.08°, compared with 25.71 ± 2.39° for GPS. The hybridized GPS+BeiDou solution delivered the best positioning performance, with average Two-Distance Root Mean Square (2DRMS), Circular Error Probable (CEP), Spherical Error Probable (SEP), and Mean Radial Spherical Error (MRSE) values of 3.93 m, 1.64 m, 3.66 m, and 4.83 m, respectively. These values represent an improvement of approximately 14
Diketopyrrolopyrrole (DPP) and its derivatives have attracted significant attention in organic electronic devices due to the advantageous features of the DPP core, including strong electron-withdrawing character, high planarity, excellent optoelectronic properties, and thermal stability. Progress in this field has been closely linked to advances in synthetic strategies, particularly chemical modification at the nitrogen position of the DPP core. N-Alkylation represents an effective approach for tuning solubility, molecular packing, optical absorption, and semiconducting behavior. This review focuses on N-alkylated DPP-based small molecules and their applications in organic field-effect transistors (OFETs) and photodetectors. The first part provides a concise historical overview of the DPP core, its key properties, and commonly employed synthetic routes for N-alkyl DPP derivatives. The subsequent part summarizes the device performance of N-alkyl DPPs and their π-expanded derivatives in OFETs and photodetector devices, highlighting structure-property performance relationships. Finally, current challenges and future perspectives are discussed, offering guidance for the rational design of high-performance organic electronic materials. This review will serve as a valuable reference for the researchers working on DPP-based materials and developing high-performance organic electronic devices.
Due to its high efficiency and free from environmental pollution Proton exchange membrane fuel cells (PEMFCs) attracted a lot of attention for clean energy conversion in rural, hospital and vehicle power applications. But the nonlinear characteristics along with the difficulties associated in water management and gas flows make it hard to develop a real-time control strategy. The research develops an integrated model that combines physics modeling, data driven parameter estimation, and control using deep neural networks for PEMFC systems. In this study, an integrated modeling approach using physics modeling, parameter estimation using data-driven techniques, and control using Deep Neural Networks (DNNs) for PEMFC systems has been developed. The main contribution of this study is to integrate a physics-based five-state PEMFC model, an LSTM surrogate model for dynamic predictions, and a DNN Controller, trained from expert PI control policies in a common framework for efficient PEMFC modeling and control. A five-state dynamic model is formulated for simulating the behavior of oxygen, hydrogen, and water vapor inside the PEMFC along with inlet-valve dynamics. Simulation results are used to train an LSTM network to learn and predict the dynamic responses of the system. The trained LSTM model achieved an RMSE of 1.7 × 10⁻⁵ and provided faster simulations than the physics-based model by more than one order of magnitude. An expert PI controller was employed to train a feedforward NN controller which successfully learned the expert’s control strategy with an RMSE of 1.52 × 10− 4 mol/s for hydrogen and 9.85 × 10⁻⁵mol/s for oxygen. Closed-loop validation of the proposed modeling and control approach reveals superior performance of the proposed DNN Controller compared to the traditional PI controllers. Specifically, the DNN Controller showed reduced settling times for tracking (1.76 s vs. 2.64 s) and disturbance rejection (0.30 s vs. 0.57 s) while decreasing overshoot from 3.71
Abstract Machine olfaction is still not embedded in portable devices. Relationships between electronic materials and chemical environments are hard to decipher to build phenomenologic models that formalize smells’ and odors’ subjectivity. Also, machine olfaction requires norms to generate large datasets: a generic technology, easy to manufacture and integrable to the broadest range of electronic devices must be identified. To this aim, this study reports on relevant materials to be processed on cost-effective electronic noses. Using a single water-soluble conducting polymer but sensitized with various p-dopants deposited directly on the circuit board, solvent vapors are successfully recognized on a USB-powered small-sized prototype. The chosen doped polymers are easy and safe to process, particularly sensitive to water, but successfully discriminate moist from acetone and ethanol vapors. Diversifying the nature of the doping metal salts increases the prototype receptive field and allows better identification of the nature of volatile chemical compounds. This study contributes to identifying relevant materials to be processed in an easy way on electronic circuits to implement machine olfaction in portable IoT systems, by the use of various mildly doped conducting polymers. By its statistical approach, this study also raises several hypotheses on the physical origin of the sensitivity of a specific class of polymer which transports both electrons and ions. The development of such cost-effective platform allows effectively to compare the performances of multiple materials in a reproducible and systematical way, to ease multi-material performance screening supported by multivariate data analysis.
