In order to enhance the performance of massive multiple-input multiple-output (MIMO) systems, the downlink channel state information (CSI) must be efficiently compressed and fed back to the base station in a frequency-division-duplex mode. The deep-learning-based CSI feedback networks can learn more complicated patterns in the data as compared with the traditional compressed-sensing techniques. Nevertheless, many of the current methods are founded on the convolutional neural network (CNN) modules that were initially developed to act on images and might not make full use of relevant channel and spatial information. The suggested model, wavelet transform convolution and spatial channel self-attention-channel network (WTSCNet), is a combination of wavelet-transform convolution and attention mechanisms. The encoder also has access to wavelet-transform convolution to generate multi-resolution features, enabling it to perform better in both space information and local details at a reduced computational cost. On the decoder part, it is presented as a spatial channel self-attention-channel reconstruction block (CARBlock) which combines multi-scale residual reconstruction with spatial-channel attention to improve the recovery of CSI. Experiments prove that WTSCNet enhances the indoor reconstruction precision by 6.41 dB, compared with CRNet, at a compression ratio of 1/4. In comparison with the Transformer-based TransNet+, WTSCNet requires about 9.9 million floating-point operations less than TransNet+ but offers a 0.28-dB improvement in the indoor normalized mean squared error (NMSE) at the same compression ratio. The simulation results indicate that WTSCNet offers a good accuracy-complexity trade-off between CSI compression and feedback.

The rapid advancement of the ultra-wideband (UWB) technology has driven the demand for high-performance antennas with compact size, flexibility, and efficient radiation characteristics. This paper presents a comparative performance analysis of a conventional flame retardant (FR4)-based antenna prototype and a polyimide (PI)-based wearable antenna designed for UWB applications. The antennas are fabricated and evaluated in terms of key performance metrics, including reflection coefficient, impedance bandwidth, gain, and radiation pattern, for wearable applications. The results demonstrate that the FR4-based antenna exhibits stable performance and ease of fabrication, and offers superior flexibility, conformability, and reduced weight, making it highly suitable for body-worn applications. Furthermore, the wearable antenna maintains a wide impedance bandwidth with compactness. These findings provide valuable insights into the trade-offs between rigid and flexible antenna materials, highlighting the potential of PI substrates for the next-generation UWB wearable communication systems.

Energy auditing is critical for sustaining the performance and reliability of solar power plants, yet current practices are fragmented and lack a standardised approach for industry application. This study introduces the SMART (specific, measurable, achievable, relevant, and time-bound) framework, a first-of-its-kind, standards-aligned methodology that unifies verified performance equations, benchmark thresholds, and mitigation strategies drawn from internationally recognised standards and peer-reviewed literature. Unlike conventional audits that assess isolated metrics, the SMART framework integrates data acquisition, performance evaluation, loss analysis, and corrective planning into one reproducible process. It benchmarks key performance indicators, including system efficiency, specific yield, performance ratio (PR), and temperature and mismatch losses, to provide clear, actionable insights into plant operation. Designed for direct industry use, the framework is practical, scalable, and adaptable across diverse plant sizes and configurations. By transforming scattered academic knowledge into a structured, field-ready toolkit, this study establishes a groundbreaking pathway for transparent and replicable solar power plant auditing.

Timing variability induced by device manufacturing fluctuations has become a pervasive and critical challenge for low-voltage logic circuits, which severely impairs the timing margin and operational accuracy of core peripheral circuits such as decoders. In sub-1-V low-voltage applications, gate-level delay discretization in decoders will be magnified into wordline arrival-time mismatches, posing insurmountable obstacles to the realization of low-power and high-reliability integrated systems. To address this issue, this paper proposes a closed-loop paradigm integrating device modeling-circuit implementation-robustness evaluation-feedback design for ferroelectric negative-capacitance field-effect transistors (NCFETs) in low-voltage logic applications. First, under a strict charge-conservation framework, the Landau-Khalatnikov (LK) ferroelectric dynamics is coupled with the MIT virtual-source (MVS) channel conduction model to construct a simulation program with integrated circuit emphasis (SPICE)-compatible compact model for NCFETs that can characterize device-to-device (DTD) variations. Second, DTD variations are explicitly introduced to simulate process fluctuations in ferroelectric layer thickness (t_FE), and the resulting statistical perturbations are propagated from the device level to the gate-level 2−to−4 decoder circuit. Finally, with the 2−to−4 decoder as a case study, the propagation delays of all output signals are quantified, and a “DTD-centric” feedback path aimed at design-window convergence is formulated. The results show that within a process window of approximately t_FE ≈ 10 nm, the proposed model successfully reproduces experimentally comparable I−V characteristics; the NCFET-based 2−to−4 decoder maintains correct functionality under low voltage drain drain V_DD, and the coefficient of variation (C_V) of its propagation delay is controlled within 3%–7%, exhibiting a substantial design optimization window for low-voltage logic design.

