Cheng Shiu University

Background This work explores the design of a ternary heterojunction photocatalyst, g-C3N4/TiO2/ZrO2, synthesized with varying proportions of graphitic carbon nitride (TZ1-TZ4). The motivation was to improve visible-light absorption and photocatalytic efficiency for wastewater treatment, particularly targeting the degradation of organic dyes under sunlight. By integrating TiO2, ZrO2, and g-C3N4 through interface engineering, the study aimed to create a durable and environmentally friendly photocatalyst. Methods The composite was fabricated using a chemical wet method, producing nanoparticles uniformly anchored on g-C3N4 nanosheets. Structural, morphological, and optical properties were characterized using XRD, XPS, FESEM, HRTEM, and UV-Vis spectroscopy. XRD confirmed anatase TiO2, monoclinic ZrO2, and graphitic g-C3N4 phases, with an average particle size of similar to 27 nm. Bandgap analysis revealed a decrease from 3.2 eV (TZ1) to 2.9 eV (TZ4). Photocatalytic activity was evaluated by degrading crystal violet dye under sunlight for 180 min, while radical-scavenging experiments identified hydroxyl radicals and photogenerated holes as the main reactive species. Significant findings The ternary heterojunction exhibited enhanced photocatalytic performance, with TZ4 achieving the highest efficiency of 79% due to increased g-C3N4 content and improved visible-light absorption. Radical-scavenging studies confirmed hydroxyl radicals and holes as the dominant contributors to dye degradation. After three cycles, the catalyst retained most of its activity, demonstrating good stability. Overall, the TiO2-ZrO2/g-C3N4 heterojunction proved to be an efficient, durable, and eco-friendly photocatalyst suitable for wastewater treatment applications.

This paper develops a hybrid adaptive control framework for improving low-voltage ride-through (LVRT) performance and transient stability in grid-connected hybrid renewable energy systems. The proposed approach combines an adaptive least squares recursive Chebyshev fuzzy neural network (ALTSRCFNN) with super-twisting sliding mode control (STSMC) to achieve robust voltage regulation under grid faults and parameter uncertainties. A Static Synchronous Compensator (STATCOM) is integrated to provide dynamic reactive power support during voltage sags. The ALTSRCFNN is employed to estimate system nonlinearities and update control gains online, while the STSMC ensures finite-time convergence and mitigates chattering during severe disturbances. The effectiveness of the proposed controller is evaluated in MATLAB/Simulink under symmetrical and asymmetrical grid fault conditions and is benchmarked against conventional proportional-integral-derivative (PID) and feedforward neural network controllers. Simulation results demonstrate that the proposed method reduces voltage recovery time by approximately 25% and improves transient voltage stability by about 15% while satisfying LVRT grid code requirements. These findings confirm the suitability of the proposed control strategy for fault-tolerant operation of large-scale hybrid renewable power systems.

In pursuit of high-performance and energy-efficient dielectric ceramics for next-generation wireless communication systems, this work examines the effect of Co2+ substitution for Mg2+ on the microstructure and microwave dielectric properties of (Mg0.95Zn0.05)TiO3 ceramics. A series of [(Mg1-xCox)0.95Zn0.05]TiO3 compositions (x = 0.1-0.4) was synthesized via solid-state reaction and characterized by synchrotron XRD, Raman spectroscopy, SEM, and dielectric measurements. The optimal composition, [(Mg0.8Co0.2)0.95Zn0.05]TiO3, sintered at 1350 degrees C for 2 h, achieved epsilon r = 18.42, Qf = 190,000 GHz, and tau f = -51 ppm/degrees C, representing a 16.2% Qf enhancement and a 43.7% reduction in overall thermal budget compared with literature benchmarks. These results are attributed to improved densification, increased lattice polarizability, and a stable phase composition. The findings demonstrate a viable route toward low-loss, energy-efficient microwave ceramics suitable for 5G and future communication devices.

Traumatic impacts resulting from vehicular collisions, sports activities, occupational accidents, and falls frequently lead to severe injuries of the cranium, brain, and cervical spine, posing a significant global public health burden. Accurate injury prediction requires high-resolution anatomical modeling supported by advanced sensing technologies. In this paper, we presented a comprehensive biomechanical impact analysis using a high-fidelity integrated cranial-brain-cervical (CBC) finite element model, with particular emphasis on the application of optical sensor systems in model development. The CBC model was established through reverse engineering using the Breuckmann SmartSCAN 3D system, which integrates industrial-grade CMOS/CCD imaging sensors to capture high-precision surface geometry. A commercially available 3B Scientific C18 five-part brain anatomical model was digitized, and the reconstructed geometry was further refined on the basis of a prior validated modeling work. The sensor-acquired data ensured high spatial resolution and geometric fidelity, directly enhancing computational accuracy. Modal analysis was conducted to determine the fundamental natural frequencies and mode shapes, followed by impact simulations under both damped and undamped conditions. Injury severity was quantified using the head injury criterion (HIC), peak linear acceleration, and velocity in accordance with standards established by the National Highway Traffic Safety Administration. Simulated HIC and peak acceleration results showed strong agreement with published validation data, confirming the predictive reliability of the sensor-informed CBC model. The proposed framework demonstrates how advanced 3D optical sensing can support biomechanical modeling, injury assessment, and the sensor-integrated design of next-generation protective equipment.

This study empirically validates the Human–AI Co-Intelligence Learning (HAIL) framework, which explains how university students convert generative AI support into autonomous, ethical, and effective learning. Using a mixed-methods design, data were collected from 320 undergraduates in Taiwan who had integrated ChatGPT or similar tools into their coursework. Partial Least Squares Structural Equation Modeling (PLS-SEM) revealed that AI Support significantly enhanced Trust Calibration, which in turn strengthened Self-Regulated Learning (SRL) and promoted Learning Agency and Outcomes. SRL partially mediated the trust–agency relationship, while Ethical Awareness and Cultural Orientation moderated these effects in opposite directions. Thematic interviews corroborated the quantitative findings, showing that students developed reflective reliance on AI—balancing efficiency with integrity—while hierarchical cultural norms constrained autonomy. The validated HAIL model extends self-regulation and trust theories into AI-mediated education, positioning trust calibration as the cognitive bridge between algorithmic support and human autonomy. The findings provide theoretical and practical guidance for designing human-centered, ethically sustainable, and culturally responsive AI learning environments in higher education.