Abstract For terahertz (THz) applications, this paper presents a novel approach to temperature regulation in Dual-Channel Double-Gate High Electron Mobility Transistors (DCDG-HEMTs). The continuous scaling of high-frequency semiconductor devices introduces significant self-heating effects that can reduce performance and reliability. To tackle this challenge, we investigate how an optimal drain-side recess length (Lrd) affects heat dissipation, electrostatic control, and RF performance. The proposed device uses a dual-channel structure based on In0.7Ga0.3As/InAs to enable enhanced carrier mobility and improved heat dissipation routes. By combining a robust Schottky barrier with a buried metal gate, localized heating is prevented and thermal resistance is further reduced. Numerical simulations using the Sentaurus TCAD tool demonstrate significant improvements in important performance metrics, including a peak transconductance of 4.77 mS/μm, a maximum drain current of 2.203 mA/μm, and high-frequency capabilities with fT of 810 GHz and fmax of 900 GHz. Our findings indicate that Lrd's precision engineering maintains electrical performance while enhancing heat distribution and lowering thermal deterioration. The results validate the viability of the proposed DCDG-HEMT structure for next-generation THz applications, where efficient heat management and fast operation are essential.
Personal safety systems increasingly rely on artificial intelligence for hands-free distress detection, yet most existing solutions depend on cloud-based processing, raising privacy concerns and requiring stable internet connectivity. This paper presents a complete edge-based danger detection system implemented on the STM32 B-U585I-IOT2A microcontroller, which recognises a user-defined spoken phrase entirely on-device and triggers an emergency SMS alert via a GSM SIM900A module. The system employs a lightweight depthwise separable CNN (60.42 kB Flash, 144.66 kB RAM) that detects the keyword “happy” spoken three times within a 10-second window, achieving 73.3
The operational reliability of isolated mini-grids is constrained by voltage instability arising from weak networks, high renewable energy penetration, and evolving rural demand. Demand Response (DR) emerges as a cost-effective and flexible approach to stabilising such systems by reshaping electricity consumption. This study proposes a conceptual Voltage-Centric Demand Response (VCDR) framework, derived from a structured critical review, to examine how demand-side flexibility can contribute to voltage stability in isolated mini-grids. It employs simulation and field studies to evaluate DR control logics, load classes, system structures, and reported performance gains. Although modelling results predominantly support cost savings, reliability improvements, and demand control, a notable gap persists between the theoretical foundations of voltage-centric DR and its practical implementation in mini-grids. Key challenges remain underexplored, including communication architectures, socio-technical constraints, and the co-design of renewable energy systems. Emerging technologies such as artificial intelligence, edge computing, blockchain, and digital twins are identified as potential enablers of scalable and resilient DR systems. Conceptually, the study proposes a hierarchical control framework and an adaptive DR algorithm; however, these lack empirical validation in real-world community settings. While DR has the potential to stabilize voltages in isolated mini-grids, its practical implementation remains fragmented and at an early stage, indicating significant research and deployment gaps. Restates demand response as a voltage control measure with a Voltage-Centric Demand Response (VCDR) paradigm specific to isolated, high R/X mini-grids. Integrates simulations, experimental, and field results to demonstrate how voltage-constrained demand modulation can be more effective than supply-only control when using rural infrastructure. Identifies decentralised and hybrid control architectures as the most resilient and scalable pathways for voltage stability in inverter-dominated mini-grids. Discloses socio-technical/governance barriers as driving influences on the performance of voltage-centric demand response in the real world.
This work presents GaN-on-Si metal–insulator–semiconductor high-electron-mobility transistors (MIS-HEMTs) featuring a quaternary InAlGaN barrier and gate field plate (GFP) for power device applications. The proposed device has achieved a 40 _ON ) compared to conventional AlGaN/GaN HEMTs, due to a high carrier density of ∼ 1.9 × 10^13cm^-2, high mobility of 1540 cm^2 /V·s. These enhancements yield a low specific ON-resistance (R _ON,sp ) of 2.27 m Ω · cm ^2 . The device also exhibits excellent breakdown performance, with a drain breakdown voltage of ∼ 950 V without Gate Field Plate (GFP) and >1500 V with GFP optimized at 4 μ m, enabled by a low sheet resistance (R _sh ) of ∼ 215 Ω / □ . Furthermore, thermal reliability is confirmed by a minimal threshold voltage (V _TH ) of ∼ 0.4 V and a variation of approximately 20 _D,max ) at 150 ^∘ C. Moreover, this study attains a state-of-the-art achievement with a tradeoff of a high device figure-of-merit (FOM) on BV _DS^2 /R _ON, sp of ∼ 1100 MW/cm ^2 . In addition, we have also carried out the Long-term reliability under positive and negative bias temperature instability (PBTI/NBTI) stress tests under the specific bias of V _GS =10 V, V _DS =0 V and V _GS = -30 V, V _DS =0 V, respectively. It reveals that all GFP-equipped devices exhibit smaller V _TH shifts under PBTI, while NBTI induces a slightly higher degradation, indicating distinct charge-trapping mechanisms in both processes.