In an industrial environment, sound is generally an important source of data that results from the movement and vibration of machines during operation. Each machine operation condition often generates a specific sound, for example, a familiar normal working sound, an unfamiliar sound, a breaking sound, or even the absence of sound, which may indicate a stop. This paper proposes a supervised convolutional neural network (CNN) model with early stopping and learning rate (LR) scheduling to achieve a more accurate and cost-effective trained deep learning (DL) model for machine sound classification. The proposed model is trained and evaluated using the MIMII dataset, which contains unbalanced real-world data and signal-to-noise ratios (SNRs). To ensure a fair (i.e., representative) data splitting strategy and robust coverage across all hierarchical structures of the MIMII dataset, this study adopts a hierarchical multi-stage stratified splitting protocol incorporating pre-split shuffling to prevent order bias. The synthetic minority oversampling technique (SMOTE) is used to balance the training set after feature extraction. Compared with state-of-the-art works implemented on the same dataset, the proposed model achieves optimal performance in terms of accuracy, area under the curve (AUC) , F1-score, precision, recall, Matthews correlation coefficient (MCC), and Kappa without noise processing methods at an SNR of 6 dB, and maintains robust, competitive performance under severe noise conditions (0 dB and −6 dB).

Reliable state-of-health (SOH) estimation depends not only on the nonlinear mapper but also on whether its inputs are compact, informative, and physically interpretable. We develop an equivalent circuit model (ECM)-guided feature-selection strategy that uses fractional-ECM parameters offline as physical anchors for charging-response and incremental capacity analysis (ICA) features. The predictive evaluation comprises 4259 cycle samples from ten commercial nickel–cobalt–aluminum (NCA) cells tested at 25 °C, 35 °C, and 45 °C under a 0.5 C protocol. A 13-dimensional (13D) full input is compared with five equal-cardinality 8-dimensional (8D) schemes using gated recurrent unit (GRU) and temporal convolutional network (TCN) nonlinear mappers. Ten-fold leave-one-battery-out (LOBO) validation provides the primary evidence. Among the 8D schemes, the Pearson correlation coefficient (PCC-8D) gives the lowest GRU macro root-mean-square error (RMSE; $0.005272\pm 0.002064$). Physically anchored ECM-8D gives the lowest TCN macro RMSE ($0.005882\pm 0.003804$). ECM-8D is therefore competitive rather than universally most accurate. The two retained charging-time ratios are deterministically related, and 7D removal sensitivities reveal battery- and architecture-dependent effects. ECM fitting is required only during offline feature design; inference uses charging, ICA, and temperature variables. These results support a bounded route to interpretable SOH input construction within the tested NCA protocol, not cross-chemistry or wider-regime generalization.

Securing network infrastructure against cyberattacks is essential in rapidly evolving digital landscapes. This paper proposes an AI-based real-time intrusion detection system (IDS) for wireless sensor network (WSN), cloud computing (CC), and Internet of things (IoT) environments. The framework employs classical machine learning (ML) models, including decision trees (DTs), K-nearest neighbors (KNN), and logistic regression, optimized using the energy valley optimization (EVO) algorithm on the Network Security Laboratory − Knowledge Discovery in Databases (NSL-KDD) dataset. Feature selection (FS) reduced the number of input features from 42 to 18 without compromising the detection performance. The system achieved high accuracy rates of 98.95% with a decision tree, 98.47% with KNN, and 88.84% with logistic regression, along with high precision, recall, and F1-scores. In addition, the training and testing times were significantly reduced, supporting real-time deployment. Class imbalance was addressed through downsampling to ensure balanced detection across attack types. The proposed model offers a scalable, low-latency, and accurate intrusion detection solution aligned with modern cybersecurity requirements.