This study aims to develop a low-disruption prefabricated strengthening method to improve the seismic performance of existing masonry walls by externally attaching precast concrete panels. Four masonry wall specimens are tested under quasi-static cyclic loading, including one unstrengthened control wall and three strengthened walls with grouting, shear-key, and combined grouting and shear-key connections. Failure mode, crack development, hysteretic response, energy dissipation, displacement ductility, bearing capacity, and lateral stiffness are evaluated. Results show that the strengthened specimens increase ultimate bearing capacity by 5.92, 5.70, and 5.46 times, respectively, relative to the control specimen, and lateral stiffness by 2.76, 1.94, and 2.30 times, respectively, compared with the control specimen. Grouting improves capacity and stiffness, while shear keys enhance ductility but induce severe local damage. The combined connection provides better deformation compatibility, crack control, and a more balanced seismic performance.
This study aims to improve resonant-frequency tracking in medium-frequency inverters subjected to switching harmonics and load-parameter variations. The proposed method combines recursive least squares (RLS) feedforward estimation with dual second-order generalized integrator phase-locked loop (DSOGI-PLL) feedback. The RLS algorithm identifies the equivalent resistance and inductance online and calculates the resonant-frequency feedforward value from the estimated inductance. In parallel, two SOGIs extract the fundamental components of the total voltage of the resistor-inductor-capacitor (RLC) circuit and the resonant current, and a normalized cross-product phase detector generates the feedback correction. The method reduces frequency drift after a phase-shift step by approximately 93% and shortens the convergence time after an inductance step from 9.29 ms to 2.32 ms. These results demonstrate that the proposed method can improve resonant-frequency tracking accuracy and dynamic response under phase-shift modulation and load-parameter variations.
This study proposes an efficient you only look once (YOLO)v11s-based framework for real-time oil palm fresh fruit bunches (FFB) detection, addressing the high computational cost that typically limits YOLO-based detectors on edge devices. The framework integrates structured detection-path pruning (SDPP) and 8-bit Integer (INT8) post-training quantization. SDPP analyzes the object-scale distribution of FFBs in UAV imagery to identify the detection paths that contribute most effectively to the target object characteristics. Based on this analysis, redundant detection branches with limited contribution to detection performance are removed, reducing the computational burden while preserving accuracy. The pruned model is subsequently quantized to INT8 precision to further improve inference efficiency on resource-constrained devices. Experimental results show that SDPP reduces computational complexity by 53.5% while maintaining an mAP@0.5 of 0.987. With INT8 quantization, the optimized model achieves 30.42 FPS on the NVIDIA Jetson Orin Nano, demonstrating its potential for real-time FFB detection on edge devices.
This study aims to investigate the dynamic shear response and damage mechanisms of reinforced concrete (RC) columns under soft impact. High-fidelity finite element (FE) models are developed and validated against pendulum impact tests, followed by simulations. The results show that, unlike a hard impactor, a soft impactor absorbs kinetic energy through its deformation, which mitigates the dynamic effect and thus decreases the probability of one-way/punching shear failure in RC columns. Dynamic shear force in RC columns increases with increasing impactor stiffness under low-velocity impact, but decreases beyond a critical stiffness under high-velocity impact. This trend results from the transition from one-way shear to punching shear. The inertial force of shear plug is mainly affected by impactor stiffness and impact velocity, whereas other factors exert negligible influence. The punching shear resistance model proposed for hard impact demonstrates good applicability for evaluating the one-way/punching shear failure of RC columns under soft impact.
This study aims to develop a deep learning framework for intelligent fault detection, classification, and multi-level localization in rooftop photovoltaic (PV) systems. The proposed approach uses a hybrid CNN–LSTM architecture that combines a one-dimensional convolutional neural network (1D-CNN) to identify local spatial patterns with long short-term memory (LSTM) units to capture long-term temporal dependencies in time-series data. The framework was evaluated on a multivariate dataset collected from a grid-connected experimental rooftop PV prototype under various fault scenarios. Experimental results demonstrate that the hybrid model significantly outperforms standalone 1D-CNN and LSTM architectures across all evaluation criteria. Specifically, the proposed framework achieved classification accuracies of 99.96% for fault type identification, 98.76% for string-level fault localization, and 99.21% for module-level fault localization. These findings highlight the effectiveness and robustness of the hybrid framework for real-time monitoring and reliable fault diagnosis in photovoltaic installations.