Integrating biosensors into everyday garments offers a non-invasive pathway to continuous physiological monitoring. This review organises recent advances by garment type–covering headwear, eyewear, footwear, wristwear, vests, and mouthguards–assessing sensing mechanisms, fabrication materials, and validation rigour for each platform. Studies are stratified by validation level, from in vitro bench work to controlled human trials, and a cross-platform suitability matrix maps each garment category to six clinical domains: chronic disease management, neurological monitoring, mental health, sports and fitness, occupational safety, and rehabilitation. Materials such as biofunctional hydrogels, MXene films, and conductive textiles have enabled architectures ranging from EEG-integrated caps and glucose-sensing contact lenses to thermometric diabetic socks. Despite considerable engineering progress, obstacles in clinical validation, device durability, cohort representativeness, and data security remain, and each is examined critically across all six platforms.
Effective service restoration in modern Active Distribution Networks (ADNs) is challenged by the high penetration of volatile Distributed Energy Resources (DERs) and stringent thermal constraints. This paper proposes a novel multi-objective, security-constrained restoration framework that balances maximum load recovery with real power loss minimization and network thermal safety. Unlike conventional methods relying on linearized models, this study employs a full Alternating Current (AC) power flow formulation to ensure physical feasibility in stressed post-fault states. The optimization is driven by the Artificial Protozoa Optimizer (APO), a bio-inspired metaheuristic tailored for non-convex switching configurations. The framework was validated on modified IEEE 33-bus and IEEE 123-bus systems. For the IEEE 33-bus system, the APO identified a security-constrained optimal configuration that enhanced thermal operating margins while maintaining reliable load restoration. The obtained solution reduced total system losses by 42.1 ≈ 10^-6 ) and computational speed ( <9 seconds), proving its suitability for real-time distribution management systems.
Accurate direction of arrival (DOA) estimation is critical for effective source localization in radar, sonar, and wireless communication systems. However, conventional Minimum Variance Distortionless Response (MVDR) beamformers suffer significant performance degradation due to steering vector mismatches arising from DOA errors, noise uncertainty, and dynamic interference. To address these limitations, this study proposes a Robust Optimum Diagonal Loading (ODL) beamformer in which the diagonal loading factor is analytically derived from steering vector mismatch and noise variance, enabling adaptive robustness without empirical tuning. This research introduces and evaluates a Robust Optimum Diagonal Loading (ODL) beamformer to mitigate these challenges, employing a Uniform Linear Array (ULA) for a detailed comparative analysis with MVDR. The investigation encompasses a wide range of operational conditions, including array sizes from 5 to 30 elements, azimuth angle variations, noise power levels from 0.001 to 8, and DOA mismatch errors up to ± 2°. Simulation results reveal that ODL consistently surpasses MVDR by preserving high-fidelity main lobe alignment, restricting output power fluctuations to ≤ 1.8 dB, and maintaining robust Signal-to-Interference-plus-Noise Ratio (SINR) across diverse elevation, azimuth, and array scaling scenarios. Notably, ODL achieves SINR enhancements of 20–40 dB under mismatch conditions and demonstrates stability in low-noise environments where MVDR performance deteriorates. Overall, the proposed ODL beamformer provides a robust and computationally efficient framework for DOA estimation and adaptive beamforming under non-ideal operating conditions. These results establish ODL as a promising candidate for practical array signal processing applications requiring resilience to steering errors, interference, and noise uncertainty. Future research will explore the extension of ODL to wideband beamforming and the integration of data-driven techniques for real-time optimization. Analytically derived mismatch-aware diagonal loading beamformer for robust DOA estimation. The proposed ODL method maintains stable SINR and main-lobe alignment under steering vector errors. Robust performance across varying array sizes, noise levels, and interference conditions. The ODL beamformer outperforms conventional MVDR in diverse conditions.
LoRa (Long Range) is widely adopted for low-power, long-range Internet of Things applications, yet its strict payload and duty-cycle constraints pose challenges for secure data transmission. With the growing integration of Edge AI for on-device inference and real-time decision-making, these constraints become even more critical, as intelligent processing at the edge often requires transmitting compressed and encrypted data efficiently. This study investigates the combined impact of data compression and encryption on the efficiency, energy consumption, and reliability of image transmission at the LoRa physical layer. A Variational Autoencoder was employed to compress high-resolution environmental image into compact binary representations suitable for LoRa transmission. The compressed data were evaluated under three configurations: compression only, compression with lightweight encryption using ChaCha20, and compression with hybrid post-quantum security using CRYSTALS-Kyber key encapsulation combined with ChaCha20 encryption. Experiments were conducted using real LoRa hardware under controlled laboratory conditions, with each configuration repeated 100 times. Results show that lightweight encryption introduces negligible overhead in airtime and energy consumption, maintaining performance comparable to the unencrypted baseline. In contrast, the post-quantum configuration incurs substantial airtime and transmission energy overhead due to key encapsulation ciphertexts, increasing fragmentation and occasionally affecting reliability. These findings demonstrate that compression enables security-ready LoRa transmissions while highlighting the current limitations of post-quantum cryptography for ultra-low-power LoRa deployments.
